# Demo
Source: https://docs.hazel.ai/demo
February 13, 2026
In this Office Hours, you’ll see a full demo of Hazel:
* **Ask Hazel:** Our powerful chat interface that meets industry security and privacy requirements.
* **Meeting Assistant:** Hazel helps with your meetings, from prep to debrief.
* **Email Assistant:** Drafts, summarizes, and helps you respond faster in your tone.
# Best Practices
Source: https://docs.hazel.ai/guides/ask-hazel-chat/best-practices
### Ask Anything
Hazel is designed to handle any type of question — from broad to highly specific. Keep in mind that the more context you provide, the more tailored and detailed Hazel’s answers will be.
**Best practices:**
* Even vague questions will return results, but context (client names, dates, goals) helps Hazel dig deeper and deliver sharper answers.
* Hazel can pull across systems — CRM, Emails, Previous Meetings, Web Search — so you don’t need to worry about where the data lives.
* Don’t worry about perfect phrasing. Hazel understands natural language, so speak as you would to a colleague.
* Use action-oriented prompts like *“Draft,”* *“Summarize,”* *“List,”* or *“Prepare”* when you want Hazel to create something for you.
Use Hazel like you would an assistant. If you only remember part of a detail — a client’s first name, or that a conversation happened “last month” — that’s enough to start. Hazel will fetch results, and you can refine with follow-ups.
### Guide the Conversation
Think of Hazel like an assistant who remembers the flow of your instructions and connects details across systems. That means you don’t have to repeat client names or details every time — you can ask follow-ups naturally.
**Best practices:**
* Start with a broad request, then narrow down with follow-ups.
* Layer your questions step by step — Hazel keeps track of who and what you’re talking about.
* Be clear on the format you’d like — an email, bullet points, or a meeting agenda.
Hazel shines when you chain together multiple steps — pulling data, analyzing it, and drafting outputs — all in the same conversation. This turns Hazel into a true assistant.
### Build an Additional Layer of Context
Hazel doesn’t just pull from your existing systems — you can also **add your own notes, tables, and files directly into Hazel** to give it richer context. This makes Hazel smarter and more personalized.
**Best practices:**
* Add **files** in the Chat and have Hazel analyze and provide context.
* Add **additional notes or tables** within the **Household section** of Hazel. Use this extra context to power chats, prep, and agendas going forward.
* Think of it as giving Hazel additional information about a client that isn’t stored anywhere else but in your head.
The more client-specific context you add, the more Hazel has to work with. As Hazel receives new data — from emails, CRM, or meetings — it will add it to their client profile. For additional thoughts or context, add your own notes for depth. Over time, Hazel becomes a single source of truth you can rely on for deeper insights and faster prep.
### Scope questions to your own book
If your firm assigns advisor roles in your CRM (Lead Advisor, Senior Advisor, Paraplanner, CSA, etc.), you can ask Hazel to limit results to households where a specific person holds that role. This is useful when you only want to see clients you own, or when you’re reviewing another advisor’s book.
**Best practices:**
* Reference yourself with *“my,”* *“mine,”* or *“me”* — for example, *“List my Lead Advisor households”* or *“How many clients am I the Lead Advisor for?”*
* Reference a colleague by name or email — for example, *“Show households where Jennifer Akhlaghi is the Lead Advisor.”*
* Combine the role and person filters — for example, *“Which of Jennifer’s Senior Advisor clients haven’t had a meeting in 6 months?”*
* If you’re not sure which role labels your firm uses, just ask: *“What advisor roles are used across our households?”*
Advisor-role filtering reads from your CRM, so the role names available depend on how your firm has them set up (e.g. Redtail, Wealthbox, Salesforce). If a role doesn’t appear, confirm it exists on your households in your CRM.
# Overview
Source: https://docs.hazel.ai/guides/ask-hazel-chat/overview
Hazel is designed to help you think, plan, and achieve outcomes with speed and clarity, so you can stay engaged and be proactive. Instead of spending time digging through files, emails, and notes, Hazel anticipates what you need in seconds.
Your accumulated data is the knowledge that Hazel uses to surface what matters most, right when you need it — whether that’s prepping for a client meeting, recalling an interaction, or prioritizing an email.
Think of Hazel as your partner that helps you move from “what do I do?” to “here’s what’s next”, all from a simple nudge.
## How Hazel can help you:
### Writing & Communication
* Emails: I can prioritize your client emails and draft intelligent responses using available data.
* Summaries: Paste articles or reports or ask about a previous meeting, and I can turn them into a digestible summary.
* Follow-ups: I can generate reminders or recap emails for teams and clients.
### Research & Analysis
* Document Analysis: Drop in files (like estate documents or tax returns), and I can summarize, highlight observations, or help you break down next steps.
* Quick Explanations: Ask me to break down a new concept in simple terms, or compare options (e.g., tax, political policies, financial planning strategies).
* Industry Insights: I can summarize news or best practices from trusted sources.
### Brainstorming & Creativity
* Idea Generation: Whether it’s marketing ideas or prospecting strategies, I can be your thought partner.
* Problem-Solving: You can describe a challenge, and I’ll suggest different approaches while referencing your clients details on file.
* Content Outlines: For presentations, meeting agendas, or financial planning strategies, I can give you a starting point.
### Planning & Organization
* Daily Priorities: I’ll prioritize your day and update you as items arise.
* Project Outlines: I can help create step-by-step plans or roadmaps.
* Templates: Need a framework for reports, agendas, or documentation? I’ll draft it.
# Guardrails
Source: https://docs.hazel.ai/guides/financial-planning/assumptions/guardrails
Your firm's never-do list: the rules nothing overrides. What ships turned on, how to write your own, and where a plan records them.
Your firm has a handful of things it will never put in a plan, whatever the client's situation or however a request is phrased. Write those down as guardrails, so no plan anyone at your firm builds can cross them.
## Where you set them
Open **Settings → Planning** and select **Base template**, the first entry under **Plan templates**. The never-do list sits directly below **Hazel instructions**.
* **Adding a rule.** Select **+ Add a rule**, which shows "Never recommend variable annuities" as the shape to aim for. Type your rule and press Enter, or click away and it saves anyway. Escape drops what you typed.
* **Suggestions.** Beside that button sit four common rules. Select one and it joins your list, then drops out of the row.
* **Removing a rule.** Each row carries an X at the right. A row cannot be edited in place, so to reword a rule, remove it and add the new wording.
The four suggestions:
* Never recommend variable annuities
* Never push a conversion into a higher IRMAA tier
* Never assume Social Security benefits are cut
* Never recommend leverage or margin
Every change saves as you make it.
## The two rules that ship
Your firm starts with two, and both apply until you remove them:
* **"Never recommend a specific named security or ticker."** A plan can still recommend an allocation change or a category of holding, but it will not name an instrument.
* **"Never propose a Roth conversion that pushes the client into the next IRMAA tier."** IRMAA is the income-related surcharge that raises a client's Medicare Part B and Part D premiums, and a conversion that clears a tier line can cost more in premiums than it saves in tax.
You can remove either one.
## How guardrails differ from instructions and assumptions
| | Instructions | Assumptions | Guardrails |
| --------------------------- | -------------------------------- | -------------------------------------- | ---------------------------- |
| **What it is** | Guidance Hazel weighs | The numbers the planning math runs on | Rules |
| **Can a persona change it** | Its own list replaces the firm's | Some settings, one at a time | It can **add**, never remove |
| **Can one plan change it** | In effect, by asking Hazel | Yes, inside the range your firm allows | **No** |
Hazel treats a guardrail as a hard rule, declines a request that would break one, and names the rule that request ran into. They are rules Hazel follows rather than a mechanical block, so if you ever see a plan cross one, tell whoever manages your firm's planning settings.
## Writing a good one
"Never recommend surrendering an in-force life policy."
"Never model borrowing against a retirement account."
"Be conservative." A preference. Put it in instructions.
"Use a 90% success threshold." A number. Put it in Solver assumptions.
"Always include an insurance section." Structure. Build it into the plan design.
Test each one:
1. **Would you accept an exception in any client situation?** If yes, it is guidance. Write it as an instruction.
2. **Could a colleague tell whether a plan broke it?** If not, it is too vague to hold.
Keep the list short. Two to six rules is usually enough, and past that some entries are really instructions or assumptions.
## Personas and single plans
Guardrails are firm-wide. Nothing on a persona drops one of yours: whichever persona Hazel matches a household to, every firm rule still applies. For a rule you only want for one kind of client, write it in that persona's **Hazel instructions**, which Hazel weighs as strong guidance rather than an absolute limit.
## Checking which rules a plan was built under
Open the plan's **Appendix** and read the **Applied settings** block. Its **Guardrails** row lists the rules that applied, and **Set by** names where each came from: your firm, or the matched persona where that persona added its own. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
The record travels with the plan version, so an older version shows the list as it stood then. A rule you add today applies to plans built from now on. To bring an existing plan under it, ask Hazel in chat to update the plan, or use **Regenerate** on the household's **Financial Planning** tab.
## Where to go next
* [Instructions](/guides/financial-planning/settings/personas/instructions): house style and preferences, not rules.
* [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions): the numbers behind every figure.
* [Base template](/guides/financial-planning/settings/base-template): the pane your never-do list lives on.
* [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from): reading the Appendix and a figure's sources.
* [What it can and can't do](/guides/financial-planning/overview/what-it-can-and-cant-do): the limits that hold whatever your guardrails say.
# Market assumptions
Source: https://docs.hazel.ai/guides/financial-planning/assumptions/market-assumptions
The return, volatility, inflation and tax figures every plan is built on: how your firm edits the set, how much detail to model, and how a plan shows which figures it used.
You are going to put a projection in front of a client, so the capital-market figures behind it should be your firm's own.
Market assumptions live in **Settings → Planning → Market assumptions**, a firm owner's screen. Like the rest of Planning settings, there is no Save button: edits save themselves.
## One set, edited in place
Your firm has one set from the moment the screen first opens, already filled in with Hazel's published figures, and every plan is built on it. This screen edits it directly: there is no draft copy and no separate step that switches a new set on.
Along the top of the screen:
* The **name** field labels the set for whoever opens the screen next.
* A line counting how many values you have edited.
* **Reset all to source**, which appears once there is something to reset and returns every row to Hazel's figure.
* **Publish to firm**, unavailable until you have changed a row. Pressing it flashes "Published" and changes nothing else, because the edit saved when you made it. An edited row wears a badge reading "Applies to all new plans when published", and it already applies.
## The four sections
| Section | What it sets | Simple | By sub-asset class |
| -------------- | ------------------------------------------ | --------------------------------------------------------- | ---------------------------------------------------------------------------- |
| **Returns** | Expected returns used to grow assets | Three rows: Equity, Fixed income, Cash & equivalents | Nine rows: four equity classes, four fixed-income classes, cash |
| **Volatility** | Variability market simulations draw on | The same three rows | The same nine rows |
| **Inflation** | General price growth on spending and goals | One row | One row, so switching changes nothing on screen |
| **Taxes** | Effective rates for tax-aware projections | One row, **Effective tax rates**, a single blended figure | **Ordinary income**, **Long-term capital gains** and **Qualified dividends** |
Return rows are nominal long-term figures, one compound annual rate per asset class, and volatility rows are annualized standard deviations. Inflation is modeled separately, so enter the nominal returns your committee agreed: a real return here removes inflation twice.
### How much detail to model
Each section carries its own **Simple** or **By sub-asset class** toggle, so the levels can differ. If you went looking for a capital-gains rate and found one blended number, the Taxes toggle is what you were missing.
Hazel's Simple Equity and Fixed income figures are the equal-weight average of that section's four sub-asset-class rows. Past that the levels are independent: each keeps its own rows, and only the rows visible at your selected level reach a plan. An edit at the level you are not showing is remembered, and still counts in the edited-values line, so that count can run ahead of the screen.
## Changing a row
Every row is a slider: drag it or nudge it with the arrow keys. The reading beside it is a display, not a field to type in. Returns, volatility and inflation move in tenths of a percent and tax rates in half points, so a figure between two steps cannot be set here, and **Reset to source** is the only way back to the exact published figure. The [Assumptions reference](/guides/financial-planning/assumptions/reference) lists every range.
An edited row says so: its provenance line changes from the source label to **Set by You**, a badge appears, and **Reset to source** sits beside it. The section header gains a **Custom** badge naming what the section is based on.
## The Monte Carlo switch
A switch labeled **Monte Carlo** sits in the header of the Volatility card and covers the whole set, whatever detail level each section is on. On, Hazel models many market paths, which is where a range of outcomes and a probability of success come from. Off, expect plans to lean on single-path figures instead of a spread. Leave it on unless you have a specific reason not to.
## What plans do with these
Select a figure that carries its workings and **Provenance and calculations** names the assumption behind it, its value, and where it came from: Hazel's default set until your firm moves that row, your firm afterwards. A figure with no workings renders as plain text and does not open.
The plan's **Appendix** lists them under **Applied settings**: a **Market scenario** row, then the rows your firm or this plan changed. The scenario row identifies the set rather than repeating the name you typed in settings. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
One plan can differ from the firm set. Open a plan from its chat and the **Plan settings** panel lists the market rows your firm has hand-edited, adjustable for that plan alone and committed with **Apply and replan**. Hand-edit nothing and that panel has no market rows to adjust. **Plan settings** is chat-only: a plan opened on the household's **Financial Planning** tab has no such control, so make the change from that plan's own chat.
Changing a figure here changes what new plans and rebuilds are built on. A plan version you leave alone keeps the figures it was built with, so nothing you have shown a client moves under you, though an older version can disagree with this screen. To bring a household onto new figures, use **Regenerate** in the top bar of the household's **Financial Planning** tab.
## Reviewing them
Once a year, with your investment committee: mid-year changes are hard to explain to a client, and there is no draft to circulate, so bring agreed numbers. Rename the set for the year you adopted it.
Then rebuild two plans you know, one near retirement and one accumulator, with **Regenerate** on each household's **Financial Planning** tab, and back out surprises with **Reset to source** or **Reset all to source**.
## Where to go next
* [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions): the planning inputs the math runs on.
* [Assumptions reference](/guides/financial-planning/assumptions/reference): every shipped figure and range.
* [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from): how a plan attributes a figure.
* [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly): changing one figure and recalculating.
# Assumptions reference
Source: https://docs.hazel.ai/guides/financial-planning/assumptions/reference
Hazel's shipped value for every planning assumption, what it drives, and which layer can change it.
You are setting up Planning settings, so you need to know what Hazel already assumes and which layer can change each value. Every default below is what a firm gets before configuring anything.
**Scope** names the layers that can change a value. **Firm** is the Solver assumptions and Market assumptions screens in Planning settings, a firm owner's surface. **Persona** is a Client persona's own value, replacing the firm's for matching households. **Plan** is one plan only. Each plan's Appendix records the winning layer in its **Applied settings** block: **Firm**, **Persona** with its name, or **This plan**.
## Probability and simulation
| Setting | Default | What it drives | Scope |
| ------------------------------ | ------------------------------------------------------ | ------------------------------------------------------------------------------------ | -------------------- |
| Success threshold | 90% | Lowest probability of success that counts as on track; below it the plan is flagged. | Firm, Persona, Plan |
| Monte Carlo simulations | 1,000 | Market paths each projection runs. More paths, more precision, slower. | Firm |
| Success-figure display ceiling | 99%, unless your firm has a different ceiling in place | Highest probability of success a plan will print. | Firm-wide, no slider |
Ranges: Success threshold 50% to 100% in steps of 1. Monte Carlo simulations 100 to 5,000 in steps of 100.
Where a projection models a figure above the ceiling, the plan prints `>99%` with a short note under the figure saying a cap applied.
## Fees and taxes
| Setting | Default | What it drives | Scope |
| ---------------------- | ------- | ------------------------------------------------------------------------------------------ | ----- |
| Advisory fee | 1.00% | Annual fee drag on assets under management, in every projection. | Firm |
| Terminal heir tax rate | 24% | Rate heirs are assumed to pay on inherited pre-tax balances, valuing the after-tax estate. | Firm |
Ranges: Advisory fee 0% to 3% in steps of 0.05. Terminal heir tax rate 0% to 50% in steps of 1.
## Decumulation strategy
| Setting | Default | What it drives | Scope |
| -------------------------------------------------------------- | --------------------------------- | -------------------------------------------------------------------------------------------------------- | ------------------- |
| Default reserve length | 3 yrs | Years of retirement spending held in safe assets, so the client never sells equities into a down market. | Firm, Persona, Plan |
| Reserve refill policy, under the Default reserve length slider | Ships with wording in it | How the reserve is held and replenished. | Firm, Persona |
| Retirement withdrawal guardrails, the one box with a heading | Ships with Guyton-Klinger wording | How spending flexes as markets move. | Firm, Persona |
| Default Roth conversion strategy | Stop under first IRMAA tier | The ceiling Hazel uses when recommending annual Roth conversions. | Firm, Persona, Plan |
| Conversion elaboration, under the Roth strategy choices | Ships with wording in it | Plain-language nuance refining the chosen strategy. | Firm, Persona |
Range: Default reserve length 0 to 10 years in steps of 1. The three free-text boxes ship filled rather than empty, and [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) quotes each word for word.
## What an advisor can change on one plan
**Plan settings** in the plan's top bar stages values for one plan, and **Apply and replan** in its footer rebuilds the plan with them. The panel belongs to the chat plan view, so from the household Financial Planning tab use **Open chat** in the top bar first.
Editable there: Success threshold, Reserve length, Roth conversion strategy (six choices), and one row for each market row the firm has hand-edited, so a firm that has hand-edited nothing gets no market rows. Withdrawal guardrails is displayed for reading, and changes at the firm or persona layer instead. A read-only row names the assumption set the plan drew on, and it reads as Hazel's default set rather than the name anyone typed on the Market assumptions screen, which is not a sign your firm's edits were ignored.
## Market assumptions
| Setting | Default |
| ----------------- | ---------------------------------------------------------------------------------------- |
| Assumption set | One live set, its name field pre-filled with "Name this assumption" until you replace it |
| Returns detail | Simple: Equity, Fixed income, Cash & equivalents |
| Volatility detail | By sub-asset class |
| Inflation detail | Simple |
| Taxes detail | Simple: one Effective tax rates row |
| Monte Carlo | On |
| Hand-edited rows | None |
All firm settings, though an advisor can change hand-edited rows for one plan. The set arrives fully populated, with nothing hand-edited yet. At the shipped granularity:
| Row | Shipped value |
| --------------------------- | ------------- |
| Returns, Equity | 7.2% |
| Returns, Fixed income | 4.8% |
| Returns, Cash & equivalents | 3.1% |
| Inflation | 2.4% |
| Taxes, Effective tax rates | 24.0% |
Switch Returns to by-sub-asset-class and nine rows appear instead: four equity classes, four fixed-income classes and cash. Rows are edited on their own, so changing one leaves the Simple figure where it was.
Volatility opens at by-sub-asset-class detail, listing all nine rows, from cash at 0.7% up to US small cap at 21.1%.
Switch Taxes to by-sub-asset-class and the blended row splits into Ordinary income at 24%, Long-term capital gains at 15% and Qualified dividends at 15%. Statutory brackets are applied as written, not a firm setting, and the card is badged with the bracket year.
Row ranges: returns 0% to 15%, volatility 0% to 30%, inflation 0% to 10%, tax rates 0% to 50%.
## Guardrails
Two hard rules ship, in the card for things Hazel must never do:
* "Never recommend a specific named security or ticker."
* "Never propose a Roth conversion that pushes the client into the next IRMAA tier."
A persona can add guardrails on top, and no persona, plan or chat request can lift one. A firm owner can remove either shipped rule from the base template.
## Layout and shape
| Setting | Default |
| ------------------------- | ----------------------------------------------------------------------------------------------------------------------------------- |
| Plan sections | Summary, Today, Goals, Getting there, Decisions, plus an Appendix reached from the bottom of the plan's nav |
| Tabs inside Getting there | Your priorities, then Cash flow, Retirement, Insurance |
| Enabled modules | Eight: Net worth & balance sheet, Emergency fund, Goal funding, Insurance, Cash flow, Debt paydown, Retirement, Monte Carlo success |
| Strictness | Flexible |
| Allow additions | On |
Every module outside those eight is off until you add it, meaning the topic gets no dedicated treatment.
**Exact** locks **Allow additions**, the switch beside the mode, off. [Strictness](/guides/financial-planning/settings/personas/strictness) covers what each of the three modes does.
## Client disclaimer
Nothing ships. The **Client disclaimer** card opens with a paragraph Hazel drafted as a preview: until you edit it, plans and PDFs carry no footer at all. Type your own wording, or edit the draft, and it renders word for word under every plan and in the exported PDF. A **Restore Hazel default** button brings the drafted paragraph back.
## Persona overrides
The three pre-installed Client personas differ from the firm defaults above:
| Persona | What it changes |
| ---------------------------------- | --------------------------------------------------------------------------------------------------------------------------------- |
| HENRY, ages 28 to 44 | Success threshold 85% |
| Pre-Retiree, ages 55 to 67 | Reserve length 5 yrs; Roth strategy Fill to top of 24% bracket, with an elaboration naming the 22% bracket for ages 60 through 72 |
| Retiree / Decumulator, 65 and over | Strictness Guided; Reserve length 7 yrs; its own Guyton-Klinger wording, with a ceiling in goal years |
They are yours to review, rename, edit or delete. Anything a persona leaves alone inherits the firm value.
## Also shipping empty
The base template carries no **Hazel instructions** until you add one, and no **Planning assumptions** until you add a label and value row.
## Where to go next
* [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions)
* [Market assumptions](/guides/financial-planning/assumptions/market-assumptions)
* [Guardrails](/guides/financial-planning/assumptions/guardrails)
* [Default templates](/guides/financial-planning/settings/personas/default-templates)
* [Planning settings](/guides/financial-planning/settings/overview)
* [Sharing and delivering](/guides/financial-planning/the-plan/sharing-and-delivering): where the disclaimer shows up.
# Solver assumptions
Source: https://docs.hazel.ai/guides/financial-planning/assumptions/solver-assumptions
Your firm's planning policy in one place: probability, fees and taxes, and decumulation strategy.
Your firm has a house view on how much cash a retiree holds, when spending gets trimmed, and how far to push Roth conversions. **Solver assumptions** are where you write it down once, so every plan runs on it rather than on whoever built it.
They live in **Settings → Planning → Solver assumptions**, a firm owner's screen. There is no Save button: edits save when you move off a field, and an indicator reads "Saving…" and then "Saved". Firm owners can also open the screen from the **Planning settings** gear in the top bar of a plan open in its chat.
A small tag beside each field name reads **Overridable per persona** or **Set Firm-wide**, the field's scope.
## Probability & simulation
* **Success threshold** (Overridable per persona). The probability of success a plan solves toward, shipped at 90%. Below it, the plan is flagged.
* **Monte Carlo simulations** (Set Firm-wide). How many market paths each projection models. More are more precise and slower. The shipped value is 1,000.
Raise the success threshold and plans turn more conservative: more saving, later retirement dates, lower sustainable spending. It is the bar the math must clear, not a ceiling on what a plan may display, which is a separate firm setting in the [Assumptions reference](/guides/financial-planning/assumptions/reference).
## Fees & taxes
* **Advisory fee** (Set Firm-wide). Annual fee drag applied to assets under management in every projection.
* **Terminal heir tax rate** (Set Firm-wide). The rate heirs are assumed to pay on inherited pre-tax balances, used to value the after-tax estate.
## Decumulation strategy
Four fields, three of them free text, all overridable per persona.
A slider, in years: how much retirement spending the plan carves into safe assets, so a client never sells equities into a down market. It ships at three years.
The text box under the reserve-length slider, with no heading of its own, says what the reserve holds and when it is topped back up. What ships:
"Hold the reserve in cash, T-bills, and short-duration bonds. Refill it from equities in years the portfolio finished up."
Your spending policy in retirement, written as rules that adjust as markets move. What ships:
"Use Guyton-Klinger guardrails: trim spending 10% when the withdrawal rate drifts 20% above its initial target, and raise it 10% when it falls 20% below. Skip the inflation raise in any year the portfolio finished down."
Three suggestion chips sit under the box: **Savings rate**, **Debt-free date**, and **Simple assumptions**. Each appends a sentence to whatever the box already holds, so clear the box first to use a chip's wording alone.
This is spending policy. Rules Hazel must never break belong in [Guardrails](/guides/financial-planning/assumptions/guardrails).
Six choices, one at a time, setting the default ceiling for annual conversions. The chips read exactly:
* **Stop under first IRMAA tier**
* **Cap at ACA-subsidy MAGI**
* **Drive off the heir's bracket**
* **Fill to top of 24% bracket**, fixed at 24 percent rather than a bracket you pick
* **Low-income years only**
* **No default**
The shipped choice is **Stop under first IRMAA tier**, elaborated in the box below it: "Convert each year up to just under the first IRMAA tier, so the client avoids a higher Medicare Part B/D premium."
Hazel honors that box over the chip where the two disagree, so a house rule like filling the 22 percent bracket for ages 60 to 72 goes there, with whichever chip is closest.
## The three layers
Every value a plan uses comes from one of three layers:
* **Firm.** Everything on this screen.
* **Persona.** The matched Client persona's own values, for the fields tagged **Overridable per persona**. A persona carries only what it changes and inherits the rest, so it does not go stale when you edit a firm number.
* **This plan.** One household's plan, changed in that plan's **Plan settings** panel in the chat view.
The three personas Hazel installs already use that layer, with their own success thresholds, reserve lengths, Roth strategies and withdrawal wording. A persona copied from another brings those values along; one started from the **Base template** inherits everything.
## Changing one plan
Open the plan in its chat, and the sliders icon labeled **Plan settings** in the top bar opens a panel on the right. A household's **Financial Planning** tab has no sliders icon, so press **Open chat** first.
Every row carries a chip: **Firm**, **Persona** with the persona's name, or **This plan**. **Withdrawal guardrails** appears read-only, naming the layer that supplied the wording.
Three solver fields are adjustable per plan: **Success threshold**, **Reserve length**, and **Roth conversion strategy**, alongside any market row your firm has hand-edited. A value outside the allowed range is refused with the range named, not quietly rounded.
The advisory fee, terminal heir tax rate, and simulation count never leave the firm, so the panel leaves them out.
Nothing changes until you press **Apply and replan** in the panel's footer. A reset arrow on a changed row restores the inherited value, and **Discard changes** drops everything staged. The button stays unavailable until you change something, and while Hazel is mid-turn in that chat. Applying posts a short message in the chat.
The plan's **Appendix** carries the same provenance: every row under **Solver assumptions** in the **Applied settings** block is tagged **Firm**, **Persona**, or **This plan**. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
## Where to go next
* [Market assumptions](/guides/financial-planning/assumptions/market-assumptions): returns, volatility, inflation, taxes.
* [Guardrails](/guides/financial-planning/assumptions/guardrails): the rules nothing overrides.
* [Assumptions reference](/guides/financial-planning/assumptions/reference): every shipped default.
* [Client personas](/guides/financial-planning/settings/personas/overview): how a household gets matched.
# Best practices
Source: https://docs.hazel.ai/guides/financial-planning/best-practices
The habits behind plans that need fewer corrections: firm setup, household data, the brief, the audit, settings upkeep.
You want each plan to need fewer corrections than the last.
## Set your firm's shape before you scale
Every plan starts from the shape in **Settings → Planning**: your [**Plan templates**](/guides/financial-planning/settings/overview), with your **Base template** and your **Client persona** entries.
* **Build on the defaults first.** Let Hazel build one plan before you change a setting, and read it as a client would. Then fix in order: structure in [Plan design](/guides/financial-planning/plan-design/overview), the [instructions](/guides/financial-planning/settings/personas/instructions) that set the voice, then the [assumptions](/guides/financial-planning/assumptions/solver-assumptions).
* **Write your guardrails.** The rules Hazel must never break, however a request is phrased. Two ship already, and you add yours on the base template. See [Guardrails](/guides/financial-planning/assumptions/guardrails).
* **Sort out the personas you have.** Your firm arrives with three: HENRY, Pre-Retiree, and Retiree / Decumulator. Review, rename, edit or delete them first. Each carries its own plan design and instructions, so base-template work does not reach a household that matched one; **Re-sync from base** puts it back on the base design. See [Personas](/guides/financial-planning/settings/personas/overview) and [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
* **Set the Client disclaimer before anyone exports.** Nothing prints under a plan until your firm's own text is in the **Client disclaimer** card, and an edit reaches plans already published, next time one is opened.
## Keep each household's profile current
A plan reads the household's [Intelligent Client Profile](/guides/financial-planning/your-data/how-hazel-uses-the-profile) first, and that profile learns only from recorded meetings, client emails, uploaded documents, notes, and what you tell Hazel. A CRM or custodian connection puts no facts there.
* **Record the meetings where goals come up.**
* **Clear the review queue before you build.** Verifying a fact lets a plan state it plainly instead of labeling it. The queue is **Review**, on the household's **Data** tab. See [Reviewing and verifying data](/guides/intelligent-client-profiles/managing-client-data/reviewing-and-verifying).
* **Tell Hazel what changed, as it changes.** In the household's chat: "They paid off the HELOC and opened a 529 for Maya." Confirm the update Hazel offers back and it joins the profile. See [Adding client data](/guides/intelligent-client-profiles/managing-client-data/adding-client-data).
## Front-load the brief
One paragraph before Hazel starts covers it: who the plan is for and who reads it, the one question it has to answer, the goals and the constraints. [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief) has a template to paste.
## Build early, then fill the gaps
* Where an answer would move a number, the plan carries a questions card with at most three questions. **Looks right** accepts the figure in use; edit the answer and the button becomes **Use this**.
* The household's **Financial Planning** tab lists what Hazel has and what it is missing; its **Prepare data** card turns those gaps into a document request through **Prepare data for full financial plan**.
Upload the documents that settle those points, then use **Regenerate** in the top bar of the household's **Financial Planning** tab. See [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating).
## Carry a plan into the next review
Working the plan you have keeps the household's planning history in one place; a new plan starts again at V1.
Before a review, add what has happened since: new statements, the new tax return, the meetings recorded. Then regenerate instead of rebuilding, and ask Hazel in the plan's chat what changed between the two versions.
A regenerate can land on a different persona and pick up firm settings changed since the first build, so treat the refreshed plan as unreviewed and run the audit below.
## Audit before you present, every time
The info button in the plan's top bar opens **Plan orientation**, a brief on what the plan concludes and why, on the versions that have one. Then read **Applied settings** in the plan's **Appendix**: its **Persona** row names the Client persona the plan was built from. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
Take the plan's probability of success, one balance-sheet number, one income number. A figure underlined with dots opens onto its sources, its Confidence and, where Hazel recorded them, the arithmetic behind it.
Settle an **Unverified** figure under your main recommendation before the meeting. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from) for what each confidence label means.
The bar at the bottom offers **Save** for text and value edits, **Recalculate** when the math has to run again. **Export PDF** is grayed out while anything is unsaved, and the PDF is the last saved version.
## Use levers in the room
Where a plan carries dials, every figure bound to them moves with them, and nothing saves. **Reset to optimized** in the panel puts them all back, and a reload loses them, so settle the values before you leave. If the dials belong to a recommendation, approving it records what is on screen, and **Save** in the bottom bar makes that the plan's next version. Where the panel notes that projections do not respond to a control yet, that dial is illustration. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios).
## Ask for substance, edit for polish
| Situation | Do this |
| --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------- |
| A sentence is nearly right | Hover the text, click the pencil, fix it in place |
| A section is framed wrong | Ask Hazel: **Edit** in the top bar in chat, or **Open chat** on the household's **Financial Planning** tab. Both open the same conversation |
| A figure is wrong | Open it, edit the assumption underneath and **Queue change**, or correct the fact it came from in the profile |
| Every plan has this problem | Change your firm's settings |
## Review your settings on a schedule
* **Market assumptions**, once a year, with whoever owns your firm's market view. One live set, saved as you type, so every plan built after a change uses the new number; **Publish to firm** only confirms what has already saved. Agree the numbers, then regenerate one plan you know well before telling the team. Published plans keep what they were built with until you rebuild them.
* **Personas**, once or twice a year: which ones you have customized, whether those changes still match how you plan, whether any two now describe the same household.
* **Instructions and guardrails**, whenever you notice yourself giving the same correction twice.
Keep a note of why each setting is what it is.
## Team habits
* **One owner per design at a time.** A save writes only the design you are in, so a colleague's work on another persona is safe. Inside one design the later save wins, so agree who has it. When your copy goes stale, **Settings → Planning** offers **Reload**, which takes their version and drops what you have done since.
* **In the layout designer, nothing is saved until you press Save.** Leave through the **Planning settings** back link, which offers **Keep editing**, **Discard and leave** and **Save and leave**. Closing the tab loses the session.
* **Train new hires on the sample plan.** **See financial plan example** on the **Financial Planning** page opens a complete plan with no client data. It is fixed, so it will not reflect a settings change.
* **Use the thumbs when you read a plan in chat.** They sit in the plan's top bar in the chat view only, so open the plan in chat to send feedback. See [Anatomy of a plan](/guides/financial-planning/the-plan/anatomy-of-a-plan).
## Where to go next
* [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief)
* [Prompt library: Financial Planning](/resources/prompt-library/financial-planning)
* [Troubleshooting and FAQ](/guides/financial-planning/troubleshooting-and-faq)
# Glossary
Source: https://docs.hazel.ai/guides/financial-planning/glossary
Plain-English definitions for the words the Financial Planning Agent uses, from plans and versions to templates, strictness and assumptions.
You are about to build a plan or walk a client through one, so these words need to mean one thing.
## The plan
**[Plan](/guides/financial-planning/the-plan/anatomy-of-a-plan).** The document Hazel builds for a household, which can hold several.
**Section.** A top-level part of a plan, listed under Plan sections.
**Appendix.** The plan's full detail, reached from the foot of that list.
**Tab.** A section inside another, shown as a tab of Getting there.
**[Block, chip](/guides/financial-planning/the-plan/visual-reference).** One thing on the page, and a chip in a design.
**Plan orientation.** A short brief on what this plan concludes and why.
**Figure.** A number kept in sync with the plan's math. You cannot type over one.
**[Provenance and calculations](/guides/financial-planning/your-data/where-numbers-come-from).** The panel behind a figure: sources, methodology, assumptions.
**Confidence.** A column in that panel: **Verified**, **Unverified** or **Modeled**.
**[Lever](/guides/financial-planning/the-plan/levers-and-scenarios).** A dial you move in the browser; it recalculates figures and saves nothing.
**Edit.** The chat view's top-bar button, which opens the conversation with Hazel about this plan. Not an editing mode, and the household's Financial Planning tab has none.
**[Staged changes](/guides/financial-planning/the-plan/editing-a-plan-directly).** Edits made but not saved, counted in a bar at the bottom of the plan.
**Save.** Commits your staged edits as one new version.
**Recalculate.** Sends the staged changes that need the math run again.
**Discard.** Throws staged changes away. Saved versions are untouched.
**[Version](/guides/financial-planning/the-plan/versions-and-regenerating).** A saved state of the plan. Older ones stay readable, and read only.
**Regenerate.** Rebuilds the plan from the household's latest information. On the household's Financial Planning tab only.
**Status.** **Ready to review**: a version is there to open. **Draft in progress**: none yet.
## Firm configuration
**[Base template](/guides/financial-planning/settings/base-template).** Your firm's default plan. Every Client persona builds on it.
**[Client persona](/guides/financial-planning/settings/personas/overview).** A client type your firm serves. Hazel matches each household to the closest.
**[Plan design](/guides/financial-planning/plan-design/overview).** The card showing a template's shape, block by block.
**[Design layout](/guides/financial-planning/plan-design/using-the-editor).** The button that opens the full layout editor.
**[Module](/guides/financial-planning/plan-design/modules).** A named piece of planning. Left out means no dedicated treatment of it, never a banned topic.
**[Instructions](/guides/financial-planning/settings/personas/instructions).** Your firm's guidance, in an ordered list: Hazel instructions on a template, Instructions to Hazel on a block.
**[Strictness](/guides/financial-planning/settings/personas/strictness).** How tightly a layout binds Hazel: **Exact**, **Guided** or **Flexible**. Flexible by default.
**[When it appears](/guides/financial-planning/plan-design/block-settings).** **Always**, or **When it applies**. Called When it shows up in the layout editor.
**Leave out when.** The one named condition under which a block may be dropped.
**Sample data.** The layout editor switch that fills a design with a made-up household.
**[Customized](/guides/financial-planning/settings/personas/inheritance).** A persona whose layout is its own copy, no longer following the base template.
**Re-sync from base.** Puts that layout back on the base template. A discard, not a merge.
**Plan settings.** The sliders panel in the chat view, for overriding a value on one plan. The household's Financial Planning tab has none.
## Assumptions
**[Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).** The inputs the math runs on: probability, fees and taxes, decumulation strategy.
**Probability of success.** How often a plan works across many modeled market outcomes. Displayed capped at ninety-nine percent unless your firm sets another ceiling.
**Monte Carlo simulations.** How many of those runs, one thousand by default, set firm-wide.
**Success threshold.** The probability a plan is judged against, ninety percent by default.
**[Market assumptions](/guides/financial-planning/assumptions/market-assumptions).** Return, volatility, inflation and tax figures, in one live set.
**Market scenario.** The Applied settings row naming the set a plan drew on.
**Planning assumptions.** Extra firm figures, a label and a value each, on the base template.
**[Guardrails](/guides/financial-planning/assumptions/guardrails).** Your firm's never-do rules. A persona can add one, never remove one.
**Withdrawal guardrails.** A different thing: the solver assumption that flexes spending as the portfolio moves.
**[Applied settings](/guides/financial-planning/your-data/where-numbers-come-from).** A plan's effective settings, in its Appendix, tagged **Firm**, **Persona** or **This plan**.
**Apply and replan.** Commits those overrides and recalculates, as a new version.
**Client disclaimer.** Your firm's footer, rendered as written under every plan and in the PDF.
## Data
**[Intelligent Client Profile](/guides/intelligent-client-profiles/overview/introduction).** A household's lasting facts, from emails, meetings, notes and documents.
**[Financial Plan document readiness](/guides/financial-planning/overview/getting-started).** The eight checks on a household's Financial Planning tab. Seven decide Ready to build.
**Prepare data.** Uploads documents you hold, then drafts a request email for the rest.
**Clarifying points.** The middle build step, where Hazel asks what would move the plan.
## Where to go next
* [Introduction](/guides/financial-planning/overview/introduction)
* [Visual reference](/guides/financial-planning/the-plan/visual-reference)
* [Assumptions reference](/guides/financial-planning/assumptions/reference)
* [Troubleshooting and FAQ](/guides/financial-planning/troubleshooting-and-faq)
# Access and availability
Source: https://docs.hazel.ai/guides/financial-planning/overview/access-and-availability
Where planning opens, who can open a plan, and what a firm owner sets for everyone.
You are about to build a plan for a client. Financial planning opens in three places: the **Financial Planning** tab on every household, the **Financial Planning** page in the left nav, and **Ask Hazel** with the chat set to **Financial Planning** mode. A request in plain Ask mode will not start a plan.
## Who can build a plan
Anyone at your firm who can see a household can build a plan for it, with nothing planning-specific to switch on per advisor. A household holds as many plans as you need, each with its own version history.
## Who can open a plan you built
Plans are personal for now, so a colleague with full access still sees the household's **Financial Planning** tab as a household with no plan yet. No permission setting changes this, and shared financial-plan chats are a fast follow. A colleague can build their own plan for the household, leaving yours untouched, and **Export PDF** in the plan's top bar produces the finished document to send to anyone once your changes are saved.
If a colleague cannot see the household at all, that is ordinary household access. See [Roles and permissions](/guides/settings/members/roles-and-permissions).
## What a firm owner sets for everyone
**Settings → Planning** belongs to the firm owner, and everything set there applies to every plan the firm builds. Firm Admin is a separate role that does not include it, so a Firm Admin who cannot find the Planning tab should ask the firm owner to make the change. The sidebar has three groups: **Plan templates**, **Client persona**, and **Assumptions**.
| Setting | Effect |
| ------------------------ | -------------------------------------------------------------------------------------- |
| **Base template** | The blueprint every plan starts from |
| **Plan design** | Which sections, tabs, and blocks a plan has, in what order |
| **Hazel instructions** | Voice, emphasis, and content guidance in your words |
| **Planning assumptions** | Named figures Hazel plans against |
| **Guardrails** | Your never-do rules. Hazel will not cross them, and a persona can add but never remove |
| **Client disclaimer** | The footer under every plan and in the PDF |
| **Client persona** | How the shape and assumptions change by client type |
| **Solver assumptions** | How a projection is run and judged: success threshold, fees, taxes, drawdown |
| **Market assumptions** | Return, volatility, inflation, and tax figures |
**Hazel instructions**, **Planning assumptions**, and **Guardrails** are blocks inside the **Base template** card; **Plan design** and **Client disclaimer** are cards of their own below it. Everything saves as you go.
The **Client disclaimer** box shows wording Hazel has drafted for you, but it is not applied until you edit the box yourself, so a firm that never opens it publishes plans with no footer. Set yours before the first client-facing export. Hazel reads the disclaimer fresh each time a plan opens, so an edit reaches earlier plans too.
Advisors who are not firm owners never see these panels, though every plan they build follows them. A bookmarked link to the plan design editor opens, then reports that they cannot reach it. They can still read the [Applied settings block](/guides/financial-planning/your-data/where-numbers-come-from) in the plan's **Appendix**, which tags each setting with the layer that set it.
An advisor can still tune a single plan. **Plan settings** is a chat-view control, so from a household's **Financial Planning** tab, **Open chat** takes you there first. The override belongs to that one plan, and other plans for the household still inherit the firm and persona values. See [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).
## Where to go next
* [Getting started](/guides/financial-planning/overview/getting-started): build your first plan, and read the finished example first.
* [Planning settings](/guides/financial-planning/settings/overview): the firm owner's full tour.
* [Anatomy of a plan](/guides/financial-planning/the-plan/anatomy-of-a-plan): what a plan contains.
* [Roles and permissions](/guides/settings/members/roles-and-permissions): how household access works.
# Getting started
Source: https://docs.hazel.ai/guides/financial-planning/overview/getting-started
Where to start a plan, what happens while it builds, and what to look at when it lands.
You are going to build a plan and walk a client through it. Starting one is a single click from the household, the draft usually lands inside fifteen minutes, and you do not have to watch it.
## Look at the finished example first
The welcome card on the **Financial Planning** page carries **See financial plan example**, a finished plan built on a made-up family. A six-callout tour runs on first open, and **Take the tour** at the top replays it. If your own plans have replaced that card, **Sample household data** in the **New financial plan** flow opens the same plan.
Almost everything works: click through the sections and the tabs inside **Getting there**, move a lever to watch the numbers respond, open the **Appendix**, and select a figure for the panel behind it, headed **How this was calculated** where there is arithmetic to show. What does not: **Approve** and **Ignore** are grayed out, the question cards take no answer, and there is no editing, version history or PDF download.
## The three entry points
A plan belongs to whoever built it, so build the plan you are going to present: a colleague with full access to the household still cannot open yours. See [Access and availability](/guides/financial-planning/overview/access-and-availability).
### From a household
Best with a client in front of you. The **Financial Planning** tab opens on a readiness list headed **Financial Plan document readiness**, and nothing on it blocks a build. It checks Accounts & Holdings, Calendar, Tax documents, Uploaded documents, Meetings & transcripts, CRM, Email and Business valuation. Rows read a tick, **Incomplete**, **Missing**, **Loading…** while a check runs, or **unavailable** where Hazel could not reach the source.
Calendar, CRM and Email are about your own setup, so **Missing** names what is absent, such as no mailbox connected; connect that tool to Hazel to clear it. Calendar reads **Incomplete** when it is connected but nothing is booked. See [Integrations](/guides/integrations/overview). Business valuation is shown whatever the household owns, and **Missing** means the plan labels the figure as an assumption.
Seven of the eight decide your card, Business valuation aside:
* Complete on all seven: **Ready to build**, summarizing what Hazel has to draw on.
* Short on any: **Begin right away**, naming what is connected and what is missing, beside a **Prepare data** card for uploading documents onto the household and drafting an email asking the client for the rest.
The button is **Draft financial plan** either way, and building early is fine: missing data and assumptions are labeled in the finished plan.
### From the Financial Planning page
Best for a first plan or a book of clients: this route runs from no data to a finished plan. It opens from **Draft a plan** on the welcome card, or **New financial plan** once you have a plan of your own.
Three cards: **From my CRM**, **Manual entry** and **Sample household data**.
**From my CRM** leads to **Choose your CRM**: Salesforce, HubSpot, Wealthbox or Redtail. If Hazel already holds clients you go straight to your client list and can build now; if not, **Your CRM is connected** says the import is running, usually five to ten minutes, with **Done** to leave while it does. The manual route stays open.
HubSpot takes one extra step, headed **One more step in Settings**: choose which records Hazel should import, and the first sync starts when you save.
Pick a client, or create one. On the manual route the client comes before the documents, so the button reads **Continue** and uploading is next. On the CRM route the data is already in Hazel, so this is the last step: **Draft financial plan** starts the build.
A client whose record has no contacts on it cannot take documents, so if the button will not enable, add a contact first.
Hazel asks for brokerage and custodian statements as PDFs, and recent tax returns (Form 1040). Files are read as they arrive, a few minutes each, and **Draft financial plan** stays unavailable until at least one has been read. Close the flow and come back while Hazel reads. If nothing could be read, the step says so and you will need a different document.
Either route hands you off to the household's **Financial Planning** tab, which starts the build and shows progress.
#### The page is also your list of plans
Once you have built something, the welcome card gives way to a table of your plans, and selecting a row opens that household's **Financial Planning** tab. **Ready to review** means the plan has published a version you can open; **Draft in progress** means it has not, because Hazel is still building or a build stopped part way. If a row has sat there a while, open it.
### From Ask Hazel
Best when planning comes up mid-conversation. Chat starts a plan only in **Financial Planning** mode. On a new chat's start screen, choose it on the segmented control under the message box, alongside Ask and Tax Planning. A chat keeps the mode it started in, so switch before your first message.
Say what you want before Hazel starts, rather than correcting the draft:
```text theme={null}
Build a financial plan for the Hendersons. They want to retire in 2032,
they care most about whether they can keep the lake house, and Ann's
deferred compensation is the piece I want covered properly.
```
Everything here shapes the first draft. See [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief) and the [prompt library](/resources/prompt-library/financial-planning).
## What happens during a build
A three-segment rail marks progress, **Reading files**, then **Clarifying points**, then **Writing plan**.
| Step | On screen |
| ------ | ----------------------------------------------------------------------------------------------------------- |
| 1 of 3 | **Reading your paperwork**. Hazel reads your files against the client profile on file, usually 5-12 minutes |
| 2 of 3 | **A few questions before I draft** |
| 3 of 3 | **Drafting your plan** |
When it publishes, the card reads **Your plan is ready**, with a **View plan** button.
The card carrying **Draft financial plan** estimates 5-15 min, though a household with a lot of documents can run past twenty minutes: the badge estimates time left, then reads **Taking longer than usual** once a build is over budget.
### The questions step
Hazel pauses part way through to ask a few things the documents did not settle. On the household's **Financial Planning** tab an unanswered question falls back to a reasonable assumption after thirty seconds, and hovering or focusing the card pauses that countdown. The badge reads **Waiting for answer**, over a line naming what Hazel has read so far, or a count such as **Question 2 of 4**. Questions come one at a time: answer, or press **Skip this one** or the pill labeled **Make reasonable assumption**, which doubles as the countdown to it.
A build started in Ask Hazel never answers its own questions: they arrive in the chat itself, with no countdown and a plain **Skip**, so if a chat build looks stalled, go back to that chat and look for the question card.
### While it runs
* **Leave the page.** A banner on every tab of the household says Hazel is drafting the plan.
* **Listen for the chime.** Hazel plays a short sound when the plan publishes, wherever you are inside Hazel, though not while you are watching the build. The speaker icon on the Financial Planning tab mutes it, and Hazel has to be open in a tab to hear it.
### If the build does not finish
A build that stops before it publishes leaves no version behind, so the household and any plan you have are untouched. The tab says **Build didn't finish** and offers two ways on.
* **Try again** continues from the documents Hazel already read and the questions you answered, so reach for it first.
* **Start over** discards the attempt and returns you to the readiness list, for when you would rather add more data.
While a build runs, the Financial Planning tab shows its progress in place of the **Draft financial plan** button, so you cannot start a second build for the same household by accident. Once it lands, ask for changes in the plan's chat, or press **Regenerate** in the tab's top bar to refresh it from the household's latest information.
## The first five minutes with a new plan
The info button at the top of the plan opens **Plan orientation**, a short read on what the plan concludes and why. Not every version has one; where there is none the button is absent.
The **Persona** row of the **Applied settings** block in the plan's **Appendix** names it, and the Financial Planning page repeats it in its **Persona applied** column. **None**, or a dash on the page, means your firm's base template alone. If it is wrong for this family, say so in the plan's chat and ask for a new version. See [Client personas](/guides/financial-planning/settings/personas/overview).
Figures Hazel can source carry a dotted underline and open a panel with where they came from. Check the number the summary leads with, one balance and one income figure. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
The plan works the numbers out again while you watch. Dials Hazel cannot recompute in the browser render no control at all, so ask in the plan's chat instead. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios).
Reading the plan in chat, **Edit** at the top (tooltip: **Ask Hazel to change this plan**) opens the conversation with Hazel. On the household's **Financial Planning** tab there is no **Edit** button, and **Open chat** in the top bar goes to the same conversation. Either way the reply comes back as a new version.
Prose you rewrite and assumptions you queue are different: they collect in a bar at the bottom of the plan and become a version only when you press **Save**. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
## Where to go next
* [Anatomy of a plan](/guides/financial-planning/the-plan/anatomy-of-a-plan)
* [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief): a sharper first draft.
* [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating)
* [Sharing and delivering](/guides/financial-planning/the-plan/sharing-and-delivering)
* [Planning settings](/guides/financial-planning/settings/overview): for firm owners.
* [Troubleshooting and FAQ](/guides/financial-planning/troubleshooting-and-faq)
# What is the Financial Planning Agent?
Source: https://docs.hazel.ai/guides/financial-planning/overview/introduction
Hazel reads what your firm already holds on a household, runs the math, and gives you a financial plan you can change by asking.
You have a household to sit down with, and most of the planning work is pulling together what your firm already knows about them. Hazel reads the household's statements, tax returns and meeting history, works out what this household's situation calls for, runs the math, and lays out a plan you can present. Ask for the plan to say something different and Hazel writes a new version.
## What you get
A plan opens on **Summary**, and the left rail runs through **Today**, **Goals**, **Getting there** and **Decisions**, with an **Appendix** at the foot. **Getting there** holds tabs and opens on **Your priorities**, where Hazel ranks what this household should deal with first. The rest are the topics your firm's template calls for: out of the box, **Cash flow**, **Retirement** and **Insurance**.
Accounts and holdings, tax filings, documents you upload, meetings and transcripts, emails and notes. It all reaches Hazel that way, so connecting a CRM or a custodian does not by itself put a balance in front of Hazel.
Sourced figures carry a dotted underline. Select one and Hazel shows where the number came from, the assumptions behind it, and any arithmetic.
Plans carry **levers**: dials such as a retirement age or a savings rate. Move one and the numbers rework on screen while the client watches. The change is exploratory, so it doesn't save. To keep it, ask Hazel.
Your firm owner sets the guidelines once: which sections a plan contains, the **Hazel instructions** that steer how it reads, the assumptions to plan against, and the **Guardrails** no plan is allowed to cross.
For now a plan belongs to whoever built it, so a colleague with full access to the household cannot open yours. See [Access and availability](/guides/financial-planning/overview/access-and-availability).
## How Hazel decides what goes in a plan
Hazel composes each plan around the household's facts, so a household with a large concentration of company stock can get a section about that stock, and two households at the same firm can end up with different sections. Your firm's guidelines and templates are a starting point, and a firm that wants every plan built to the same template can say so, in [Strictness](/guides/financial-planning/settings/personas/strictness).
You can talk to a plan, too. "Add a lever for retiring at 62." Each request comes back as a new version, and the earlier versions remain accessible.
## How it works
Open the household's **Financial Planning** tab, use the **Financial Planning** page, or ask in **Ask Hazel** with the chat in **Financial Planning mode**.
Hazel reads the household's client profile, its documents and its meeting history, then matches the household to the closest of your firm's **Client personas**, which informs which sections are in the plan, how it reads, and some of its starting assumptions. See [Client personas](/guides/financial-planning/settings/personas/overview).
The numbers come out of a library of planning calculations, run with your firm's **Solver assumptions** and **Market assumptions**.
Open **Plan orientation**, the info button at the top of the plan, for what this plan concludes and why. Check a few figures against their sources, move a lever, edit any sentence in place. Asking Hazel for a change takes the plan's conversation: the **Edit** button in the chat view, **Open chat** on the household's Financial Planning tab. Then download the saved plan as a PDF for the meeting.
Every figure comes either from something Hazel read or from an assumption, and the plan labels which, so read the figures resting on an assumption before you present. Where something material is missing, Hazel asks before drafting: on the household's **Financial Planning** tab, a question you don't answer becomes a reasonable assumption and the plan proceeds. A build you started in a chat pauses until you answer.
## Where to go next
* [Getting started](/guides/financial-planning/overview/getting-started): the entry points, the data you need, and what a build looks like from the inside.
* [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief): say what you want before Hazel starts.
* [Anatomy of a plan](/guides/financial-planning/the-plan/anatomy-of-a-plan)
* [Settings](/guides/financial-planning/settings/overview): plan templates, Client personas, Plan design and assumptions.
* [Planning philosophy](/guides/financial-planning/overview/planning-philosophy)
* [What it can and cannot do](/guides/financial-planning/overview/what-it-can-and-cant-do): where your sign-off is needed.
New to planning software? The [Glossary](/guides/financial-planning/glossary) gives persona, lever, solver assumption and the rest a line each.
# What it can and can't do
Source: https://docs.hazel.ai/guides/financial-planning/overview/what-it-can-and-cant-do
What the Financial Planning Agent builds, what Hazel will not invent, and which calls stay with you.
You are going to put your name on this plan and walk a client through it.
## What Hazel does
Out of the box: net worth and balance sheet, cash flow, emergency fund, debt paydown, goal funding, insurance, retirement, and probability of success. Eleven further topics, among them Investments, Roth conversion and Social Security, switch on in **Plan design**, which also sets the order.
Probability of success from a Monte Carlo simulation, a thousand market paths by default, with the range of outcomes behind it. Also withdrawal order, Roth conversion schedules, when each earner claims Social Security, required minimum distributions, college funding, and equity compensation.
Statements, tax returns, policy documents, estate documents, and meeting transcripts, scans and photographs included.
Ask in the plan's chat: rewrite a section, add a tab, swap a chart, add a lever, tighten the language. Each change becomes a new version.
## What Hazel will not invent
| Hazel will not | What happens instead |
| ----------------------------------------------- | --------------------------------------------------------------------------------------------------------- |
| Make up a balance, an income, or a premium | Asks you when the number would move the plan, and records the assumption behind any figure that uses it |
| Assume a goal nobody in the household expressed | Builds on the goals the household's record supports, and asks about the rest |
| Quietly swap in a market figure | Uses your firm's **Market assumptions**, and names the value's origin on the figure's source card |
| Treat a figure nobody has confirmed as settled | Labels it **Verified**, **Unverified** or **Modeled** on its source card |
| Print a probability of success of 100 percent | Caps it at the ceiling your firm sets, 99 percent until someone changes it, and says when the cap applied |
Hazel can still misread an unusual statement layout or attach a figure to the wrong account, so check anything you will say out loud.
## What Hazel never does
No instruction, persona, or request changes the four rules below.
* **Hazel does not place trades or move money.** A plan can recommend a change in allocation, or in which account holds what. Executing it is your job.
* **Tax and legal advice are out of scope.** Hazel models the tax consequences of a planning decision, using your firm's effective tax rates in **Market assumptions**, sourced from IRS brackets. A tax position, a document, or a filing needs a professional.
* **A plan does not reach a client on its own.** There is no client login and no share link. Both routes are yours: export the PDF and hand it over, or walk the household through it on your screen.
* **Suitability is your call.** Whether a recommendation suits this person is a judgment for you to make.
One thing here does email a client, and only when you press **Send**: the document request you compose in **Prepare data** on the household's **Financial Planning** tab. It goes from your own Gmail or Outlook mailbox and carries no plan.
## Two rules your firm sets, and can unset
Two ship switched on: "Never recommend a specific named security or ticker." and "Never propose a Roth conversion that pushes the client into the next IRMAA tier." A firm owner can remove either, or add your own. See [Guardrails](/guides/financial-planning/assumptions/guardrails).
## Where you sign off
**Approve** and **Ignore** on each recommendation stage until you press **Save**, and export is switched off while anything is staged, so save or discard first.
Hazel asks when a judgment call would move the numbers and nothing on record settles it. Some arrive during the build, at the step where Hazel pauses before drafting. See [Getting started](/guides/financial-planning/overview/getting-started).
A plan can also carry a **Questions** list, at most three, each with Hazel's answer filled in and a line saying what it does to the figures. **Looks right** accepts that answer, so it is a decision. Change it and the button becomes **Use this**, and the correction queues for recalculation.
Hazel matches the household to the closest **Client persona**, and the **Appendix** names the one it used. If the match is wrong, tell Hazel in chat which persona fits and ask for a rebuild. See [Client personas](/guides/financial-planning/settings/personas/overview).
Select the figure and **Provenance and calculations** opens: its sources with a confidence label on each, the methodology, and the assumptions used. A figure with no source renders as plain text and does not open. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
Until someone at your firm edits the **Client disclaimer** card, a plan carries no footer at all, on screen or in the PDF. Set it before your first client-facing export. See [Planning settings](/guides/financial-planning/settings/overview).
## Known edges
* **A household with no documents and no meetings has nothing to plan from.** Balances and incomes arrive through a document, a meeting, an email or a note, and a CRM or custodian connection alone brings none. Upload statements and tax returns, or use **Prepare data** to ask the client. See [Documents and CRM data](/guides/financial-planning/your-data/documents-and-crm).
* **A household can hold more than one plan.** The household's **Financial Planning** tab opens the newest, and each plan keeps its own version history. To weigh two courses of action, ask for the comparison inside the plan you have. See [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating).
* **Only some levers move.** A lever that renders as a control recalculates the figures that depend on it in your browser. It does not rewrite the words around it, and moving it never becomes a version, so ask Hazel for a change you want to keep.
* **Editing and rebuilding are different tools.** Hover a paragraph and a pencil appears: prose, framing and emphasis are yours. **Regenerate**, in the household tab's top bar, rereads the household's data and publishes a fresh version. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
* **A plan belongs to whoever built it.** A colleague with full household access still cannot open it, so whoever built a plan exports and presents it. See [Access and availability](/guides/financial-planning/overview/access-and-availability).
* **A missing section has two causes.** Either your firm's **Plan design** leaves that module out, in which case Hazel folds the topic into a broader section rather than dropping it, or the household's facts did not support it. Check Plan design for the first, ask Hazel in chat for the second.
## Where to go next
* [Planning philosophy](/guides/financial-planning/overview/planning-philosophy): why Hazel asks instead of assuming.
* [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from)
* [Guardrails](/guides/financial-planning/assumptions/guardrails): adding your own rules.
* [Access and availability](/guides/financial-planning/overview/access-and-availability): who can open a plan.
# Block settings
Source: https://docs.hazel.ai/guides/financial-planning/plan-design/block-settings
A pinned heading, instructions for that block alone, when it shows up, and the one case it may be skipped.
You are shaping one piece of the plan, so a block's settings say what belongs in it, in your firm's words, and whether Hazel must include it at all.
Two places open them, with different fields.
* In **Settings → Planning**, the **Plan layout** card draws your design as a blueprint. Click a block chip or a tab name for a panel with three settings and a **Done** button. The top-level section bands are labels, not buttons, so a section's own settings live in the editor.
* Open the layout editor with **Design layout** on that same card, then select anything on the canvas or in the outline. The right-hand rail holds those three settings plus a **Heading** and an **Advanced** group.
## What each setting does
| Setting | What it does |
| ------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Heading** | A title pinned above the block, up to 200 characters. Hazel uses your words exactly. Leave it blank and Hazel writes one to suit the block. Only the layout editor has this field, so open **Design layout** to pin one |
| **Instructions to Hazel** | Content and style guidance for this block alone, up to 2,000 characters. The layout editor counts characters as you near the limit |
| **Leave out when** | The one case in which this block may be skipped, in plain English |
| **When it shows up** | **Always** puts the block in every plan you build. **When it applies** puts it in only if it fits the household. The **Plan layout** card panel and the blueprint legend call it **When it appears** |
**Suggested (legacy)** shows only on a block somebody set that way in the past. It lets Hazel skip the block freely and is being retired: switch that block to one of the other two and it disappears.
Instructions never set numbers. A block instruction saying "use a 4% withdrawal rate" does not change what Hazel calculates: the structured assumption does, and it wins. See [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).
## Writing block instructions
Most blocks in Hazel's default design arrive with instructions already: the summary's **Key numbers** block, headed **At a glance**, says "Exactly four headline tiles". Edit what is there rather than starting blank.
Concrete and local ones work, like this **Narrative** block in **Summary**: "Three paragraphs. Open with whether they can retire on their stated timeline, then the single biggest risk, then what we recommend."
Weak instructions could apply to anything: "be clear", "cover the important points".
Guidance about one block goes here. Firm-wide guidance goes in [Instructions](/guides/financial-planning/settings/personas/instructions) on the template, which is where the same instruction on five blocks belongs.
## Using "Leave out when"
A single condition, in your words:
* "The household has no dependent children."
* "There is no restricted stock or option grant."
Pair it with **When it applies**: the block lands when the household's facts call for it, and your sentence names the case it stays out of. The layout editor drops the field on an **Always** block. The panel on the **Plan layout** card still shows the box there, so a condition typed on an Always block is not a reliable exemption.
Two conditions joined by "or" usually means the block is really two blocks. Some default blocks arrive with a condition already written: the debt table is left out when the household carries no debt.
## The Advanced group
At the bottom of the settings rail sits an **Advanced** disclosure. Sections and tabs always have one; a block has one when there is something to put in it.
* **Module.** Which module this node is the home of, not editable. Removing the node removes that module from the plan. See [Modules](/guides/financial-planning/plan-design/modules).
* **Nav label** and **Nav icon** on a section, **Short tab name** on a tab. How the item reads in the plan's left menu, or on the tab when the heading runs long.
* **Levers.** On a **Levers to try** block only: the sliders the plan must offer, up to twelve. Each row takes a label, minimum, maximum, step, default, unit and note. Hazel may add dials of its own. The default design has no **Levers to try** block; add one from the **Building blocks** shelf in the layout editor.
* **Calculation.** A read-only note about the method behind a block's figures, which none of Hazel's designs carry. Where such a note and your assumptions disagree, the assumptions win.
## Sections and tabs
A section or a tab takes these same settings. Its **Heading** caps at 160 characters rather than 200, and it adds a **Subtitle** of one line, up to 240. See [Sections and tabs](/guides/financial-planning/plan-design/sections-and-tabs).
## How the settings interact with strictness
Strictness is set once for a whole design, in the **Strictness** control on the **Plan design** card above the blueprint. It decides how literally Hazel reads the settings above.
| | Under Exact | Under Guided | Under Flexible |
| ------------------- | -------------------------------------------------------------- | ----------------------------------------------------------------- | ---------------------------------------------------- |
| **Always** | Hazel must build it; a plan that misses it goes back for a fix | Hazel builds it, adapting where the household's facts call for it | A strong preference; Hazel composes each plan freely |
| **When it applies** | Built when the facts call for it | Same | Same |
| **Leave out when** | Read on a **When it applies** block | Read | Read as guidance |
With no strictness chosen, the design runs on **Flexible**, so choose one before you lean on **Always**. See [Strictness](/guides/financial-planning/settings/personas/strictness) for what each mode is for, and for the **Allow additions** switch that Exact turns off.
## Where to go next
* [Sections and tabs](/guides/financial-planning/plan-design/sections-and-tabs): the structure these blocks sit in.
* [Modules](/guides/financial-planning/plan-design/modules): one level up from a block.
* [Using the layout editor](/guides/financial-planning/plan-design/using-the-editor): where these settings live.
* [Visual reference](/guides/financial-planning/the-plan/visual-reference): what each block type renders.
* [Instructions](/guides/financial-planning/settings/personas/instructions): the same idea at template scale.
# Module reference
Source: https://docs.hazel.ai/guides/financial-planning/plan-design/module-reference
All nineteen modules Hazel ships: what each adds, where it lands, and what a household's data needs for it to have something to say.
You are deciding which topics a plan should treat on their own. All nineteen modules Hazel ships sit in the same four groups and order as the **Modules** tab of the plan design editor, where you add and remove them.
**Where it lands by default** holds whatever you have selected on the canvas, since Hazel's own modules ignore your selection; only a module your firm wrote follows it. Five modules own a whole tab and carry a small **Tab** badge.
**What it needs** is what the household's data has to hold for the module to say anything. Several also carry a leave-out condition, so a treatment with nothing behind it stays out rather than printing a blank table.
## Core planning
| Module | What it adds | Where it lands by default | What it needs |
| ----------------------------- | ----------------------------------------------------------------------------------- | -------------------------------------- | ---------------------------------------- |
| **Net worth & balance sheet** | Assets, liabilities, the net-worth trend, and the account detail behind it | Today | Account balances, property values, debts |
| **Cash flow** | Income, spending, monthly surplus, and a year-by-year cash-flow view | Its own Cash flow tab in Getting there | Income and spending detail |
| **Emergency fund** | Months of essential spending covered today, against a target of three to six months | Cash flow tab | Spending detail and cash balances |
| **Debt paydown** | Balances, rates, and a payoff order that puts balances above 6% APR first | Cash flow tab | Debts with balances and rates |
| **Insurance** | Term life and disability coverage in place, the gaps, and the cost of closing them | Its own Insurance tab in Getting there | Policies in force, income, dependents |
| **Goal funding** | One card per goal: funding status and progress toward the target | Goals | Stated goals with amounts and dates |
## Retirement & tax
| Module | What it adds | Where it lands by default | What it needs |
| ---------------------------------- | --------------------------------------------------------------------------------------------------- | --------------------------------------- | ------------------------------------------------------------------------ |
| **Retirement** | Progress toward the retirement target, a range of outcomes, and readiness figures | Its own Retirement tab in Getting there | Retirement accounts, contributions, a target retirement date |
| **Retirement income sequencing** | Which accounts fund spending, in what order, and where required withdrawals force the order | Retirement tab | Which accounts are taxable, tax-deferred or Roth, plus a spending target |
| **Social Security optimization** | Claiming ages compared for each earner, with the lifetime trade-off | Retirement tab | Dates of birth, plus an earnings history or a benefit estimate |
| **RMD calculation** | Required minimum distributions by year and account, with each earner's start year | Retirement tab | Tax-deferred balances and dates of birth |
| **Roth conversion** | Year-by-year conversions with each year's bracket impact, following your firm's conversion strategy | Retirement tab | Tax-deferred balances and a current income and tax picture |
| **Healthcare coverage estimation** | Coverage before 65, the move onto Medicare, and estimated annual costs | Retirement tab | Ages, retirement dates, coverage in force |
## Investments & analysis
| Module | What it adds | Where it lands by default | What it needs |
| ----------------------- | ------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------- | -------------------------------------------------------------- |
| **Investments** | The current allocation against the target mix, and how that mix shifts over time | Its own Investments tab in Getting there | Holdings, and which accounts are taxable, tax-deferred or Roth |
| **Equity compensation** | Vesting lanes for restricted stock, options and employee purchase plans, plus single-stock concentration against the target mix | Investments tab | Grant schedules and current share prices |
| **Monte Carlo success** | One figure for the chance the plan holds up across many varied market futures | Summary | A goal with an amount and a date |
## Wealth transfer & legacy
| Module | What it adds | Where it lands by default | What it needs |
| ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------- | ----------------------------------------------------- |
| **Education funding** | Funding targets against projected education costs, one card per child | Goals | An education goal, children's ages, savings set aside |
| **Charitable giving** | A giving strategy sized to the household's tax picture: donor-advised funds, qualified charitable distributions, gifts of appreciated stock | Estate plan tab | A stated giving intention and appreciated holdings |
| **Gifting & wealth transfer** | Annual gifting capacity, lifetime exemption use, and planned transfers to heirs | Estate plan tab | Net worth, named heirs, a stated intention |
| **Estate plan** | Documents, beneficiaries and titling: what is current, what is missing, what to fix first | Its own Estate plan tab in Getting there | Estate documents and how accounts are titled |
Adding a module whose home is not in your design creates that home: put **Roth conversion** into a design with no Retirement tab and the tab appears. Each card says where the add will land, and when a section module would fill a tab sitting there empty its button reads **Furnish the Retirement tab** instead of **Add to plan**.
## What the default plan design includes
Hazel's default design carries eight modules: the six under Core planning, plus **Retirement** and **Monte Carlo success**. So a default plan has Cash flow, Insurance and Retirement tabs but no Investments or Estate plan tab.
Adding every module is rarely right: a narrow module takes its topic over from the broader section that would otherwise cover it. [Modules](/guides/financial-planning/plan-design/modules) covers how to choose.
Two numbers here are yours to change: the 6% APR line on **Debt paydown**, in that block's **Instructions to Hazel** field, and the probability at which a plan counts as on track, in [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).
## How Hazel's personas differ
Your firm's Planning settings arrive with three Hazel personas whose designs differ from the base template and from each other; [Default templates](/guides/financial-planning/settings/personas/default-templates) covers each in full.
* **HENRY**, for high earners still building wealth, leaves out the standalone success figure: one goal's probability undersells a household juggling several.
* **Pre-Retiree** leaves out debt paydown.
* **Retiree / Decumulator** opens its Getting there tabs with retirement income sequencing and healthcare costs and drops the saving-phase treatments: no accumulation trajectory, no emergency fund, no debt paydown.
## Your firm's own modules
Modules you create sit under **Your firm's modules**, with the name and description you gave them. Start from scratch or copy one of Hazel's. Write one as a tab with no blocks of its own and its card says Hazel will compose it from your instructions.
## Where to go next
* [Modules](/guides/financial-planning/plan-design/modules): adding and removing them, and what absence means.
* [Block settings](/guides/financial-planning/plan-design/block-settings): instructions and leave-out conditions on a module's pieces.
* [Sections and tabs](/guides/financial-planning/plan-design/sections-and-tabs): how tabs sit inside Getting there.
* [Visual reference](/guides/financial-planning/the-plan/visual-reference): the charts, tables and cards a module places.
* [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from): how these facts reach Hazel.
# Modules
Source: https://docs.hazel.ai/guides/financial-planning/plan-design/modules
How to add a module, remove one, build your own, and what leaving one out means.
You want this plan to treat Roth conversion on its own, so you add that module instead of building it block by block. One move brings the year-by-year conversion schedule and the bracket work.
## The two shapes
Five of Hazel's modules are a whole tab: **Cash flow**, **Retirement**, **Insurance**, **Investments**, **Estate plan**. Adding one creates that tab inside **Getting there** with the blocks that belong in it, and a **Tab** badge marks them in the library.
The other fourteen add their pieces inside a section or tab, the way **Net worth & balance sheet** goes into **Today**. Each module's card names its default home, and you can move it once it lands.
## Finding what you can add
Open the editor with **Design layout**, then the **Modules** tab in the left rail, or from **Settings → Planning**, **See what's left out** on the **Plan layout** card. Hazel's nineteen modules are grouped by planning area, with your firm's own underneath. Use it to audit a design you inherited:
* Modules in the design carry an **In plan** badge, counted in each group header ("4 of 6 in plan"), and offer **Jump to it** in place of an add button.
* The **Not in plan** chip filters down to what you are missing; **Search modules…** finds one by name.
* A module you do not have names its landing spot: "Adds to the Cash flow tab", or "Adds to the Retirement tab (creates it)" when that tab is not in your design yet.
* **Brings** lists the pieces the module carries; **Preview** draws their shapes.
## Adding a module
Choose **Add to plan** on the module's card. Hazel drops it in, scrolls to it, and names where it landed in a message with an **Undo**.
**It arrives set to When it applies**, so Hazel includes it only when the household's facts call for it. For a topic every plan should cover, use the chevron beside **Add to plan** for the always-include option, or switch it later from the module's row in the **Outline**. On **Flexible** strictness, where a new firm starts, Always is a strong preference rather than a rule. See [Strictness](/guides/financial-planning/settings/personas/strictness).
If the tab already exists, a dialog headed **Already in this design** offers **Add it there** or **Create a second one**, so you cannot end up with two Retirement tabs by accident. If an earlier add left that tab empty, the button offers to furnish it.
Your firm's own modules follow your selection, since a custom module has no default home: a blocks module joins the tab or section you have selected, and an own-section module becomes a tab of the top-level section your selection sits inside. With nothing selected, a blocks module falls into the plan's first section. Hazel's own nineteen ignore your selection.
## Removing a module
Open the small icon at the left of a module's **Outline** row and choose **Remove from plan…**. For a tab module, that row is the tab itself, and the same action sits behind the more-actions button on the module's card. A tab a module owns has no trash can, so the menu is its only way out.
A removal takes the module's untouched pieces, plus any other module living inside a tab it owned. Hazel names everything that went and says which edited pieces it kept.
If you had edited the module's blocks, Hazel asks first, under **Save as a custom module first?**: **Save & remove** files your version in your firm's library under a name you choose, or **Remove without saving**, or **Cancel**.
## What leaving one out means
A module that is not in your layout means no dedicated treatment of that topic, not that the topic is banned. Broader modules still fold narrower ones in: with **Roth conversion** absent, a retirement or tax section can name a conversion opportunity in passing, but no year-by-year conversion schedule of its own.
It runs the other way too: add **Retirement income sequencing** and the sequencing story moves out of the general retirement narrative into a treatment of its own.
## Modules of your own
Build one when your firm has a treatment of its own, like a section your compliance review requires. **New module…** sits at the bottom of the Modules tab and asks for:
* **Name** and **Description**, both shown on the module's card.
* **How it appears in the plan**: **Its own section**, which becomes a tab, or **Adds blocks to sections**, one block joining a section or tab already there.
* **Instructions to Hazel**: what this module should cover whenever it appears.
* **Start from**: **Let Hazel write it** leaves the tab empty and has Hazel compose it from your instructions every time. **Copy existing** starts from another module's blocks. **Design now** puts it in the design to shape by hand. The blocks shape offers the last two only.
**Create module** files it in your firm's library right away, available on the base template and every persona. Under **Let Hazel write it** or **Copy existing** it waits there until you **Add to plan**.
## Choosing what to enable
Add the specialists your clients need, like equity compensation for technology employees or education funding for school-age families, then stop: every module you add is another section a client crosses to reach the decision.
When only some of your clients need a module, add it to the persona that covers them, not the base template. See [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
## Where to go next
* [Module reference](/guides/financial-planning/plan-design/module-reference)
* [Sections and tabs](/guides/financial-planning/plan-design/sections-and-tabs)
* [Block settings](/guides/financial-planning/plan-design/block-settings)
* [Using the layout editor](/guides/financial-planning/plan-design/using-the-editor)
# Plan design
Source: https://docs.hazel.ai/guides/financial-planning/plan-design/overview
Set the shape every plan starts from, and read the design your firm already has.
You want every plan your firm delivers to carry your firm's shape, not a default. Plan design sets that shape once, and Hazel fills it with each household's analysis.
Nothing here touches a plan you have already delivered. A design shapes the plans Hazel builds from now on, and an existing plan stays until someone rebuilds it.
Find it in **Settings → Planning**, on the **Base template** every plan the firm builds starts from, and on each persona under **Client persona**. Planning settings belong to the firm owner, so ask whoever owns your firm's Hazel account if you cannot see the section.
## Reading the blueprint
The card headed **Plan layout** draws your whole design at once.
| What you see | What it is |
| ---------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Bands** | Top-level sections, which become the plan's main navigation. A band without tabs shows its module count. |
| **Indented sub-bands** | Child sections, which become tabs inside the parent, whose row says "Shown as tabs". |
| **Chips** | Blocks. One chip is one piece of the plan: writing, a chart, a table, key figures, levers. |
| **Markers** | When each piece appears, under the legend headed **When it appears**: a solid check for **Always**, in every plan; a dashed circle for **When it applies**, only where it fits the household. |
A chip carrying a module shows the module's name, sometimes with a badge: "2 levers" for the what-if controls that piece offers, or one starting with "Calc" for the calculation behind its figures.
A section holding no pieces reads "Hazel composes this section freely". Under the drawing, Hazel counts your sections, the planning modules in play, and the pieces set to Always. **See what's left out** opens the module library on the rest.
### Strictness, and whose design this is
**Strictness**, in the card's header, sets how closely Hazel follows the blueprint: **Exact**, **Guided** or **Flexible**, and new firms start on **Flexible**. The same panel holds one switch, **Allow additions** ("Hazel may add sections or blocks you didn't design"), which Exact locks off. [Strictness](/guides/financial-planning/settings/personas/strictness) covers what each mode does and how it plays against each piece's own setting.
Before anyone changes it, the base template wears a **Hazel default** badge. A persona nobody has changed wears **Follows base**, faded, tracking the base template live. Your first edit gives that persona a copy of its own, and **Re-sync from base** hands it back whenever you want.
Every chip can be moved, removed or reworded, including the ones Hazel put there. A design keeps at least one section, and removing a section or tab takes everything inside it.
## The two ways to change it
A panel opens with three settings for that chip: **When it appears**, **Instructions to Hazel**, and **Leave out when**, the one condition it may be dropped. Press **Done** when finished. Use it to shape how a piece is written, without leaving the settings page.
Opens the full editor, for changing what is there: add and remove modules, reorder blocks, restructure sections and tabs, pin exact headings, and preview the shape filled in with a made-up household.
Planning settings have no Save button. What you type saves a moment after you click away, and a quiet **Saved** appears at the top of the page. The full editor is the exception: changes there wait for its own save button.
## Design, instructions and assumptions
* **Plan design**: what a plan contains, and where. Binds as tightly as your strictness setting.
* **Instructions**: how it is written, and what to emphasize. Guidance Hazel weighs.
* **Assumptions**: the numbers behind every figure. Used as given, and a persona can override a firm value while one plan can override both.
Making a topic appear in every plan takes two settings: the piece on **Always**, and strictness on **Guided** or **Exact**. A module you add starts on **When it applies** unless you use the add option that always includes it.
## Where the default shape comes from
Out of the box a plan navigates **Summary** (its own heading reads "Executive summary"), **Today**, **Goals**, **Getting there** and **Decisions**. Getting there opens on Your priorities, then three tabs your modules put there: Cash flow, Retirement and Insurance. Investments and Estate plan wait in the module library.
That shape comes from eight modules: Net worth & balance sheet, Emergency fund, Goal funding, Insurance, Cash flow, Debt paydown, Retirement, and Monte Carlo success.
## Where to go next
* [Using the editor](/guides/financial-planning/plan-design/using-the-editor): end to end.
* [Sections and tabs](/guides/financial-planning/plan-design/sections-and-tabs): building the spine.
* [Modules](/guides/financial-planning/plan-design/modules): adding and removing topics.
* [Block settings](/guides/financial-planning/plan-design/block-settings): the per-block controls.
* [Module reference](/guides/financial-planning/plan-design/module-reference): what every module needs.
* [Glossary](/guides/financial-planning/glossary): blocks, levers, personas and the rest.
# Sections and tabs
Source: https://docs.hazel.ai/guides/financial-planning/plan-design/sections-and-tabs
Top-level sections, the tabs inside them, and how to add, order and name each one.
You are laying out the destinations a client will click through, so start with the two levels a layout gives you: a **Section** becomes an entry in the plan's left navigation, and a **Tab** sits inside a section.
Blocks, the paragraphs, charts, tables and figure rows a plan is made of, sit in either one. Nothing nests deeper: a design that seems to want a third level wants another section.
Two blocks hold other blocks, **Tab group** and **Priority list**; a **Tab group** draws a real row of chips you can click between. The outline lists sections and tabs, never blocks, so look inside a container on the canvas.
## The default spine
Out of the box, a plan runs five sections in this order.
| Section | What it holds |
| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Summary** | The opening narrative, the chance-of-success gauge, and four headline figures. |
| **Today** | Net worth and the balance sheet, with account-level detail underneath. |
| **Goals** | One card per goal, with its funding status and progress toward the target. |
| **Getting there** | The working topics as a row of tabs: **Cash flow**, **Retirement** and **Insurance** to begin with. **Investments** and **Estate plan** are [modules](/guides/financial-planning/plan-design/modules) you add when you want them. Most of a plan lives here. |
| **Decisions** | What Hazel would do first, and the log of what was decided. |
Reorder them if your firm tells the story differently.
**Getting there** carries no blocks of its own by default, so Hazel writes whatever sits above the tab bar. A section can hold both: its own blocks lead, then the tabs.
## Adding a section or a tab
Structure is added in the layout editor, which opens from **Design layout** on the **Plan layout** card.
**A top-level section.** You build this one by hand: open the **Modules** tab in the left rail, expand **Building blocks** at the foot of the list (it starts closed), and drag **Section** from the **Structure** group onto the canvas. It arrives titled "New section", set to **Always**. No module makes a top-level section: the five section-shaped modules each add a tab to **Getting there** instead, and their cards wear a **Tab** badge.
**A tab.** Drag **Tab** from the same shelf into the section that should hold it, or add one of those five modules and let it bring its own tab and blocks. A tab that arrives with a module starts on **When it applies**; one you drag in starts on **Always**.
An empty tab works: give it a heading and **Instructions to Hazel**, and Hazel composes that tab from them.
## Ordering
Reorder from the **Outline** tab in the left rail. Hover a row and two arrows appear, moving it up or down among its siblings. Moving a block to a different parent is a drag on the canvas.
## Naming
Section and tab names are client-facing, so name them the way you would say them out loud. "Cash flow" beats "Cash flow and budgeting analysis."
Select a section and the **Editor** panel on the right offers:
* **Heading.** The words above the section's content.
* **Subtitle.** One optional line under the heading.
* **Nav label**, under **Advanced.** A shorter label for the left navigation. Hazel's own first section uses it: "Executive summary" above the content, **Summary** in the navigation.
A tab's **Heading** names the tab and heads its content. If the right name runs long, keep it and add a **Short tab name** under **Advanced**: the full heading leads the content, the short one rides the tab.
## Headings inside a section
You name the section, and Hazel writes the headings on the blocks inside it unless you pin them one at a time. Pin them where your firm has standardized. See [Block settings](/guides/financial-planning/plan-design/block-settings).
## A worked example
Say your firm presents tax planning as its own destination rather than a tab. Drag in a **Section**, set its **Heading** to "Tax planning", and move it after **Getting there**. Add **Roth conversion**, plus **RMD calculation** if you serve retirees: both land in the Retirement tab, so open that tab on the canvas and drag each block across. Put a **Narrative** block ahead of the schedules, then set the section's **When it shows up** to **Always** if it belongs in every deliverable.
## If a section you designed is missing from a plan
* **The design was never saved.** The save button still reads **Save changes**.
* **The plan predates the change.** A design shapes the plans built from now on.
* **The section is set to When it applies**, and this household's facts did not call for it. Switch it to **Always** if it belongs everywhere.
* **Strictness is doing what you asked.** **Flexible**, where a new firm starts, treats your design as a starting point. See [Strictness](/guides/financial-planning/settings/personas/strictness).
* **You edited a different design.** A plan is built from the base template or from the persona Hazel matched the household to, and each carries its own layout. See [Inheritance](/guides/financial-planning/settings/personas/inheritance).
## Where to go next
* [Modules](/guides/financial-planning/plan-design/modules): what to put in a section.
* [Block settings](/guides/financial-planning/plan-design/block-settings): headings, instructions, and visibility.
* [Using the layout editor](/guides/financial-planning/plan-design/using-the-editor): the rest of the mechanics.
* [Module reference](/guides/financial-planning/plan-design/module-reference): all nineteen modules and where each lands.
# Using the layout editor
Source: https://docs.hazel.ai/guides/financial-planning/plan-design/using-the-editor
The Design layout editor: the rails, the canvas, Sample data, adding and removing, undo, zoom, and saving.
You open **Design layout** to set what a plan contains and in what order, so every plan Hazel builds from then on comes out the way your firm presents advice. Nothing you do here reaches a plan until you save.
Only firm owners can open a plan design. The header names the one you are editing: **Plan design**, then **Base template** or the persona's name.
## What is where
The left rail opens on **Outline**: your design as a list of sections, tabs and pieces, with a search box above it. Beside it, **Modules** is the library of everything you could add.
The canvas draws your design the way a plan is drawn. Click anything to select it, and its settings open in the **Editor** panel on the right. With nothing selected, that panel describes the whole design: its strictness mode, how many of Hazel's nineteen planning modules it uses, and **See what's left out**.
**Strictness** shows in the header as a badge but is set on the **Plan design** card in **Settings → Planning**. See [Strictness](/guides/financial-planning/settings/personas/strictness).
## What fills the canvas
The **Sample data** switch in the header decides. Off, the default, every piece draws itself with stand-in content, and a piece with no heading of its own names itself at the top of its card. On, Hazel fills the design with a made-up example household, so you can read it roughly the way a client will. A band across the top names that household and confirms it is fictional, never a real client, and on a persona the example fits the persona.
Zoom sits beside the switch: **Fit width**, the default, reflows the design to your window, and the percentages from 50% to 150% magnify one fixed layout instead.
## Working in it
In the **Modules** tab, find what you want and press **Add to plan**. Each card says where it will land, and Hazel selects it and scrolls to it.
A module arrives set to **When it applies**, and the chevron beside the button adds it as always included instead.
Single pieces, and bare sections and tabs, come from the **Building blocks** shelf at the foot of that tab, the one place you drag from.
In the **Outline**, hover a section or tab row and use the up and down arrows that appear. To reorder pieces inside a section, or move one into another section, drag it on the canvas.
Select the piece and the **Editor** panel holds its settings: a **Heading**, **Instructions to Hazel** for that piece alone, **Leave out when**, **When it shows up**, and the rarely-touched extras under **Advanced**.
**Leave out when** disappears once a piece is set to **Always**. See [Block settings](/guides/financial-planning/plan-design/block-settings).
| To remove | How |
| ------------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| One piece | Select it, then **Remove from layout** in the Editor panel. |
| A whole module | The circle icon at the left of its row in the Outline, then **Remove from plan…**. Its other pieces go too, apart from any companion piece you have edited yourself. |
| A section, or a tab you added | The trash icon that appears when you hover its row in the Outline. A section's icon grays out with an explanation once only one section is left; a design keeps at least one. |
| A tab a module brought with it | No trash icon. Use the circle icon on that row and **Remove from plan…**, so the module's other pieces leave with it. |
Every removal shows a message naming what went, with an **Undo**. If the module has been customized here, Hazel asks first whether to save your version to your firm's module library.
The button at the top right reads **Saved** when there is nothing to save, **Save changes** when there is, and **Saving…** while it works. Until you press it, your work lives in this editor only: closing or refreshing the tab makes your browser ask first, and the **Planning settings** back link opens **Save your changes?**, offering **Keep editing**, **Discard and leave**, or **Save and leave**.
A module you create with **New module…** in the Modules tab does not wait for this button. It joins your firm's module library at once.
## Undo and redo
The two arrow buttons left of the save button, and Cmd+Z or Ctrl+Z outside text fields. They cover everything you do on the canvas. Escape clears the selection.
The **Undo** inside a message covers that one change, and only while nothing else has changed since. After another edit, Hazel says so and points you back at the header arrow.
## When a save is blocked
A red band under the header saying your firm's stored planning settings could not be read means Hazel is holding the save back on purpose, so saving cannot replace your real settings with defaults. Reload **Settings → Planning** and open the editor again. If the band comes back, report it.
Any other failure arrives as a message under the header: a request that timed out, a persona somebody deleted in another tab, or another owner saving in the same instant. In that last case, press save again, and Hazel re-reads the latest settings and writes only this design onto them.
Advisors who build plans cannot open or change your design, and saving writes only this one design, so it cannot touch an owner's work on a different persona. Two owners editing the **same** design will overwrite each other, so agree who has it first.
## Editing a persona that follows the base
If a persona has no design of its own, the editor opens the base template's design and the chip in the header reads **Follows base template**, and nothing changes for the persona while you look around. Press save and Hazel asks first whether to give that persona its own design. Say yes and the version on screen becomes the persona's own, stops picking up later changes to the base, and the chip reads **Customized from base**. Editing a chip on the **Plan design** card does the same at once, since that card saves as you go. See [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
## Where to go next
* [Sections and tabs](/guides/financial-planning/plan-design/sections-and-tabs): the structure you are arranging.
* [Modules](/guides/financial-planning/plan-design/modules): what to add, and what leaving one out means.
* [Block settings](/guides/financial-planning/plan-design/block-settings): the per-piece controls.
* [Strictness](/guides/financial-planning/settings/personas/strictness): how closely Hazel follows the design.
# The base template
Source: https://docs.hazel.ai/guides/financial-planning/settings/base-template
Your firm's default plan: your plan shape, house instructions, firm-specific planning assumptions, and the guardrails no plan may cross.
Set the base template once, so every plan Hazel builds already carries your firm's shape, voice, numbers and rules. The **Base template** is your firm's default plan: what a household's plan looks like when all you know is that they are your client.
It lives in **Settings**, under **Planning**, first under **Plan templates**. Everything here applies firm-wide, and every Client persona is built on it, so a persona carries only what differs.
## What the base template holds
The blueprint: sections, tabs, modules, and what each block is told to do.
Your house guidance in plain English, as a numbered list. It ships empty. Select **Add instruction** to start.
Label and value rows for firm-specific figures the Solver and Market panels do not model.
The never-do rules, firm-wide. A persona can add its own, and no persona, plan, or request can remove one.
Two more boxes sit below those four:
* **Client disclaimer**: the footer clients read under a plan and in the exported PDF ([Sharing and delivering](/guides/financial-planning/the-plan/sharing-and-delivering)). It opens on Hazel's plain default wording, and nothing reaches a plan until you edit the box.
* **Used by**: the personas still following this design, so a change here reaches all of them. A persona with its own plan design drops off it.
Every edit saves itself, with "Saving…" then "Saved" beside the heading. Changes shape the plans Hazel builds from now on; plans you have already delivered stay as they are.
Instructions are the one part a persona does not add to: once Hazel matches a household to a persona, Hazel follows that persona's list, not the base list. Plan design, planning assumptions and the assumption numbers carry through until a persona changes them; guardrails carry through and stay. See [Inheritance](/guides/financial-planning/settings/personas/inheritance).
## Plan design
The upper card carries the **Strictness** dial, plus a **Hazel default** badge while you are still on the shipped design, which places eight of Hazel's nineteen planning modules with strictness **Flexible**. See [Strictness](/guides/financial-planning/settings/personas/strictness).
**Plan layout** below it draws your blueprint: sections as bands, tabs as indented sub-bands, blocks as labeled chips. A line under it counts your sections, the modules in play and the always-on blocks, then ends in a **See what's left out** link into the layout designer's **Modules** panel.
Two ways to work with it:
* **Select any chip** to edit it in place: **When it appears**, **Instructions to Hazel**, **Leave out when**, then **Done**. No layout designer needed.
* **Select Design layout** for the layout designer, where you add, remove, and reorder modules and blocks. See [Using the editor](/guides/financial-planning/plan-design/using-the-editor).
## Planning assumptions
For example:
| Label | Value |
| -------------------- | ----- |
| Longevity age | 95 |
| Healthcare inflation | 5.5% |
| Education inflation | 5% |
Select **Add planning assumption**, give the row a **Label** and a **Value**, and it joins every plan built on this template. The value is free text: a number, a percentage, or a phrase such as "Stop under the first IRMAA tier".
These rows are firm-wide, with no per-row override on a persona. When one kind of client needs a different figure, write it as an instruction on that persona in the same label and value shape, for example "Education inflation: 6%". For a single plan, ask Hazel in that plan's chat.
"Longevity age: 95" belongs in assumptions; "Frame longevity as a spending question, not a portfolio question." belongs in instructions. Where the two disagree on a number the assumption wins, unless the instruction uses the label-and-value shape above.
## The order to set it up
Your base template arrives with a plan design and the firm-wide solver and market numbers set, but no instructions. Build a plan for a household you know well, from that household's **Financial Planning** tab or the **Financial Planning** page, and read it as a client would. See [Getting started](/guides/financial-planning/overview/getting-started).
Hazel matches each household to the closest of the three Client personas your firm already has. The plan's **Appendix**, under **Applied settings**, names the persona it was built from, or none if the base template alone was used, so you edit the screen that shaped what you read.
Anything structurally missing or unwanted goes in Plan design.
Add instructions for what you would have written differently. See [Instructions](/guides/financial-planning/settings/personas/instructions).
Set [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) and [Market assumptions](/guides/financial-planning/assumptions/market-assumptions), then add any firm-specific planning assumptions the panels do not model.
Write down the rules your firm would never break, however a request is phrased.
Create a persona only where a kind of client needs a different shape or different numbers. See [Creating a persona](/guides/financial-planning/settings/personas/creating-a-persona).
Test a change by building a plan. The sample plan reads the same whatever you change here. If the household already has a plan, **Regenerate** in the top bar of its **Financial Planning** tab rebuilds it on your current settings.
## Where to go next
* [Plan design](/guides/financial-planning/plan-design/overview)
* [Client personas](/guides/financial-planning/settings/personas/overview)
* [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions)
# Planning settings
Source: https://docs.hazel.ai/guides/financial-planning/settings/overview
Set how your firm builds plans once, in Settings → Planning: Plan templates, Client personas, and the two assumption editors.
Decide once how your firm builds plans, so no advisor rebuilds the same choices household by household. **Settings → Planning** sets what a plan contains, in what order, in your words, and against which numbers.
Only firm owners can open this tab. An advisor still has room on one household, adjusting most numbers in that plan's own **Plan settings** panel in the chat view, within the ranges you set. See [Access and availability](/guides/financial-planning/overview/access-and-availability).
## The four panes
The left rail groups them under **Plan templates**, **Client persona**, and **Assumptions**.
Your firm's default plan. Every Client persona is built on it.
Your client archetypes. Your firm has three, installed the first time anyone opens Planning settings. Hazel matches each household to the closest one.
The planning inputs behind every figure, in three cards: **Probability & simulation**, **Fees & taxes**, and **Decumulation strategy**.
The figures every plan is built against, under **Returns**, **Volatility**, **Inflation** and **Taxes**.
Two more settings sit on the **Base template** pane, below the instructions list.
**Guardrails.** The rules your firm will not break, however a request is phrased. A persona can add one, never remove one. See [Guardrails](/guides/financial-planning/assumptions/guardrails).
**Client disclaimer.** The footer on every plan, on your screen and in the PDF you hand over. The card opens with wording Hazel has drafted, which is not in force until you edit and save it, so save it once or your plans publish with no footer. **Restore Hazel default** brings the drafted wording back.
## Which layer wins
A plan tags every planning value with the layer that set it, and Hazel ships a default for every value, so anything you have not changed reads as **Firm**. The tags are in the plan's **Appendix**, in its [Applied settings block](/guides/financial-planning/your-data/where-numbers-come-from).
| Layer | What sets it |
| ------------- | ------------------------------------------------------------------------------------------------------------------------------------------- |
| **Firm** | The base template, your **Hazel instructions**, your guardrails, and the two assumption editors |
| **Persona** | The matched persona's own instructions, its own plan design if you give it one, and any planning values it wants different from your firm's |
| **This plan** | An advisor's adjustment for one plan, staged in that plan's **Plan settings** panel in the chat view |
A per-plan adjustment belongs to that plan alone, so a second plan for the same household still inherits your firm and persona values. [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) covers the panel and the few values it leaves out.
Numbers merge one value at a time, but shape and wording replace: a persona with its own plan design stops following the base design, and its instruction list stands in for the base list rather than adding to it. See [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
**A number beats a sentence.** Where an instruction and an assumption disagree about a figure, the value set in **Solver assumptions** or **Market assumptions** wins.
## Where to change what
| You want | Change this |
| ------------------------------------------------ | ------------------------------------------------------------------------------------- |
| A section or module in every plan | **Plan design** on the base template |
| Hazel to follow your design more or less closely | **Strictness**, on the Plan design card |
| Every plan to lead with the answer | **Hazel instructions** |
| A different success threshold or reserve length | **Solver assumptions** |
| Different return or inflation figures | **Market assumptions** |
| A change for one plan only | That plan's **Plan settings** panel in the chat view, or ask Hazel in the plan's chat |
If you ask Hazel for the same change on every plan, that change belongs in settings.
## How saving works
There is no Save button on these panes. An edit saves when you leave the field, so click or tab out of a box before closing the tab, and watch for "Saving…" then "Saved" beside the **Plan templates** heading.
Two exceptions: the **New persona** dialog writes nothing until you select **Create**, and the layout designer, which opens from **Design layout**, has its own **Save changes** button and asks before you leave with work unsaved.
If your planning settings changed somewhere else since your screen loaded, in another tab, in the designer, or by a colleague, Hazel stops saving and shows a red notice asking you to reload. What you typed after that is discarded, so note it down first.
## When a change takes effect
A settings change shapes the next plan Hazel builds, and the next change to a plan that already exists; plans you have already delivered stay as they are.
To move an existing household onto a new shape, open that household's **Financial Planning** tab and select **Regenerate**. That adds a version rather than replacing what your client has seen. See [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating).
The **Client disclaimer** is the exception. Hazel reads it fresh each time a plan is opened, so editing it also changes the footer on plans you published earlier.
## Where to go next
Start with the [base template](/guides/financial-planning/settings/base-template), then [Client personas](/guides/financial-planning/settings/personas/overview) and [Plan design](/guides/financial-planning/plan-design/overview) in full.
# Creating a persona
Source: https://docs.hazel.ai/guides/financial-planning/settings/personas/creating-a-persona
Add a Client persona of your own: pick a starting template, name it, write the description Hazel matches households against, then give it instructions and a plan design.
You plan for the same kinds of clients again and again. Create a persona for one of them, so every plan Hazel builds for that client type starts from your firm's thinking about it.
Personas live in **Settings → Planning**, under **Client persona** in the left nav, and **New persona** adds one. Planning settings are the firm owner's surface.
## Before you start
Have a real client in mind, not an abstraction: "the Okonkwos, and the four households like them". Settle your [base template](/guides/financial-planning/settings/base-template) first, since a persona is defined by how it differs from the base.
## Create it
The **New persona** dialog has four fields.
* **Base template**, the default, copies your firm's Hazel instructions in and leaves the plan design following the base.
* A **Hazel preset** copies that archetype's instructions, the planning figures it sets and its own plan design, and fills in the name, description and age range for you to review. It brings none of your house instructions, so add back any you rely on after creating.
* **Copy from** an existing persona takes everything that persona has, as an independent copy. Editing either one afterwards leaves the other alone.
A preset appears here only when no persona of that name exists, so with [Hazel's three](/guides/financial-planning/settings/personas/default-templates) installed the presets group is usually empty.
Short and recognizable, in your firm's own language: Business Owner, Physician Household. Up to 80 characters, with a counter as you approach it. **Base** and **Base template** are refused, and so is a name close to one you already have: Pre Retiree alongside an existing Pre-Retiree is refused as too close.
This is what Hazel matches households against, as the field says underneath it.
**Age min** and **Age max** are optional and take anything from 0 to 120, with the minimum no higher than the maximum. Hazel will not refuse a better match on age alone. The persona's own pane carries no age field, so "typically late fifties" goes in the description instead.
Select **Create**. The persona saves right away and its pane opens.
## Then finish it
The ordered list of guidance Hazel follows for this client type. It stands alone rather than adding to the base template's, so edit it down to what this type needs. See [Instructions](/guides/financial-planning/settings/personas/instructions).
A new persona follows the base design until you change something here, and the card says which state it is in. The strictness control sits on the same card. See [Plan design](/guides/financial-planning/plan-design/overview), [Strictness](/guides/financial-planning/settings/personas/strictness) and [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
Read the finished plan for the household you had in mind, and check the persona named in the **Applied settings** block of its **Appendix**. Adjust, then read again.
Edits here save as you make them, with "Saving…" then "Saved" beside the **Plan templates** heading. **Delete** removes a persona immediately and cannot be undone, though deleting one of Hazel's three returns it to the **Starting template** list.
## Writing a description that matches well
Write a situation, not a label. Include the stage of life or career, the largest asset, the central planning question, and what sets this type apart from your others.
"Business owners."
Nothing to match on: every self-employed household fits equally well.
"Owner-operator of a closely held business, five to fifteen years from an exit. The business is the largest asset, so liquidity, concentration and the pre-exit tax picture lead."
## Numbers, and where they are set
A persona's pane carries no number fields. The planning figures are firm-wide, on two screens under **Assumptions**: [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) for the success threshold, the cash reserve, fees and the default Roth conversion strategy, and [Market assumptions](/guides/financial-planning/assumptions/market-assumptions) for returns, volatility, inflation and tax rates.
A persona can still carry values of its own, and Hazel's three do, from HENRY's 85% success threshold to Retiree / Decumulator's seven-year cash reserve. See [Default templates](/guides/financial-planning/settings/personas/default-templates). A copy of one of those personas keeps its values, and **Applied settings** shows them attributed to **Persona**.
The advisory fee, the terminal heir tax rate and the number of Monte Carlo simulations never vary by persona.
For one household rather than a whole client type, change the figure on the plan itself: select it, edit the assumption behind it, and recalculate. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly). For a whole client type, set the firm-wide figure to what most of your book should use and handle the exceptions on the plan.
The never-do rules live on the base template, apply to every plan Hazel builds, and nothing on a persona loosens them. See [Guardrails](/guides/financial-planning/assumptions/guardrails).
## Common mistakes
* **A persona that does not differ from the base.** Base list unchanged, plan design still following the base: give it real differences, or delete it.
* **Two personas that overlap.** If both descriptions fit the same household, sharpen what separates them or merge them.
* **A plan design forked before the base is settled.** A persona with a design of its own stops picking up base improvements. See [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
# Default templates
Source: https://docs.hazel.ai/guides/financial-planning/settings/personas/default-templates
The three Client personas your firm starts with: HENRY, Pre-Retiree, and Retiree / Decumulator. What each one emphasizes, what it changes, and which ones to keep.
You already have three Client personas, so your first decision is which ones to keep. **HENRY**, **Pre-Retiree**, and **Retiree / Decumulator** are in **Settings → Planning** the first time you open it, and Hazel matches households against all three from the start.
They are ordinary firm personas from that moment: rename them, rewrite their instructions, redesign their plans, or delete them.
When a persona below leaves a topic out, no part of its plans is devoted to that topic on its own. Hazel still covers it where it belongs in the plan, and can still answer a question about it in chat.
## HENRY
High earner, not rich yet. Ages 28 to 44.
Its instructions tell Hazel to:
* Lead with the savings-rate target and how it moves the retirement date.
* Call out concentrated single-stock risk when restricted stock units or an employee share purchase plan appear.
* Show tax-advantaged headroom, such as 401(k), mega-backdoor, and HSA capacity, before taxable saving.
* Compound efficiently toward competing goals at once: a house, college, financial independence.
* Show what each extra dollar of savings buys.
What it changes:
* Its plan design is Hazel's default set of modules minus **Monte Carlo success**.
* It sets its own success threshold of 85%, where Hazel's firm-wide default is 90%.
* Its design ships at **Flexible** strictness.
Keep it if you serve high-income professionals whose plans turn on how much they can save.
## Pre-Retiree
Five to ten years out. Ages 55 to 67.
Its instructions tell Hazel to:
* Anchor to the household's stated retirement year and frame every trade-off around it.
* Model Social Security at both 67 and 70 for each earner.
* Stress-test spending at a 4% real return.
* Surface Roth conversion capacity in the 12% and 22% brackets.
* Answer a question like "can I retire at 60?" with a defensible number.
* Build and lock the reserve ladder so rebalancing does not erode the floor.
What it changes:
* Its plan design is Hazel's default set of modules minus **Debt paydown**, usually resolved by this age.
* It holds a five-year cash reserve, against Hazel's default of three years.
* It sets the Roth conversion strategy to **Fill to top of 24% bracket**, with a note to fill the 22% bracket between ages 60 and 72.
* Its design ships at **Flexible** strictness.
Keep it if your clients are asking when they can retire.
## Retiree / Decumulator
In retirement or transitioning. Ages 65 and up.
Its instructions tell Hazel to:
* Fund spending from taxable accounts first, then tax-deferred, then Roth, unless the household says otherwise.
* Call out the year required minimum distributions start, for each earner.
* Show healthcare cost inflation at 5.5% separately from general inflation.
* Show a band of modeled outcomes rather than a single success figure.
* Draw down tax-efficiently without running out.
* Smooth income to avoid bracket creep and IRMAA cliffs, the income thresholds that raise Medicare premiums.
What it changes:
* Its plan design is built for the income phase, so it leads the tabs inside **Getting there** with Retirement income, then Cash flow, Investments, and Insurance.
* It drops the **Retirement**, **Emergency fund**, and **Debt paydown** modules, and adds **Retirement income sequencing**, **Healthcare coverage estimation**, and **Investments**.
* It holds a seven-year cash reserve, the longest of the three, against Hazel's default of three years.
* It replaces the firm's withdrawal guardrail wording with its own: Guyton-Klinger plus or minus 10%, with a 20% ceiling in goal years.
* Its design ships at **Guided** strictness rather than Flexible.
Keep it if you serve retirees, and expect its plans to look structurally different from the other two.
## Deciding which ones to keep
An unused persona still competes to match. Open it in **Settings → Planning** and select **Delete**. It leaves your firm's Plan templates straight away and cannot be undone, but the preset stays available: select **New persona** and pick it from **Hazel presets** in the **Starting template** list.
Pick a household you know well and read the finished plan, then check the Persona row in **Applied settings**, in the plan's **Appendix**.
The description is the main matching signal, and the field says as much. Rewrite it in your firm's words, for the clients you have. The age range under the name is fixed at creation.
Small, specific changes. If you are rewriting a persona wholesale, your base template is probably what needs fixing.
All three shipped personas carry a plan design of their own, so each reads **Customized** on its **Plan design** card and offers **Re-sync from base**. They will not pick up a base-template change until you re-sync.
## Where to go next
* [Creating a persona](/guides/financial-planning/settings/personas/creating-a-persona): building one of your own.
* [Instructions](/guides/financial-planning/settings/personas/instructions): writing instructions Hazel can act on.
* [Strictness](/guides/financial-planning/settings/personas/strictness): what Exact, Guided, and Flexible do.
* [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance): how a persona relates to the base template.
# Inheritance and re-syncing
Source: https://docs.hazel.ai/guides/financial-planning/settings/personas/inheritance
How to tell whether a persona follows your base template or has a design of its own, and how to put it back on the base.
You put work into your firm's base template, so you need to know which Client personas will pick it up.
## Follows base, or Customized
No design of its own, so the persona uses your firm's current base design, live. On the **Plan design** card the blueprint is drawn faded, a **Follows base** badge sits beside the heading, and the line under the strictness control reads "This persona follows the base template".
Its own design replaces the base, so base-template changes do not reach this persona. The line under the strictness control reads "Customized", and a **Re-sync from base** button appears below it.
## Where your firm starts
The three personas your firm arrives with each ship with a plan design of their own, so all three read **Customized** before you touch anything. The left nav of **Settings → Planning** shows it: under **Plan templates**, the subtitle beneath **Base template** reads "No persona follows it right now", and the base pane's **Used by** card is hidden entirely.
A persona follows the base when you create it with **Starting template: Base template**, or when you re-sync one of the three.
A plan for a household that matched no persona uses the base design, and every persona you create from **Base template** starts on it.
## What gives a persona its own design
A persona takes its own copy on the **first** change to its design:
* Editing a block on the **Plan design** card: **When it appears**, **Instructions to Hazel**, or the condition for leaving the block out.
* Changing **Strictness** on that card.
* Saving in the layout designer, which you open with **Design layout**. Adding, removing, reordering, and the fields in the settings rail all count.
On the **Plan design** card the change lands as soon as you make it, because Planning settings autosave. A notice appears the first time, reading "This persona now has its own plan design." and "It no longer follows the firm's base design. Re-sync from base any time."
In the layout designer nothing is committed until you save, and that first save asks you first: a dialog says the persona currently follows the firm's base template and that saving gives it a design of its own, with **Keep editing** and **Save design** as the choices. The header chip reads **Follows base template** before that save and **Customized from base** after it.
## What does not follow the base
**Hazel instructions.** A persona's instruction list is its own from the moment the persona exists: creating a persona from **Base template** copies your base instructions across once, and later edits to the base list do not follow. A house instruction you rely on has to appear on each persona.
**Assumptions.** The planning figures live on two firm-wide screens, **Solver assumptions** and **Market assumptions**, and personas inherit them value by value: a persona carries only the values that differ for it, so changing a firm figure reaches every persona that does not set that value. The never-do rules on your base template apply to every plan too, and nothing on a persona loosens them.
## Re-syncing
**Re-sync from base** sits on the persona's **Plan design** card, under the strictness control, and only while the persona reads **Customized**. It puts the persona back on your current base design and keeps it there until the next change. A confirmation dialog headed "Re-sync from base?" comes first.
Re-sync discards rather than merges: anything the persona's design carried that your base does not have goes with it. Instructions and assumption values are untouched, since neither lives in the design.
Move anything you want to keep into the base template before you re-sync. There is no undo for it.
### When to re-sync
* You improved the base and want this persona on the new shape.
* The persona's design has become a pile of one-off edits rather than a shape you chose.
* The persona took its own design by accident.
Leave it customized while its design is doing real work, or when you want one thing from the base and nothing else: make that change on the persona directly.
## If a base change did not reach your plans
Open the plan's **Appendix** and read the **Applied settings** block: it names the persona the plan was built from, or shows none.
If it reads **Customized**, your base change never reached it. Re-sync it, or make the same change on the persona.
Settings changes apply to plans Hazel builds from then on. A delivered plan keeps the design it was built with until you rebuild it with **Regenerate**, in the top bar of the household's **Financial Planning** tab.
## Where to go next
* [Base template](/guides/financial-planning/settings/base-template): the design a persona inherits.
* [Default templates](/guides/financial-planning/settings/personas/default-templates): the three personas your firm has.
* [Plan design](/guides/financial-planning/plan-design/overview): where a design is edited.
* [Strictness](/guides/financial-planning/settings/personas/strictness): Exact, Guided and Flexible.
# Instructions
Source: https://docs.hazel.ai/guides/financial-planning/settings/personas/instructions
Your firm's house guidance for how Hazel writes plans, as an ordered list: what belongs in it, which list runs when a persona matches, how to write entries Hazel can act on, and how the editor behaves.
You want every plan to read the way your firm would have written it, so you give Hazel that guidance once and it applies on every build. **Hazel instructions** is that list, on two panes in **Settings**, then **Planning**: **Base template**, and every persona.
Each entry is one short statement in plain English: what to lead with, what to always call out, voice and length. The rows are numbered, and the field says "Hazel follows the list in order, top to bottom," so put what matters most first.
## What instructions cannot do
| Do not write this in instructions | Set it here instead |
| ------------------------------------- | ------------------------------------------------------------------------------- |
| "Assume 4% real returns" | [Market assumptions](/guides/financial-planning/assumptions/market-assumptions) |
| "Use a 90% success threshold" | [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) |
| "Hold five years of cash" | Solver assumptions, as the default reserve length |
| "Never recommend annuities" | [Guardrails](/guides/financial-planning/assumptions/guardrails) |
| "Always include an insurance section" | [Plan design](/guides/financial-planning/plan-design/overview) |
The last two rows catch people out: an instruction is guidance Hazel weighs, so anything that must happen on every plan belongs in a guardrail or in the layout.
If an instruction and **Solver assumptions** or **Market assumptions** state different numbers, the assumption wins. For a different number on one household rather than the whole firm, change the assumption on that plan and recalculate. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
## Which list Hazel follows
Hazel picks the persona closest to the household, and that match decides which single instruction list runs.
* **A persona matched.** Hazel follows that persona's instructions, not the base template's.
* **Nothing matched.** Hazel follows the base template's instructions.
Your firm starts with three personas, each with its own instructions, so most households match one and the base list is the fallback. Anything you want on every plan goes on the base template and on each persona, or becomes a guardrail instead: guardrails and **Planning assumptions** rows are firm-wide, with no persona-level copy, and hold whichever persona matched.
**Starting template** gives a new persona your base list, or another persona's with **Copy from**. That copy is taken once, at creation, so a base instruction you write next month has to be added to each persona by hand.
## Writing instructions Hazel can act on
| Vague | Actionable |
| ---------------- | -------------------------------------------------------------------------- |
| "Be concise" | "No section longer than one page. Cut adjectives." |
| "Focus on taxes" | "Show the bracket each conversion year fills, and name the marginal rate." |
* **Say it once, and keep the list short.** Repeating an instruction in other words does not strengthen it. Nothing caps how many you add, but three to eight is a good target.
* **Resolve contradictions yourself, and keep each entry to a sentence or two.** A row that runs long is usually two instructions.
## Using the editor
**Adding.** Select **Add instruction** and type. Enter commits the row and leaves the next open; clicking away also commits, and Escape discards. One row holds one instruction, so Enter never starts a second line and line breaks in pasted text become spaces.
**Editing and order.** Select any row and type; Enter or leaving the row saves it. A row you clear and click away from disappears. There are no drag handles, and a new instruction lands at the bottom, so decide the order as you write and retype rows to change it.
**Deleting.** The X at the end of a row removes it straight away, and a message headed "Instruction removed" offers **Undo**. Undo works until you change the list again; after that the message says to add the row back yourself.
**Length.** One instruction can run to 400 characters, with a counter as you type and typing stopping at the limit. A paste that would push a row over is refused rather than trimmed, with a note under the field: "Over the 400-character limit for one instruction, so it wasn't pasted. Split it into separate instructions." Your text is still on the clipboard.
**Saving.** **Saving…** and then **Saved** appears beside the **Plan templates** heading as each edit lands.
## If an instruction did not land
* **Check which list ran.** Open the plan's **Appendix** and read the **Persona** row in **Applied settings**. A persona named there is the list Hazel followed, not your base list.
* **Check you have not written a number.** Set it in Solver assumptions or Market assumptions instead.
* **Check whether you asked for structure.** A section your plan design does not include only appears if the design's strictness lets Hazel add to it. Under **Exact** it cannot. See [Strictness](/guides/financial-planning/settings/personas/strictness).
* **Check it is specific.** "Be thorough" gives Hazel nothing to do. Say what to include, where it goes, and in what order.
## Where to go next
* [Base template](/guides/financial-planning/settings/base-template): where your firm's list lives, next to guardrails.
* [Creating a persona](/guides/financial-planning/settings/personas/creating-a-persona): choosing a new persona's starting list.
* [Inheritance](/guides/financial-planning/settings/personas/inheritance): what a persona takes from the base template.
* [Block settings](/guides/financial-planning/plan-design/block-settings): the **Instructions to Hazel** field for one part of the plan.
# Client personas
Source: https://docs.hazel.ai/guides/financial-planning/settings/personas/overview
Describe the client types your firm plans for once, so Hazel matches each household to the right one and builds the plan that fits.
You plan differently for an accumulator in their thirties, a couple three years out from retirement, and a household already drawing income. Describe each of those client types once, as a **Client persona**, and Hazel matches every household to the closest one and builds the plan accordingly.
Personas live in **Settings**, then **Planning**, under **Client persona** in the left rail. Your firm already has three, installed the first time anyone opens Planning settings. Planning settings are open to firm owners; see [Roles and permissions](/guides/settings/members/roles-and-permissions).
## What a persona holds
| Field | What it does |
| ---------------------- | -------------------------------------------------------------------------------------------------------------------------------- |
| **Name** | What you call this archetype. Unique within your firm |
| **Description** | Who this persona is for, and what their planning centers on. Hazel matches households against this |
| **Age band** | An optional matching hint, such as 55 to 67. A hint, never a hard rule |
| **Hazel instructions** | An ordered list of guidance Hazel follows for this client type |
| **Plan design** | This persona's layout, when it differs from the base template |
| **Solver assumptions** | Planning values this client type uses instead of your firm defaults, such as **Success threshold** or **Default reserve length** |
Open a persona to edit its name, description, **Hazel instructions** and **Plan design**, or to delete it.
**The age band is set while you create the persona.** The **New persona** dialog has **Age min** and **Age max** fields, both optional, and the persona screen afterwards has no age control. The persona's row in the left rail shows the band.
**Solver assumptions are not edited on the persona screen.** The three personas your firm starts with carry their own, such as HENRY's 85% success threshold. Change a planning value for the whole firm in **Solver assumptions**, or on a single plan by editing the assumption behind the figure it drives. A persona you create with **Starting template** set to a Hazel preset, or to **Copy from** an existing persona, starts with that persona's solver assumptions.
Read [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) for what each value does and its default, and [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from) for editing an assumption on one plan.
## How the match happens
At the start of a build, Hazel picks the closest fit, weighing two signals together rather than in order:
* **The description.** What the persona says it is for, held against what Hazel knows about this household, including the goals it has stated.
* **The age band.** All else being equal, Hazel matches a 58-year-old to a persona banded 55 to 67. A better fit elsewhere still beats a match on age alone.
Both read from what Hazel already knows, so a thin profile matches poorly. See [How Hazel uses the profile](/guides/financial-planning/your-data/how-hazel-uses-the-profile).
If nothing fits well, Hazel builds on the **base template** alone and records that no persona was used. That usually means you are missing a persona for a client type you serve.
### Hazel records the match
Open the plan's **Appendix** and read the **Applied settings** block: a **Persona** row names the matched persona, or reads **None**, and each planning value under it carries the layer that set it. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
The record is stamped when that version is published, so an older version shows what it was built with rather than today's settings.
## Overriding the match
Name the persona you want up front:
```text theme={null}
Build a plan for the Okonkwos using the Business Owner persona,
not the pre-retiree one. He is 58 but the exit drives everything.
```
Or name it afterwards on a built plan, and Hazel rebuilds against it.
A persona that keeps getting mismatched needs a sharper description. Rewrite it to say plainly who the persona is for.
## What a persona inherits
A persona is built on the base template and follows it until you change something. Its solver assumptions work value by value: the persona carries only what differs and inherits the rest from your firm settings. Its plan design follows the base design live until the first edit on its **Plan design** card, which forks a copy for that persona. See [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
Guardrails are firm-wide, and a persona can add one but never remove one. See [Guardrails](/guides/financial-planning/assumptions/guardrails).
## Where to go next
* [Default templates](/guides/financial-planning/settings/personas/default-templates): the three your firm already has, and what each one does.
* [Creating a persona](/guides/financial-planning/settings/personas/creating-a-persona): build your own, field by field.
* [Instructions](/guides/financial-planning/settings/personas/instructions): how to write guidance Hazel follows.
* [Strictness](/guides/financial-planning/settings/personas/strictness): how tightly a persona's layout binds Hazel.
# Strictness
Source: https://docs.hazel.ai/guides/financial-planning/settings/personas/strictness
How closely Hazel follows your plan design: Exact, Guided, or Flexible, and how to tell which setting a plan was built on.
You put real time into a plan design, so you need to know how closely Hazel will follow it. **Strictness** decides that.
It is set per design, not per firm: your base template has its own, and so does every Client persona with its own design. Find it on the **Plan design** card in **Settings → Planning**, in a panel headed **Strictness** showing the current setting and its one-line consequence. Select it to change.
## The three settings
| | Exact | Guided | Flexible |
| ---------------------------------------------------- | ------------------------------ | ------------------------------------------------------------------------- | -------------------------------------------------------------------- |
| On screen | Plans must match this design | Hazel follows this design, adapting where a household's facts call for it | Your design is a starting point, and Hazel composes each plan freely |
| Must include every **Always** block | Yes, checked before publishing | Yes, unless the household's facts say otherwise | Treated as a strong preference |
| May add a section or block your design does not have | No | Yes, unless you turn **Allow additions** off | Yes |
| May leave a block out | Not an **Always** one | Yes, when the facts call for it | Yes |
## Check which one you are on
A new firm starts on **Flexible**: the base template, and two of the three Client personas your firm arrives with. Only **Retiree / Decumulator** ships on **Guided**, so if plans keep arriving in a shape you did not ask for, look here first.
Guided is where to move once you have a design you trust.
## Allow additions
Below the three settings sits a single switch, **Allow additions**, described on screen as Hazel adding sections or blocks you did not design. It starts on. Turn it off under **Guided** when you want your shape followed and nothing extra.
Under **Exact** the switch turns off and grays out, whatever it was set to before. Move that design back to **Guided** or **Flexible** later and it is still off, so turn it back on if you want Hazel covering ground your design does not.
There is no matching switch for leaving things out; that is each block's own setting.
## Choosing
| Pick | When |
| ------------ | -------------------------------------------------------------------------------------------------------------------------------------------- |
| **Guided** | You want your shape respected, with room for Hazel to notice what is different about a household: no children, so skip the education section |
| **Exact** | The design is a commitment: a deliverable format your firm has standardized on, or a shape a compliance review has approved |
| **Flexible** | The design is a preference, while your firm is still working out its template, or on a persona covering households with little in common |
Exact checks the finished plan against your design and sends it back to be fixed, though now and then one still publishes with something not matching, so read the first few plans built on an Exact design before you present one. An **Always** block covering something a household's facts cannot support gets produced thin rather than dropped.
## Strictness and each block's own setting
Strictness governs the whole design; each block also carries its own setting for whether it belongs in a plan. Select a chip on the **Plan design** card and the first field is **When it appears**.
| Setting | What it means |
| ------------------- | -------------------------------------------- |
| **Always** | In every plan you build. |
| **When it applies** | Only when the household's facts call for it. |
The same control is labeled **When it shows up** in the layout editor. An older design may show a third choice, **Suggested (legacy)**, which Hazel may skip freely. Open that block in the layout editor and the field says the setting is retiring and asks you to pick one of the two above. The third choice then disappears from both screens.
A block set to **When it applies** stays the household's call under every setting, Exact included.
See [Block settings](/guides/financial-planning/plan-design/block-settings).
## Telling whether your strictness is right
Open a plan, look at **Plan sections** in the left nav, and compare it with the chips on your **Plan design** card. A section you designed that is not there, or a section you did not design that turned up, is Hazel's read of that household.
Check which design the plan was built on first: the **Applied settings** block in the plan's **Appendix** names the persona Hazel matched, or reads **None**. A persona with its own design held that plan to that persona's strictness, and otherwise your base template's applied.
Then act on what keeps repeating:
* The same section keeps disappearing. Your design is asking for something your households cannot support. Move that block to **When it applies**, or take it out.
* Hazel keeps adding the same topic. Add it to your design so you control how it is written, or turn **Allow additions** off.
## Where to go next
* [Block settings](/guides/financial-planning/plan-design/block-settings): the per-block controls, including when a block appears.
* [Plan design](/guides/financial-planning/plan-design/overview): the design that strictness applies to.
* [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance): changing strictness on a persona that follows the base template gives that persona its own copy.
# Anatomy of a plan
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/anatomy-of-a-plan
What each part of a plan does: the rail, the five sections, the top bar on both surfaces, the Appendix, plan settings, the staging bar, and feedback.
You are going to walk a client through one long document, so you need to know what each part does and which controls you get where.
## The navigation rail
Above the section list, the rail carries the plan's name, the household with its members, and the date the plan was written. Selecting a section scrolls to it and expands it. **Appendix** sits apart below the list.
## The five default sections
| Section | What it holds |
| ----------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Summary** | The narrative opening, four headline figures, and the success gauge: paired rings comparing the chance of success as things stand with the chance once the plan's recommendations are in place |
| **Today** | Where the household stands now: net worth, the balance sheet, account-level detail |
| **Goals** | One card per goal, with its funding status and progress toward the target |
| **Getting there** | The analysis itself, split across tabs |
| **Decisions** | What Hazel would do first, and the log of what you approved or ignored |
Your firm can change this shape in [Plan design](/guides/financial-planning/plan-design/overview), and each Client persona can carry one of its own, so a plan may not have all five sections.
### Getting there and its tabs
Most of the plan lives here, usually leading with **Your priorities**: one card per priority Hazel ranked for this household. Select a card and its detail opens underneath: the recommendation, the figures behind it, and the dials that move it.
The tabs after that come from the modules your firm has enabled: **Cash flow**, **Retirement** and **Insurance** out of the box, with **Investments** and **Estate plan** once your firm adds them. See [Modules](/guides/financial-planning/plan-design/modules).
## The top bar
On the left: the plan's title, the version you are reading, and an **N unsaved** badge whenever you have staged work. The controls on the right depend on where the plan is open, in **Ask Hazel** or on a household's **Financial Planning** tab.
| Control | What it does | Where |
| --------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------- |
| **Plan version** | Steps back and forward through the plan's history, reading as "V3 (2 of 4)". An older version is read-only, under a banner with **Back to latest**. One version means a plain badge and no arrows. [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating) | Both |
| **Plan orientation** | The info button opens a short generated brief: the household's situation and what the plan concludes, the trade-offs the analysis weighs, where the goals stand, up to four recommendations with their figures, and a closing **Start here** line naming what to settle with the client first | Both |
| **Export PDF** | Turns the version on screen into the client-ready PDF, older versions included. Grayed out while you have unsaved changes, with a tooltip telling you to save or discard first. [Sharing and delivering](/guides/financial-planning/the-plan/sharing-and-delivering) | Both |
| **Edit** | Opens the conversation with Hazel about this plan, under the tooltip "Ask Hazel to change this plan". There is no editing mode to switch on. [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat) | Ask Hazel |
| **Plan settings** | This plan's own assumptions, with the few you can override for one household. Covered below | Ask Hazel |
| **Planning settings** | The gear beside it leaves the plan for your firm's planning settings, opened on Solver assumptions: the defaults behind every plan, not this one. It appears only if you own the firm. [Settings overview](/guides/financial-planning/settings/overview) | Ask Hazel |
| **Feedback** | A thumbs-up and a thumbs-down on the plan itself. Covered below | Ask Hazel |
| **Regenerate** | Rebuilds the plan from the household's latest information | Financial Planning tab |
| **Open chat** | Takes you to the same plan in Ask Hazel, with the rest of the controls | Financial Planning tab |
The Financial Planning tab also carries the control that mutes the chime when a build finishes.
## Inside the document
**Blocks.** Every part of a section is a block: a narrative, key numbers, a chart, a table, a lever panel. Your firm's layout decides which appear and in what order. See [How Hazel visualizes data](/guides/financial-planning/the-plan/how-hazel-visualizes-data).
**Figures.** A dotted underline means a source is recorded behind the figure: select it for where the number came from, how it was calculated, and the assumptions it rests on. Figures without it do not open. A figure inside a sentence is locked, with the tooltip "Kept in sync with the plan": change the number by changing the assumption behind it. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
**Editing text in place.** Hover the plan's prose for a pencil at its top right: click it, type, then click away to stage the edit. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
**Levers.** Dials you can move in front of a client. Figures tied to a lever move as you drag it, but nothing is stored and no version is created.
**Dragging goal cards.** Goal order changes nothing the plan calculates, so a drag stages as an edit for **Save** to commit. Where order does feed the numbers the grip is grayed out, so ask in chat, and a dial that needs Hazel to re-solve the plan renders no control at all. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios).
**Recommendations.** **Approve** and **Ignore** record your decision on the card and in the Decisions log straight away, and stage it to save with the rest of your changes.
**Questions Hazel asks you.** A new plan may carry a card with up to three assumptions Hazel had to make, each with a line saying what the answer changes. **Looks right** accepts the figure as it stands; change it and the button becomes **Use this**, which queues a recalculation. Where Hazel had nothing to go on the box reads **Unknown**, and **Looks right** stays grayed out until you type an answer.
**The Appendix.** The plan's full detail on one page: an **Applied settings** table naming every planning setting this version was built with, its value and the layer that set it, then one card per figure with its sources, arithmetic and assumptions.
**The disclaimer.** Your firm's client-facing footer, rendered at the bottom of the plan as written and carried into the PDF. Until someone at your firm writes one, nothing renders there.
## The plan settings panel
The sliders icon in Ask Hazel opens it, and the household **Financial Planning** tab has no such control. The panel shares the right edge with a figure's detail panel, so opening one closes the other.
Every row shows a value and a chip for the layer that set it: **Firm**, the persona by name, or **This plan**. Reset a saved override and the chip reads **Inherited** until the next recalculation settles whether the persona or the firm value takes over. A market row reads **Firm** at once, since a persona never carries market rows.
Under **Solver assumptions** you can change the success threshold, the reserve length and the Roth conversion strategy; under **Market assumptions**, the scenario this plan's projections draw from plus any rows your firm has adjusted. **Apply and replan** commits your overrides, **Discard changes** in the same footer drops them, and the commit arrives as a visible turn in the plan's chat. See [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) for the fields and their ranges.
## The staging bar
Stage anything and a bar appears at the bottom of the plan, in both views, listing what is queued and offering **Save**, **Recalculate** and **Discard**. Nothing is saved until you press **Save**, and staged work does not survive a reload. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
## Feedback
In Ask Hazel, thumbs-down opens a short form: pick what went wrong, add the detail. Each version takes one vote and it cannot be changed, so cast it on the version you reviewed. Feedback goes to Hazel's team, not into how your firm's plans get built; for that, edit your base template or a persona's instructions. See [Instructions](/guides/financial-planning/settings/personas/instructions).
## Where to go next
* [How Hazel visualizes data](/guides/financial-planning/the-plan/how-hazel-visualizes-data)
* [Visual reference](/guides/financial-planning/the-plan/visual-reference)
* [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios)
* [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat)
# Changing a plan with chat
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/changing-a-plan-with-chat
Where to ask, how specific to be, what Hazel pushes back on, and what comes back.
You have read the plan and something needs to change before a client sees it. Ask in plain English and the plan comes back changed, as a new version with the earlier one still readable.
## Where to ask
* **From the plan in chat.** Press **Edit** in the plan's top bar and the conversation opens over the plan.
* **From the household.** On the household's **Financial Planning** tab, press **Open chat** in the plan's top bar for the same conversation.
**Edit** is absent while you read an older version, because past versions cannot be changed. Look for the top bar reading "Viewing v2 of 4", marked read-only, and press **Back to latest**.
Save your own edits before you ask. If Hazel publishes a new version while you have hand edits staged in the bar at the bottom of the plan, your only way out is **Discard my changes**. Press **Save** first.
## The five kinds of request
```text theme={null}
Show the allocation as a mix comparison, current versus target.
```
Naming the argument you want to make works better than naming the chart. See [How Hazel visualizes data](/guides/financial-planning/the-plan/how-hazel-visualizes-data).
```text theme={null}
Add a retirement-age lever from 60 to 70 in one-year steps, starting at 62.
```
Name the range, the step and the starting value. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios).
```text theme={null}
Add a healthcare tab covering the gap between 62 and Medicare.
```
```text theme={null}
Drop the education section, both kids are finished.
```
```text theme={null}
Move the income sequencing ahead of the balance sheet.
```
```text theme={null}
Rewrite the summary for a client who only wants the headline and the two decisions.
```
```text theme={null}
Compare the pension against the lump sum, side by side, with a verdict.
```
## How specific to be
Specific enough that there is one obvious reading. "Fix the summary" leaves it open. "Cut the summary to three paragraphs and lead with whether they can retire at 62" comes back the way you pictured it.
## What Hazel pushes back on
Some requests come back with a reason instead of a change. Read it before rephrasing: it usually names the alternative worth asking for.
* **A visual that cannot carry the claim.** Ask for a donut chart of income over twenty years and Hazel explains that a donut shows shares at one moment, then offers stacked columns.
* **A number Hazel cannot evidence.** With no business valuation on file, a request for that figure comes back as a request for the document. Hazel shows it as not provided rather than picking one.
* **Something a firm guardrail forbids.** Hazel names the never-do rule your request ran into. See [Guardrails](/guides/financial-planning/assumptions/guardrails).
* **Structure your firm has locked.** With the plan design set to Exact, a request to drop a section is declined and pointed back at the design. See [Strictness](/guides/financial-planning/settings/personas/strictness).
## What comes back
An accepted request republishes the plan, so the change arrives as a new version. Your own unsaved pencil edits count separately and do not move the version number until you save.
Rewording, reordering and most presentation changes come back fast; one that moves the numbers runs the projections again and takes longer. You do not have to watch: a banner on the household page says Hazel is recalculating and that the plan updates when it finishes.
Hazel works on one plan at a time, so **Regenerate** on the household's Financial Planning tab is unavailable while a change is running.
Several small requests in one message are fine, and Hazel usually composes them into a single version.
## Which tool for which change
| You want | Where to do it |
| ------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Different wording on the summary's opening paragraph, a highlight card, a goal card or a recommendation | The pencil that appears when you hover that text, then **Save**. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly) |
| Anything else about the writing, structure, visuals or analysis | Ask Hazel in chat |
| A what-if while you sit with a client | Move a [lever](/guides/financial-planning/the-plan/levers-and-scenarios), where the plan offers one |
| A different assumption for this household only | **Plan settings** in the top bar of the plan in chat, then **Apply and replan** |
| A rebuild after new documents or a meeting | **Regenerate**, on the household's Financial Planning tab. See [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating) |
| The same change on every plan your firm builds | Settings, then Planning: the [base template](/guides/financial-planning/settings/base-template), a persona's [instructions](/guides/financial-planning/settings/personas/instructions), or the [plan design](/guides/financial-planning/plan-design/overview) |
**Plan settings** is the sliders icon in the top bar of a plan you opened from its chat, and there is no such control on the household's **Financial Planning** tab. See [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).
You can also change an assumption from the figure it affects: select it, press **Edit**, then **Queue change**, and it waits for **Recalculate**. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
## Where to go next
* [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly)
* [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating)
* [Prompt library: Financial Planning](/resources/prompt-library/financial-planning): requests you can copy.
* [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief): say it up front, ask less later.
# Editing a plan directly
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/editing-a-plan-directly
Rewrite the plan's prose in your own words: what carries a pencil, how to keep an edit or drop one, why a figure cannot be typed over, and what survives afterwards.
A sentence in the plan is not how you would say it to this client, and the meeting is soon. Rewrite it in place: hover the passage for a pencil at its top right. Click it, or tab to it and press it, for a plain text box.
The **Edit** button in the top bar is a different tool: it opens your conversation with Hazel about the plan. It appears only when you open the plan from its chat; on a household's **Financial Planning** tab, use **Open chat** to reach the same conversation. The pencils are on both surfaces.
## Keeping an edit, and dropping one
| What you press | What happens |
| ---------------------------------------------- | ------------------------------------------------------------------ |
| Click anywhere outside the box | Your text is kept and joins the staging bar |
| Cmd or Ctrl plus Enter | The same, without the mouse |
| Enter, on a one-line field such as a goal name | The same. In a multi-line passage it is a new line |
| Escape | Your text is thrown away and the passage goes back to what it said |
Keeping an edit is not saving it. Nothing reaches the staging bar from an empty box, which a heading or paragraph cannot be, or from a passage you opened without changing.
Some cards, a highlight card and a recommendation among them, carry one pencil for the whole card: it opens every passage at once, leaving the card keeps them all, Escape drops them all.
## What carries a pencil
Hazel offers the pencil on prose it wrote, not on the plan's structure or arithmetic.
* **The executive summary**, the narrative opening of the **Summary** section. Narratives elsewhere carry no pencil, so a paragraph in **Retirement** is a request to Hazel.
* **Goal cards** in **Goals**: the name, the description, the label beside the progress bar, and the footnote.
* **Highlight cards**, wherever they appear: the title, the caption, and the body.
* **Recommendations**: the headline and the reasoning underneath it.
Nothing else. Section and tab names, chart titles, axis labels and legends, table cells, the tiles in a key-numbers row and every figure are Hazel's to write, so ask in the plan's chat for those and for any analysis the plan is missing.
The pencils step aside while Hazel is saving or recalculating, and they are gone on an older version. Press **Back to latest** to edit again.
## Why you cannot type over a number
A figure inside a sentence is locked to the plan's own math, and hovering it says as much: kept in sync with the plan.
Rewrite the sentence around it and the number comes along. In the executive summary's editor the figure shows as a small tag: leave it in place and the number keeps tracking the plan, delete it and the number leaves the sentence.
To move a number, change what produced it.
* **The data was wrong or out of date.** Upload the current document, or fix the household's profile, then regenerate. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
* **The assumption was wrong.** Change it in the figure's own panel.
### Changing an assumption behind a figure
Select the figure and **Provenance and calculations** opens, with an **Assumptions** list near the bottom. Where an assumption is yours to change it carries an **Edit** button: press it, rewrite the sentence, then press **Queue change**. The same list sits in the **Appendix**.
Nothing recalculates yet: the button reads **Pending recalculation** until you press **Recalculate** in the bar at the bottom of the plan, then **Recalculating** while Hazel works.
An assumption with no **Edit** button is fixed for this plan, and a button that will not open tells you why on hover.
A figure resting on an assumption you rewrote rests on your judgment rather than a document, so its sources read **Modeled** instead of **Verified**.
## Getting your work saved
The bar at the bottom of the plan counts both kinds of staged work on one line, for example "3 unsaved changes · 2 changes needing recalculation", and lists what is queued beneath.
* **Save** commits your written edits together as a single new version. Hazel confirms with "Saved as V5".
* **Recalculate** sends the queued changes to Hazel as one turn. Unsaved text edits are saved first as their own version, so a failed recalculation cannot take your writing with it.
* **Discard** clears everything, both kinds at once. It reads **Discard all** while something is queued.
While either job runs, the bar becomes the status line: "Saving your changes…" or "Hazel is recalculating your plan…".
Staged work lives on the page, so a reload or navigating away loses it. A failed save leaves your edits on screen and in the count, so press **Save** again.
If Hazel publishes a new version while your edits are still staged, **Save** is withdrawn and the bar offers only **Discard my changes**, naming the version that landed: "Hazel published V4 while you were editing". Your text is still on screen, so copy anything you want to keep, then discard, which puts you on the new version, ready to redo the edit.
This nearly always follows asking Hazel for something, or pressing **Regenerate** on the household's **Financial Planning** tab, with edits unsaved. Press **Save** first.
## What happens to your wording later
Your edit becomes part of the version it was saved into, and nothing later erases it.
* **A recalculation** is the strictest with your prose. Hazel is asked to keep every sentence exactly as you wrote it, and to update a number inside your wording rather than the wording.
* **A regenerate** keeps still-valid analysis and the content you wrote. Wording the new information has made untrue can change. See [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating).
* **Asking Hazel for a change** revises what you asked about and leaves the rest alone.
A wording preference that matters on every plan belongs in your base template or the persona's instructions.
## Where to go next
* [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat)
* [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating)
* [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from)
* [Instructions](/guides/financial-planning/settings/personas/instructions)
# How Hazel visualizes data
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/how-hazel-visualizes-data
Hazel picks a visual for the argument the plan needs to make. What each family of form is for, and how to ask for a different one.
You will present these visuals out loud, so each one has to carry the point you are making at that moment. Hazel picks the form that fits the claim: a **Donut chart** says "this total splits like this," a **Range of outcomes** says "we do not know the answer, we know the spread."
So the visual you get is sometimes not the one you would have named. Ask for a different one.
## The families of form
These are the product's own names: search the designer palette for them, or name them in a request.
**Narrative** for a few sentences in plain language, written for this household; Summary, Today and Goals each open with one. **Mission statement** for the household's own words about what the money is for, shown as a quote when the household has said it.
**Key numbers** for a row of headline figures at the top of a section. **Highlight cards** for short callouts: what is working, what needs attention. **Info cards** for a card per item: one per goal, account or action item.
**Success gauge** for probability of success today next to probability with the plan. **Probability bar** for one probability shown against a target. **Range of outcomes** for the spread of results across Monte Carlo simulations.
A modeled probability above your firm's display ceiling shows as greater than that ceiling, with a short note under the figure. The ceiling is ninety-nine percent unless your firm changes it.
**Scenario paths** for wealth over time on the current path next to wealth with the plan. **Stacked columns** for yearly columns split into their sources. **Inflows & outflows** for money in and money out, side by side over the years. **Whole-life cash flow** for every year's money in, out and remaining, as a chart plus a table. **Time spans** for when things start and end. **Timeline** for steps in order.
**Donut chart** for how a total splits across its pieces. **Breakdown chart** for one total broken into its parts. **Mix comparison** for labeled strips showing how several totals each split, the form for current mix next to target mix.
**Before & after** for key figures with and without the recommendations. **Scenario comparison** for two or three choices side by side over the same rows. **Progress to target** for stacked pieces measured against a target, where the gap is the point. **Change table** for rows moving from one value to another, with a total. **Bar chart** for amounts compared across years or categories, including a plain largest-first ranking.
**Ranked list** for the Social Security claiming decision: what each start age pays a month, plus the total collected by a planning age the household picks. For any other ranking, ask for a **Bar chart**. **Priority list** for Hazel's ranked priorities for this household, with blocks able to sit inside a priority. **Savings order** for where each next dollar of savings should go, in order.
**Recommendation** for one clear recommendation and the reasoning behind it. **Decision log** for decisions made, approved and still open. **Questions** for what the household needs to confirm or correct. **Asset location moves** for which holdings move to which accounts, and why.
**Levers to try** for dials such as savings rate or retirement age, with every figure bound to one moving as you drag it. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios). **Tab group** for turning a set of blocks into tabs. **Table** for rows and columns: accounts, amounts, schedules.
For what each form needs to work, see the [Visual reference](/guides/financial-planning/the-plan/visual-reference).
## Asking for a different visual
Ask in plain English, on any part of the plan:
```text theme={null}
Replace the retirement bar chart with the range of outcomes. I want
them to see the spread, not a single line.
```
```text theme={null}
Turn the Roth conversion schedule into a table with a column for the
bracket it fills.
```
* **Naming the argument works better than naming the chart.** "I want them to see the spread" gets you the right form even when you name the wrong one.
* **A form Hazel cannot fill will not appear.** A **Scenario comparison** needs options whose numbers genuinely differ. When the plan's data cannot support the form you named, Hazel says so and offers the one that fits. A **Range of outcomes** reads as a spread only while the **Monte Carlo** switch is on in [Market assumptions](/guides/financial-planning/assumptions/market-assumptions), where it starts on.
For a visual in every plan your firm builds, put it in your layout in [Plan design](/guides/financial-planning/plan-design/overview) rather than asking plan by plan, and set how closely Hazel follows that layout with the [strictness](/guides/financial-planning/settings/personas/strictness) control.
## Where to go next
* [Visual reference](/guides/financial-planning/the-plan/visual-reference)
* [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat)
* [Block settings](/guides/financial-planning/plan-design/block-settings): pinning a form into your firm's layout.
# Levers and scenarios
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/levers-and-scenarios
Move a control and the figures it feeds recalculate in front of you. Where levers live, what moving one does, how to put it back, how to make a moved value stick, and when to ask for a comparison.
You are sitting with a household and someone asks what waiting two more years would do. Move a lever and the figures it feeds recalculate while you watch.
## What a lever is
A lever is a labeled control wired into the plan's math, usually a slider with a range, a step and a starting value.
| What you want to move | How it usually shows up |
| ------------------------------- | ----------------------------------------------------------------------------------------------------- |
| Retirement spending | A slider, for example 110,000 to 190,000 a year in steps of 2,000 |
| What they save each month | A slider from today's figure up to a stretch figure |
| When they claim Social Security | A slider from 62 to 70 in one-year steps, or a switch between claiming early and delaying |
| The return the plan assumes | A slider that shifts returns one point either way from your firm's assumption, in quarter-point steps |
| Pension against lump sum | A row of option buttons, one per option |
The controls sit in a card titled for the question it answers, opening on the values the plan was published with. Where Hazel has not named it, the card reads **Optimized plan**.
That card usually sits on a priority card in the **Your priorities** tab of **Getting there**, beside the recommendation it belongs to. Some plans put it above the tabs on **Getting there** itself, and some carry bare sliders inside a tab. If a plan looks like it has no controls, scan **Getting there** and its tabs first.
A control that feeds two priority cards appears on both, and moving it on one moves it on the other.
## What happens when you move one
Everything that control feeds recalculates immediately in your browser. Figures in the other tabs of **Getting there** re-solve too, and switching tabs leaves your positions where you put them.
* **The prose does not rewrite itself.** A sentence can still quote the earlier number while the gauge beside it shows the new one, because the words were set when the plan was published. Ask Hazel to update the section and it comes back as a new version.
* **Nothing is written down.** Reload the page and every control returns to the value the plan published.
### Putting a lever back
A button at the bottom of the card returns every control in the whole plan to Hazel's recommendation, not only the one you touched, so it clears your exploration on the other cards and tabs too. It is labeled for what it returns to, and reads **Reset to optimized** where Hazel has not named it.
* A live Reset on a card you have not touched means you moved a control elsewhere.
* Sliders sitting bare in a tab have no Reset button. Reload the page to put those back.
* It goes to Hazel's recommendation, not to a position that has since been approved and saved, so read the values before you present them.
* It changes only what is on your screen, so it works on an older version you are reading. A control you moved travels with you between versions unless that version changed it, so Reset is how you read one exactly as published.
### When a lever moves nothing
* A line under the controls saying they are preview only means Hazel drew the control but could not wire it into this plan's math. The wording narrows to "not every control here" when only some are inert.
* A line under one visual saying it shows the plan as published means it is holding the published scenario while the figures around it re-solve. Only **Range of outcomes** and **Mix comparison** do this, and the line disappears when your controls go back.
With no notice at all, the figures do not depend on that control. In each case, ask Hazel in chat for the analysis instead.
## Making an explored value stick
Press **Approve** on the recommendation the card sits with: Approve reads whatever the controls are showing and writes a row into the plan's **Decisions** section, reading either "Approved as suggested" or "Approved with modifications". **Save** in the bar at the bottom of the plan then commits it as a new version.
If Approve is grayed out and hovering it reads "Requires plan recalculation", the change needs Hazel to re-run the math. Controls that do not sit with a recommendation have no Approve at all. Either way, ask in chat and the change comes back as a new version.
Dragging a card into a new order is the one gesture here that is not exploration: the drag stages a change straight away, the top bar shows an unsaved change, and you cannot export until you save or discard it. Where the order feeds the math the grip is inert and hovering it reads "Reordering this plan needs a Hazel round-trip", so ask Hazel in chat.
A downloaded PDF prints the card as the settled values the plan was published with, under its own heading or **Plan settings** where Hazel has not named it.
## Adding a lever
Ask in the plan's own conversation: **Edit** in the top bar in the chat view, **Open chat** on the household's **Financial Planning** tab.
```text theme={null}
Add a retirement-age lever from 60 to 70 in one-year steps, starting
at 62, and show its effect on the success gauge.
```
A vague request works too, and you get Hazel's judgment on the range. If no card appears, say so in the same chat and ask where it went.
Some requests come back as analysis instead. A control has to be something your browser can recompute on the spot: amounts, ages, percentages, switches and either/or choices. A Roth conversion amount does not qualify: which bracket it fills and what it does to a Medicare premium two years later does not move in a straight line. Hazel settles that one in the conversation and writes the answer into the plan.
Your firm can also prescribe the sliders a lever card should offer, up to twelve per card. In Plan design, open **Design layout**, select the **Levers to try** card, open **Advanced**, and add rows under **Levers**. A prescription is a floor, so Hazel may add more alongside them. See [Block settings](/guides/financial-planning/plan-design/block-settings).
## Levers compared to scenario comparisons
| | Lever | Scenario comparison |
| ------------ | --------------------------------------------------- | ------------------------------------------------ |
| **Question** | "How does this figure move if we change one thing?" | "Which of these two or three options is better?" |
| **Best for** | Exploring sensitivity in the room | Presenting a decision the household has to make |
Ask for a comparison when you want both answers on screen at once. It marks the better option row by row with a **Better** badge, can carry one closing line underneath, and its columns re-solve with your controls.
```text theme={null}
Compare taking the pension as a monthly benefit against the lump sum,
mark the better option on each row, and add a closing line saying
which you would take.
```
## Where to go next
* [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat)
* [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating)
* [Prompt library: Financial Planning](/resources/prompt-library/financial-planning)
# Sharing and delivering
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/sharing-and-delivering
Getting a plan to the client: the PDF export, presenting live, and the two firm-wide rules behind anything client-facing.
You have a plan and a client waiting. Export a PDF and send it, or present the plan on screen. Hazel never sends a plan to a client, and there is no client link or login.
A plan opens only for the advisor who built it, so until shared financial-plan chats land, send a colleague the PDF. See [Access and availability](/guides/financial-planning/overview/access-and-availability).
## Exporting the PDF
**Export PDF** sits in the plan's top bar, in the chat plan view and on the household **Financial Planning** tab alike. It exports the version you are reading, older versions included, and the file arrives named after the plan and its version number.
What the PDF carries:
* A cover masthead: plan title, who it was built for, the date and the version number.
* Every section in order, laid out for print, every chart and table drawn, tabs and priority cards opened out.
* The controls you can move on screen, under a **Plan settings** heading, at published values.
* Recommendations with their recorded decision in place of their buttons. An undecided one prints "Pending review", so settle those first.
* Your firm's disclaimer, verbatim, as the footer.
What it does not carry:
* Interactivity. The first page says so.
* The detail behind each figure. Sources, methodology, the numbered assumptions and the Appendix stay in the on-screen panels, and a card whose controls only Hazel can re-solve prints no settings. For an assumption the client needs to see, ask Hazel to state it in the plan itself, which prints.
**Export PDF grayed out?** You have unsaved changes. Press **Save** in the bar at the bottom of the plan, or **Discard**, and the button comes back.
Moving a slider, a switch or a row of option buttons is not an unsaved change and never blocks the export. Reordering cards is the exception: that drag stages a change and holds the export until you save or discard it.
### If the export fails
| What you see | What it means |
| ------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- |
| PDF generation is busy | Hazel is generating other PDFs. Wait a few seconds and retry |
| PDF export is temporarily unavailable | Try again in a few minutes |
| Couldn't export PDF | The catch-all. Something would not draw, or the request failed. Retry; if it repeats, ask Hazel in chat to rebuild the plan |
| The PDF download was invalid | Nothing was saved to your machine. Try again |
| This plan isn't ready to export | The saved plan cannot be rendered at all. See below |
For that last one, open the plan in its chat: if it will not display, Hazel offers a card headed "This plan needs Hazel's help" with **Ask Hazel to repair plan**. Press it, then export the fresh version. From the household **Financial Planning** tab, press **Open chat** in the top bar first.
## Presenting live
The info button in the plan's top bar, immediately left of **Export PDF**, opens a short brief on what the plan concludes and why, stamped with the version it describes. No info button means this version has no brief: read the Summary section, or ask Hazel in chat for a two-line brief.
The narrative and the Success gauge. Let them read it, then open the section that answers the question they arrived with.
Move the control and let them watch. Every figure wired to it recomputes as you drag; anything it does not feed stays where the plan published it, and only **Range of outcomes** and **Mix comparison** say when they are holding the published scenario. Name the figures that are answering the question.
A moved control springs back to the published value when the page reloads. To keep it, press **Approve** on the recommendation those controls sit with, then **Save** in the bar at the bottom. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios).
Controls with no recommendation have no **Approve**, so ask Hazel for those changes in the plan's chat before you leave; from the household **Financial Planning** tab, **Open chat** takes you there.
Then export the version you agreed on, not the one you walked in with.
## Two firm-wide rules
**The display ceiling.** A projection where every simulated path succeeds computes as 100%, which a client reads as a promise, so plans cap the probability of success they print. Above the ceiling the figure shows as `>99%`, with a line beneath saying the display is capped. The ceiling starts at 99%, your firm can set its own, and 100% switches the cap off. It is one figure for the whole firm, applies wherever that figure appears including your own screen, and changes only the display: the plan keeps the modeled figure underneath.
**The disclaimer.** Whatever your firm saves as its disclaimer renders under every plan exactly as written, on screen and in the PDF. It lives in **Settings → Planning**, on the **Client disclaimer** card, marked firm-wide. Planning settings belong to the firm owner, so if you cannot see Planning, ask whoever owns your firm's account to set it.
Until someone saves that text, no disclaimer renders anywhere. The wording already in the box is Hazel's suggestion and is not stored until you edit it, so before your firm's first export, open the card, type your own wording, and click outside the box to save. A **Restore Hazel default** button then appears. This is client-facing text your firm owns, so have compliance approve it.
## Where to go next
* [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating)
* [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from): the audit worth running before you export.
* [Best practices](/guides/financial-planning/best-practices): the review-meeting workflow end to end.
# Versions and regenerating
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/versions-and-regenerating
Nothing is overwritten. How to step through a plan's history, which buttons create a version, and when to regenerate.
You will change a plan several times before a client sees it. A plan belongs to the chat that built it, that chat keeps every version, and each accepted change publishes the next one on top.
An older version stays readable, presentable and exportable, but it cannot become current again: there is no Restore button. To put something back, tell Hazel in the plan's chat what you want restored, and it returns as a new version on top.
## Moving through the history
With two or more versions, arrows appear either side of the number and the label reads V3 (2 of 4): the version, then your position in the stack. One version shows a plain V1 badge. The number counts stored versions only, so it stays put while edits are in progress; the **N unsaved** badge beside it is your staged work.
While you read an older version the top bar turns orange, names the version, marks it read-only, and offers **Back to latest**, which returns you to the current plan as you left it. There is no split-screen compare, so ask Hazel in the plan's chat what changed between versions.
### What still works in an older version
Tabs, priority selection and the levers stay live, so you can present an older version freely. The grips that reorder goal cards go inert. **Export PDF** downloads the version you are reading as published, so lever positions you moved on screen are not in it.
Gone: the staging bar, the editing pencils, and, in the chat view, the **Edit** button. **Approve** and **Ignore** stay but gray out. Press **Back to latest** first to make a change.
## What creates a version
| What you do | Creates a version |
| ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------- |
| **Save** in the staging bar | Yes. Everything staged becomes one version, confirmed "Saved as V5" |
| **Recalculate** in the staging bar | Yes, once Hazel publishes. The whole queued batch becomes one version |
| Asking Hazel for a change in the plan's chat, **Regenerate** on the household's Financial Planning tab, or **Apply and replan** in the chat view's Plan settings panel | Yes, once Hazel publishes |
| **Queue change** on an assumption, or **Use this** on a question card whose value you changed | No. It waits for **Recalculate** |
| **Approve** or **Ignore** on a recommendation | No. Your decision lands straight away, commits with your next **Save**, and never joins the queue |
| **Looks right** on a question card, or dragging goal cards into a different order | No on its own. It stages like an edit and commits with your next **Save** |
| Moving a lever that is a number, a toggle or a choice | No. That kind of lever is a way of looking at the plan and stores nothing |
| Asking Hazel a question that changes nothing | No |
| **Discard** | No. It clears your staged text edits and queued changes |
Press **Recalculate** with text edits still unsaved and you get two versions: Hazel saves your edits first, then publishes the recalculated plan.
If approving a recommendation would move the plan's numbers, **Approve** marks the card decided but stages nothing, and the decision is gone when you reload. Ask for that change in the plan's chat instead.
If Hazel publishes a version while you have changes staged, the staging bar offers only **Discard my changes**. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
## Regenerate compared with asking for a change
**Regenerate** sits in the top bar of the household's **Financial Planning** tab, beside **Open chat**, not in the plan's chat view. On the sample household it reads **Draft a financial plan**, which starts one for a client of your own.
The button grays out while Hazel is already working on this plan, reading "Plan update in progress", and "Regenerating…" while it runs the full drafting workflow, so give it several minutes. If nothing is published, Hazel says the plan was not regenerated and you stay on the version you had.
Hazel revises the plan you have, leaving alone everything you did not mention. Faster, surgical, and right for almost everything.
Hazel rereads the household's profile and documents, folds what is new into the plan, and publishes it as the next version. Slower, and it revisits everything the new information touches.
### When to regenerate
* **New data landed.** A document the plan said it was missing reached the household profile.
* **The household's situation changed materially.** A job change, an inheritance, a death, a divorce.
* **Your firm changed its Plan templates** and you want this household on the new shape.
### When not to regenerate
* A number is wrong. Fix the data or the assumption behind it.
* A chart is wrong, or the plan reads long. Ask for the chart you want, or for a tighter draft.
* **The wrong persona was matched.** Regenerate carries no instructions from you, so open the plan's chat, name the persona you want, and ask Hazel to rebuild against it.
A regenerate uses your firm's settings as they stand today. If the base template, a persona or an assumption has changed since the plan was built, the rebuilt plan picks that up, and its shape or numbers can move for reasons unrelated to this household. Check [Planning settings](/guides/financial-planning/settings/overview) before you regenerate a plan going out today.
## What a regenerate keeps
* **The version history.** A regenerate lands as a new version on top.
* **Your wording.** Hazel is asked to preserve still-valid analysis and what you wrote. Wording the new information has made untrue can still change.
* **Any assumptions you overrode for this plan.** To drop one, open **Plan settings** from the plan's chat view and use the small reset control on that field, which restores your firm's or the persona's value.
A recalculation is stricter with your prose: Hazel is asked to keep your sentences as written and to say when something no longer matches rather than rewrite it. A figure inside one of your sentences follows the new numbers either way.
## Staying in one plan
A household can hold as many plans as you want: a second planning chat gives you a second plan, not a new version of the first. The household's **Financial Planning** tab shows whichever plan published most recently.
On the **Financial Planning** page, a row opens that household's tab, which shows the newest plan whichever row you clicked. To reach an older plan, open its own chat in Ask Hazel.
Plans are owner-scoped for now, so a colleague's plans do not appear on your Financial Planning tab and a household that looks planless to you may not be. Ask them to open theirs. See [Access and availability](/guides/financial-planning/overview/access-and-availability).
For two different futures, ask for the comparison inside the plan you have, not a second plan.
## Where to go next
* [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly)
* [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat)
* [Sharing and delivering](/guides/financial-planning/the-plan/sharing-and-delivering)
* [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios)
# Visual reference
Source: https://docs.hazel.ai/guides/financial-planning/the-plan/visual-reference
Every block Hazel can put in a plan, what each one says, and what it needs to work.
You know the point you want a plan to make, so you need the block that makes it and what it needs to work. Every one below can go in your firm's layout in [Plan design](/guides/financial-planning/plan-design/overview), or be asked for by name in chat.
In the layout editor they sit in the collapsed **Building blocks** shelf at the bottom of the **Modules** tab, ready to drag into a section, grouped there by kind: Narrative, Numbers & highlights, Charts, Tables, Interactive, Groups.
**Priority list** belongs in the **Your priorities** tab of Getting there, and **Probability bar** cannot go in the **Goals** section. Everything else goes wherever the plan needs it.
Once in a while Hazel builds a one-off visual that is not in this list. It carries a **Generative** badge on the plan, and you cannot place it in a layout.
## Words
| Block | What it says | Needs |
| --------------------- | ------------------------------------------------------------- | ----------------------------------------- |
| **Narrative** | A few sentences in plain language, written for this household | Nothing beyond the plan's content |
| **Mission statement** | The household's own words about what the money is for | Something the household has said about it |
## Headline numbers
| Block | What it says | Needs |
| ------------------- | --------------------------------------------------------------- | -------------------------------- |
| **Key numbers** | A row of headline figures | The few figures that matter most |
| **Highlight cards** | What's working, what needs attention | Short points to group |
| **Info cards** | A card per item: goals, accounts, opportunities, or a checklist | Comparable items |
## Probability
| Block | What it says | Needs |
| --------------------- | -------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Success gauge** | Probability of success today, next to probability with the plan | Two modeled results |
| **Probability bar** | Today's probability against where the plan lands, with an optional target marker | Both probabilities |
| **Range of outcomes** | The spread of results across simulated market paths | Results at two or more points in time. The band reads as a spread only while **Monte Carlo** is on in [Market assumptions](/guides/financial-planning/assumptions/market-assumptions), where it starts on |
## Time
| Block | What it says | Needs |
| ------------------------ | ------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------- |
| **Scenario paths** | Wealth over time: the current path against the planned one | One to four paths, and a title naming what the lines cover (portfolio balance, not net worth) |
| **Stacked columns** | Yearly columns split into their sources | A value per source per year, on one shared axis |
| **Inflows & outflows** | Money in and money out, side by side over the years | One flow above the zero line, one below, a figure per year |
| **Whole-life cash flow** | Every year's money in, out, and remaining, as a chart plus a table | A solved year-by-year ledger, or up to three to switch between |
| **Time spans** | Commitments side by side over the years, the area showing what each costs | Two or three items, each with a start, an end, and a recurring amount |
| **Timeline** | Steps in order across the years | Events, each with the year or window it lands in |
## Composition
| Block | What it says | Needs |
| ------------------- | ------------------------------------------ | --------------------------------------------------------- |
| **Donut chart** | How a total splits across its pieces | One or more slices |
| **Breakdown chart** | One total, broken into its parts | A stated total, parts that add up to it, nothing negative |
| **Mix comparison** | Labeled strips showing how each mix splits | One mix, or several to compare (current against target) |
## Comparison
| Block | What it says | Needs |
| ----------------------- | ---------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Before & after** | Up to five key figures, before and after the recommendations | Dials with a current and a recommended setting, both read from the plan's own model, never typed in. To change one, ask Hazel in the plan's chat |
| **Bar chart** | Bars comparing amounts across years or categories | A value per category |
| **Scenario comparison** | Two or three choices side by side, with a verdict | Two or three options, and rows whose figures genuinely differ |
| **Progress to target** | Stacked pieces against a target, with the space above them named | A target, the pieces against it, and a meaning for the space above: a gap, room to spare, or money accounted for |
| **Change table** | Rows shifting from one value to another, with a total | Each row's new value, plus the old one where there is one |
## Ranking and priority
| Block | What it says | Needs |
| ----------------- | ----------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------- |
| **Ranked list** | The Social Security claiming decision: the monthly payment at each start age, plus the lifetime total | Two to six start ages, today's option as the reference, and a planning-age dial. For other rankings, ask for a **Bar chart** |
| **Priority list** | Hazel's ranked priorities for this household, one panel per priority | Leave it empty and Hazel picks the priorities; pin blocks inside to require specific ones |
| **Savings order** | Where each next dollar of savings goes, in order | The rungs in order, each with an annual and a monthly figure |
## Decisions and actions
| Block | What it says | Needs |
| ------------------------ | --------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- |
| **Recommendation** | One recommendation, its reasoning, and Approve or Ignore. Approving records it in the Decision log | A decision worth putting in front of the household |
| **Decision log** | Decisions made, approved, and still open | Nothing. It starts empty and fills as decisions are approved, so place it early |
| **Questions** | Up to three assumptions for the household to confirm or correct, pre-filled with what Hazel assumed | At least one assumption Hazel wants confirmed. The plan already stands on those values, so answering is optional |
| **Asset location moves** | Which holdings move to which accounts, and why | An account to move from, one to move to, and a reason |
## Structure and interaction
| Block | What it says | Needs |
| ----------------- | -------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Tab group** | Each block inside becomes a tab, and Hazel finalizes the grouping when it builds | The blocks you want behind tabs |
| **Levers to try** | Dials the household can move, with every bound figure moving as the dial moves | One or more dials the plan defines. Only dials your browser can recompute appear as controls; ask in chat for anything needing Hazel to re-solve the plan |
| **Table** | Rows and columns: accounts, amounts, schedules | Column headings and at least one row |
## Where to go next
* [How Hazel visualizes data](/guides/financial-planning/the-plan/how-hazel-visualizes-data)
* [Block settings](/guides/financial-planning/plan-design/block-settings): headings, instructions and conditions on a block.
* [Plan design](/guides/financial-planning/plan-design/overview): placing these in your firm's layout.
# Troubleshooting and FAQ
Source: https://docs.hazel.ai/guides/financial-planning/troubleshooting-and-faq
Match what you are seeing to the symptom, and get the answers advisors and firm owners ask most.
You need to know whether what you are seeing is a problem before this plan goes in front of a client.
## Troubleshooting
Nothing is broken. Step two of three, badged **STEP 2 / 3**, is where Hazel asks the handful of things that would change the plan, one at a time and numbered. Answer them and the build carries on.
Rather not decide? **Make reasonable assumption** settles the one in front of you, and a second option covers the rest. On the household's **Financial Planning** tab an untouched question takes that assumption itself after about half a minute; hovering or typing pauses the timer. In a chat it waits as long as you need.
The household's **Financial Planning** tab says this when a build stops short. Nothing you uploaded is lost: **Try again** picks up from the documents Hazel already read and the answers you gave. **Start over** clears the attempt and returns you to the readiness view, for when the data was the real problem. If a build fails twice, contact support and name the household.
The card that starts a build estimates 5-15 min, and the reading step says 5-12 minutes; a household with a lot of paperwork runs longer. Hazel keeps working whether or not you watch, and chimes when the draft is ready; the speaker button beside the progress badge mutes the chime.
On the household's **Financial Planning** tab the badge estimates the time left, counts the pages read, and past its usual time reads **Taking longer than usual**. Do not start a second build; wait for the plan, then ask for changes.
Hazel states what it cannot evidence instead of guessing, so check in order:
1. **Could Hazel read the file?** During a build, the file list on the household's **Financial Planning** tab marks each document waiting, reading or read, and flags any Hazel could not read. Ask the client for a clean copy, or attach another in the plan's chat.
2. **Can the plan see it?** A document has to be attached to the plan's own chat or held on the household; **Prepare data** on that tab does both, uploading to the household's **Documents** tab.
3. **Is the fact unconfirmed?** Hazel uses unconfirmed facts and labels them Unverified in a figure's sources. On the household's **Data** tab, choose **Review** and use **Verify**. See [Reviewing and verifying](/guides/intelligent-client-profiles/managing-client-data/reviewing-and-verifying).
4. **Waiting on an account sync?** A sync never delivers planning figures: those come from documents, meetings, emails and notes, so upload the statement.
Then use **Regenerate**, in the top bar of the household's **Financial Planning** tab. See [Documents and CRM data](/guides/financial-planning/your-data/documents-and-crm).
A figure with something recorded behind it carries a faint dotted underline. Select it to open a detail panel:
* **Sources.** Which document, meeting, email or note it came from, each row carrying the value that source gave and a confidence chip: **Verified** where a document or a correction of your own settled it, **Unverified** where nobody has confirmed what was said, **Modeled** where nobody stated the figure and the plan worked it out. No dates here, so open the document itself if you suspect an old statement.
* **Methodology.** The calculation, step by step, each input tied to its source.
* **Assumptions.** Where one is yours to change you get **Edit**, then **Queue change**; the figure updates when you press **Recalculate** in the bar at the bottom of the plan.
A figure with nothing recorded renders as plain text and will not open. Try the plan's **Appendix**, the last entry in its left navigation, which gathers all three for the whole plan, or ask in the plan's chat, which produces no new version. See [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from).
A new version landed while your edits were staged, so **Discard my changes** is the only button in the bar at the bottom of the plan. Discarding keeps no copy, so copy anything you want out of the plan first, then read the new version and redo the edit. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
A grayed-out **Export PDF** means changes are staged: save or discard them in the bar at the bottom of the plan. Moving a lever stages nothing, so where you only moved dials, look for a goal card you dragged into a new order, which does. The PDF renders the last saved version. If you got a message instead:
| Message | What to do |
| ------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| PDF generation is busy | Wait the few seconds it names, then try again |
| PDF export is temporarily unavailable | Try again in a few minutes |
| The PDF download was invalid | Nothing reached your machine, so run the export again |
| This plan isn't ready to export | Open the plan in its chat, use **Ask Hazel to repair plan**, wait for the new version, then export. From the household's **Financial Planning** tab, press **Open chat** first |
| Couldn't export PDF | Try again, and if it happens twice use the thumbs-down in the plan's own conversation |
The reply names the reason.
* **A guardrail forbids it.** Nothing overrides your firm's absolute rules, Hazel included, and only a firm owner can change one, in **Settings → Planning**. See [Guardrails](/guides/financial-planning/assumptions/guardrails).
* **A number your firm set says otherwise.** These bend: say why you want the change and ask again. For this plan, use **Plan settings** in the plan's own conversation; for every plan, a firm owner changes it in Planning settings.
* **Nothing in the household's data supports the figure.** Add the document that carries it, or let the plan name the labeled assumption it used instead.
Ask in the plan's chat, naming the section: "Why is there no estate section in this plan?" Hazel answers from the plan itself, with no new version.
Under **Guided** strictness Hazel may leave a block out when the household's facts do not support it. Under **Flexible**, where an unconfigured firm starts, Hazel composes each plan freely from your design. The durable fix belongs to a firm owner, in **Settings → Planning**:
* In the **Plan design** card, use **Design layout** and select the block. **When it shows up** decides between **Always** and **When it applies**, and a **Leave out when** condition may be broader than you meant.
* For a section that has to appear every time, set strictness to **Exact**. Hazel then holds each finished plan against your design and sends one that misses a block back to be fixed, though now and then a plan still publishes with something not matching, so read the next one or two. A household whose facts do not support a section still gets one, so it can read thin.
See [Strictness](/guides/financial-planning/settings/personas/strictness).
Open the plan's **Appendix**: the first row of **Applied settings** names the persona this version was built from, or reads None.
For this plan, open its chat and name the one you want: "Rebuild this against our Pre-Retiree persona." A rebuild writes fresh content, so hand-edited wording does not come along.
For every plan after this one, rewrite the persona's **description**, the strongest matching signal; the age band is only a hint, so a vague description keeps matching households you did not mean. See [Client personas](/guides/financial-planning/settings/personas/overview).
Planning settings apply to the next plan, or to the next change to an existing one; a published plan stays as it was built, and its **Appendix** records the settings behind it.
Bring a household onto the new settings with **Regenerate** in the top bar of the household's **Financial Planning** tab. Grayed out and reading that a plan update is in progress means Hazel is still working in that plan's chat, so give it a minute. After a persona change, check the **Appendix** for the persona this household matched.
If it is one of the three Hazel installs, nobody edited anything: HENRY, Pre-Retiree and Retiree / Decumulator each arrive carrying their own plan design, so all three read **Customized** from the first time you open Planning settings, and base-template changes have never reached them.
A persona your firm created follows the base template live until someone edits that persona's own design, then keeps its copy; the **Plan design** card says which state it is in.
**Re-sync from base** moves a persona onto the base layout and discards the design it has now, so copy anything you want to keep first. Hazel asks you to confirm. See [Inheritance and re-syncing](/guides/financial-planning/settings/personas/inheritance).
A lever recalculates the figures wired to it, in your browser, and leaves the narrative alone. Moving one saves nothing.
If the numbers move far enough that the prose is now wrong, ask Hazel to update that section and it comes back as a new version. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios).
Today a plan belongs to the advisor who built it, so a colleague with full access to the household can open its **Financial Planning** tab and still see no plan. Export the PDF and send it, or have them build their own; sharing financial-plan chats is a fast follow. Check household access too, without which they cannot open the household at all. See [Roles and permissions](/guides/settings/members/roles-and-permissions).
The thumbs live on the plan in its own conversation; from the household's **Financial Planning** tab, use **Open chat** in the plan's top bar. Press the thumbs down, labeled Not helpful plan, and a short form opens: pick the kind of issue (UI bug, Not factually correct, Incomplete response, Didn't follow request, or Other) and say what you expected. Your note travels with the version you were reading.
Feedback is one way, so contact support as well if you need an answer.
## FAQ
### Building a plan
Three ways: the household's **Financial Planning** tab, the **Financial Planning** page, or **Ask Hazel** with the chat set to Financial Planning mode. A plain Ask chat will not start a plan. See [Getting started](/guides/financial-planning/overview/getting-started).
The household's **Financial Planning** tab checks eight things: Accounts & Holdings, Calendar, Tax documents, Uploaded documents, Meetings & transcripts, CRM, Email and Business valuation. Business valuation only matters where a business does, so it can read Missing while the tab still says **Ready to build**.
If something is short, the tab offers **Prepare data**, which uploads files and drafts a request email. Build anyway and the plan will say what it could not evidence.
Yes. Neither CRM data nor custodian balances become planning figures. Statements, tax returns, policies and estate documents give a plan this household's specifics, and scans and photographs are fine. See [Documents and CRM data](/guides/financial-planning/your-data/documents-and-crm).
Yes. The **Financial Planning** page offers **See financial plan example**, a complete plan on a made-up household, with **Take the tour** to walk you through the surface. It is read only, and carries no **Edit** and no PDF export.
### Working with a plan
Yes. A plan lives with the chat that built it, so a fresh build gives you a second plan rather than a new version of the first. Usually you want the new version: ask on the plan you have, or use **Regenerate** on the household's **Financial Planning** tab. For two different futures, ask for a scenario comparison inside one plan. See [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating).
The **Financial Planning** page lists your plans, most recently worked on first, with the household, the **Status** and the persona applied. Selecting a row opens that household's **Financial Planning** tab; **Open chat** there reaches the plan's own conversation. **Draft in progress** means no version published yet.
The prose, yes. Hover a passage, click the small pencil at its top right, make your change, then click away to stage it or press Escape to drop it. Staged changes collect in the bar at the bottom of the plan, and **Save** commits them as one new version.
A number inside a sentence is a chip kept in sync with the plan. Rewrite the sentence around it if the wording is wrong; to move the number, change the assumption behind it or the data underneath.
**Edit**, in the top bar of a plan in its own conversation, opens the conversation with Hazel about the plan rather than an editing mode. The household's **Financial Planning** tab carries no **Edit** button; **Open chat** there reaches the same conversation. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
Usually. Hazel is asked to preserve still-valid analysis and the content you wrote, so a passage you rewrote normally comes through. Wording the new information has made untrue can still change. If a preference should hold on every plan, put it in your firm's [instructions](/guides/financial-planning/settings/personas/instructions).
Both are requests in the plan's chat. For a lever, name the range, the step and the starting value: "Add a retirement-age lever from 60 to 70 in one-year steps, starting at 62." For a chart, name the point you want the client to take away rather than a chart type. See [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios) and [How Hazel visualizes data](/guides/financial-planning/the-plan/how-hazel-visualizes-data).
No. Hazel never sends a plan to a client: you export the PDF and hand it over, or present the plan on screen. Your firm's disclaimer travels with both. See [Sharing and delivering](/guides/financial-planning/the-plan/sharing-and-delivering).
### Firm setup
Firm owners only, and not firm admins; for everyone else the Planning tab does not appear. Every advisor still inherits what is set there, and can see it in the plan's **Appendix**. See [Access and availability](/guides/financial-planning/overview/access-and-availability).
Structure goes in [Plan design](/guides/financial-planning/plan-design/overview), voice and emphasis in [instructions](/guides/financial-planning/settings/personas/instructions), numbers in [Solver](/guides/financial-planning/assumptions/solver-assumptions) or [Market assumptions](/guides/financial-planning/assumptions/market-assumptions), and absolute rules in [Guardrails](/guides/financial-planning/assumptions/guardrails).
A module is in your plans exactly while your layout carries it, so there is no on and off switch: add or remove one with **Design layout** from the **Plan design** card. Taking one out removes the **dedicated treatment** of that topic; broader sections still fold it in. See [Modules](/guides/financial-planning/plan-design/modules).
Your firm starts with three that Hazel installs: HENRY, Pre-Retiree, and Retiree / Decumulator. Review, rename or edit them, and delete the ones that do not describe your book. Past that, enough for the client types you plan differently for.
When no persona matches a household, Hazel builds on the base template alone and records None in **Applied settings**.
Not the firm defaults. In the plan's own conversation, the sliders button in the top bar opens **Plan settings**, which holds the fields meant to vary. The household's **Financial Planning** tab has no such button, so use **Open chat** there first.
A value you set belongs to that plan, not the household, so a second plan for the same family keeps what it inherited. Each field prints the range Hazel accepts, a value outside it comes back with a message, and a reset control restores the inherited one. **Apply and replan** commits and recalculates.
The advisory fee, the terminal heir tax rate and the number of simulations never vary, so the panel does not list them.
Almost always because the matched persona overrides it on purpose. In the plan's **Appendix**, every value under **Applied settings** is labeled **Firm**, **Persona** or **This plan**, so you can see which layer set it.
**Settings → Planning**, in the **Client disclaimer** card. It opens holding wording Hazel suggests, and nothing prints under a plan until your firm saves its own text. It saves as you type, and renders under every plan and in the PDF exactly as written. **Restore Hazel default** brings the suggested wording back.
## Where to go next
* [Best practices](/guides/financial-planning/best-practices)
* [Glossary](/guides/financial-planning/glossary)
* [Prompt library: Financial Planning](/resources/prompt-library/financial-planning)
# Documents and CRM data
Source: https://docs.hazel.ai/guides/financial-planning/your-data/documents-and-crm
What your connected systems keep current, which documents change a plan most, what Hazel can read, and when a household has enough.
You want the plan built on this family's own numbers, so check what Hazel already holds for the household before you hand it more.
## What your connected systems keep current
| Source | What it brings |
| ------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- |
| **CRM** (Salesforce, HubSpot, Wealthbox, Redtail) | Household structure, contacts, relationships, and the notes your team writes. Wealthbox syncs hourly, HubSpot daily, Salesforce and Redtail nightly |
| **Email** | The client emails that carry planning facts |
| **Calendar** | Lets Hazel join and record your meetings |
| **Altruist** | Accounts, balances, holdings and cost basis, when your firm's Altruist connection is in place |
No sync writes a balance or a holding into the profile, so accounts held anywhere other than Altruist reach a plan through a statement you upload. [Integrations](/guides/integrations/overview) shows what connects today.
Hazel builds a household's profile from four kinds of material: a document, a recorded meeting, a client email, or a note, written in Hazel or in the CRM Hazel syncs from. A household that syncs cleanly but has never been met with and holds no documents has nothing financial in its profile yet.
## What you upload
**The household's Documents tab.** Hazel reads the file into the household's profile, where every plan for that household can use it.
**The plan's own chat.** Attach a file and Hazel reads it for that plan. It does not become a profile fact, so put a copy on the Documents tab when later plans need it.
### Worth uploading, in priority order
| Document | What it gives a plan |
| ---------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Recent account statements** | Balances, positions, cost basis, registration and contribution activity. Send the most recent statement for every account not held at Altruist |
| **The most recent tax return** | Income by type, brackets, deductions, capital gains, and what is already going into retirement accounts. Conversion and claiming work rests on it |
| **Employer benefits and retirement plan statements** | Workplace balances, contribution rates, match formulas, the investment menu, and equity grants with vesting schedules and strike prices. No custodian connection sees these |
| **Insurance policies** | Coverage amounts, policy type, premiums, riders and beneficiaries |
| **Estate documents** | Wills, trusts and beneficiary designations, which drive the estate review and the titling check |
| **A business valuation or recent K-1s** | Lets a plan treat a business as an asset instead of leaving it out |
### What Hazel can read
Multi-account statements, multi-year returns and long policy documents are all fine, and so are scans and photographs, because Hazel reads the text inside the image. The limits:
* **50 MB a file.** A very large or complex scan can also run out of time, and the row asks you to split it and upload the parts.
* **PDF, Word (DOCX), Excel (XLSX) and images.** Save an older .doc or .xls as .docx or .xlsx first, and export a spreadsheet as .xlsx rather than CSV.
* **A password-protected PDF is fine.** Hazel asks you for the password as the file uploads and unlocks it in your browser.
Drag files, or whole folders, anywhere onto the Documents tab.
### After you upload
The pill on each row shows how far Hazel has got:
* **Queued**, then **Reading** while Hazel pulls the text out, with a page count.
* **Waiting for profile**, then **Adding to profile**. Hazel adds documents to a household one at a time.
* **In profile.** Select the pill to see what Hazel learned.
* **Failed.** Nothing usable came out, and the row says why.
Most documents take a few minutes. A build reads the household's facts at the moment it starts, so wait for **In profile**, or anything still reading misses that plan.
### When two sources disagree
A household can hold two versions of the same fact: a CRM note from March, a statement from last week. Hazel does not average them; the newer source wins. [Where the numbers come from](/guides/financial-planning/your-data/where-numbers-come-from) has the detail.
## How much data is enough
The household's **Financial Planning** tab answers that before you build, on a card headed **Financial Plan document readiness**. The line under each row says what is missing: "3 documents, none read yet". [Getting started](/guides/financial-planning/overview/getting-started) has the eight rows and the two cards they decide.
* **A tick needs the source to be current, not only present.** Accounts count when they were updated in the last 90 days, tax documents when the latest filing is inside 18 months. A row can read **Incomplete** on a household you know has accounts.
* **Accounts & Holdings reads the household's profile,** so an Altruist connection on its own does not tick it. A tax return counts for both Tax documents and Uploaded documents, so one upload can move two rows.
Where Hazel has no household figure, a dated standard assumption is used or the figure is marked Unknown, and the plan labels both. The readiness list, not the plan, names the document that would fill the gap.
Read the eight checks, then open **Prepare data** on the same tab. Step one is a drop zone. Step two drafts an email asking the household for the rest, with the missing items listed and your next meeting date in the subject, ready to send from your own mailbox. Then build from what you have. When the documents arrive, upload them and press **Regenerate** in the plan's top bar on the Financial Planning tab.
Once a plan on this household is ready, the tab shows it in place of the readiness card.
## Where to go next
* [How Hazel uses the profile](/guides/financial-planning/your-data/how-hazel-uses-the-profile): what the profile adds on top of documents.
* [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief): the context no document carries.
* [Versions and regenerating](/guides/financial-planning/the-plan/versions-and-regenerating): what Regenerate keeps.
* [Adding client data](/guides/intelligent-client-profiles/managing-client-data/adding-client-data): how an upload becomes a profile fact.
# Front-loading the brief
Source: https://docs.hazel.ai/guides/financial-planning/your-data/front-loading-the-brief
Put your context, goals and constraints in the message that asks for the plan, so the first draft comes back as the draft you wanted.
You are about to ask Hazel for a plan, so put your context, goals and constraints in that first message and the draft comes back closer to the one you wanted.
"Build a plan for the Hendersons" produces a real plan, but Hazel then chooses for you on everything you did not mention: the Client persona to build against, the topics to lead with, the scenario to model, the sections this family does not need. Said afterwards, the same intent has Hazel revising a plan built on different priorities, over several passes, each one a change to review.
## Where you can front-load
Only one of the three ways into a plan gives you somewhere to type.
* **Ask Hazel.** On the segmented control under the message box, choose **Financial Planning**, then send your request and your context in the same message. That message is what the build reads when it starts. A chat left on **Ask** answers you and never builds a plan.
* **A household's Financial Planning tab.** The **Draft financial plan** button sends a plain request naming the household and its members, with nothing of yours attached. Once the build is running, **Open chat** in the plan's top bar takes you to the conversation, where you say the rest.
* **New financial plan** on the Financial Planning page. This wizard asks for your data source, who the plan is for, and any uploads, then starts the same plain request as the tab.
## Hazel stops partway to ask
The second of the three build steps is questions: at most three inputs, the ones that would move the outcome most, plus a check on how this household's goals should be ranked. A brief settles some of them in advance. Each question arrives carrying the value Hazel would otherwise use, so stepping away is safe: on a household's **Financial Planning** tab a question you leave alone settles into that value, while a build you started in a chat waits for you. [Getting started](/guides/financial-planning/overview/getting-started) has the step in full.
## What to put in the brief
Paste this into a Financial Planning chat, fill in what applies and delete the rest. If you only manage two lines, make them the question the plan has to answer and the constraints.
```text theme={null}
Build a financial plan for [household].
Treat them as [Client persona, if you have a view].
The question this plan has to answer: [can they retire at 62?].
Goals: [retirement year, second home, college, giving, business exit].
Constraints: [they will not sell the rental; no annuities, ever].
Emphasize: [lead with the income sequencing].
Leave out: [the education section, the kids are done].
Data notes: [the brokerage statement is from last year; ignore the
lapsed insurance policy].
Audience: [a first meeting with a skeptical engineer].
```
Hazel cannot infer a constraint, so name it. Repeating a goal Hazel already has costs nothing, so say it even when you are not sure.
If your firm has not set up personas yet, leave the persona line out; Hazel builds from your firm's base template.
Whether Hazel can drop a section your firm's plan design asks for depends on that design's strictness: under **Exact** the design is built as written, however you word the request. [Strictness](/guides/financial-planning/settings/personas/strictness) covers the three modes.
## What a brief does not set
The planning numbers come from firm configuration instead, set once by your firm owner and inherited by every plan: the success threshold, the default reserve length and its refill policy, the withdrawal guardrails, the Roth conversion strategy, the advisory fee, the Monte Carlo simulation count, and the return, volatility, inflation and tax figures. [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions) and [Market assumptions](/guides/financial-planning/assumptions/market-assumptions) explain what each one drives.
Where your firm has set a hard guardrail, asking for something it forbids gets a friendly refusal naming the policy. [Guardrails](/guides/financial-planning/assumptions/guardrails) has the details. For a genuine one-off, change the figure directly:
Open **Plan settings** from the top bar of a plan you are reading in its chat, which is the only view that carries it, and change the figure there for that one plan. See [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).
## Naming the persona up front
Hazel matches each household to the closest Client persona without being asked, and the plan's **Appendix** records which one. Name a different one in the brief: "Build a plan for the Okonkwos using our Business Owner persona, not the pre-retiree one." Correcting it after the build means asking Hazel to rebuild. See [Client personas](/guides/financial-planning/settings/personas/overview).
## Front-loading a change to an existing plan
"Add a lever" leaves Hazel to pick one. "Add a retirement-age lever from 60 to 70 in one-year steps, starting at 62, and show what it does to the success gauge" gets you that lever. The success gauge is the dial that shows probability of success today against probability with the plan.
## Where to go next
* [Getting started](/guides/financial-planning/overview/getting-started): the three ways into a plan.
* [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat): the follow-up vocabulary.
* [Prompt library: Financial Planning](/resources/prompt-library/financial-planning): requests to copy and paste.
* [Best practices](/guides/financial-planning/best-practices): where this habit sits in the wider workflow.
# How planning uses the client profile
Source: https://docs.hazel.ai/guides/financial-planning/your-data/how-hazel-uses-the-profile
Which household facts a plan reads, how confirmed and unconfirmed facts are treated, and what to check before you build.
Before you draft a plan, you want to know what Hazel is drafting it from: the household's [Intelligent Client Profile](/guides/intelligent-client-profiles/overview/introduction), the lasting facts captured from recorded meetings, client emails, uploaded documents and your notes. Whatever the profile holds on the day you build is what the plan rests on.
## What a plan takes from the profile
| From the profile | What it does in the plan |
| --------------------------------------------------- | -------------------------------------------------------------------------------------------------- |
| Household members, dates of birth, dependents | Who the plan covers, whose retirement and claiming ages get modeled, who is named as a beneficiary |
| Income, employment, business ownership | Cash flow, savings capacity, and whether a business needs its own treatment |
| Spending, by category | What the household lives on, which every retirement figure gets measured against |
| Accounts, balances, and ownership | The balance sheet, and which accounts fund which spending |
| Property, debts, and terms | Net worth, the order debts get paid down in, and the housing picture |
| Insurance coverage in force | Whether the coverage holds up, and which gaps to name |
| Wills, directives, trusts, beneficiary designations | The estate picture, and whether the titling matches the intent |
| Stated goals, in the household's own words | The Goals section, and what every trade-off gets measured against |
| Constraints and preferences | The lines a plan should not cross, and the tone it takes |
Hazel records a date of birth, never an age. When a client only mentioned an age, Hazel works out an approximate birth year from the date of that conversation and marks it approximate. Retirement ages, claiming ages and required distributions all hang off that date, so put the real one in on the household's **Data** tab.
## Confirmed and unconfirmed facts
Some facts count as settled the moment they arrive: anything a document said, anything you typed in or corrected yourself, and everything about the people and what they want, from who is in the household to their goals, concerns, preferences and life events.
What waits to be confirmed is mostly money: a balance someone mentioned in a meeting, an income figure in an email, the name of an executor. On screen, Hazel labels the two states verified and unverified.
**The most recent source holds the figure.** An older document saying the same thing confirms the newer figure rather than displacing it, and the fact reads verified from then on. When Hazel has the wrong figure, correct that fact on the household's **Data** tab: your correction carries today's date and arrives confirmed.
**A consequential unconfirmed figure earns a question.** The build pauses to ask you to confirm the figures that would move the outcome most; small ones get used without a question. See [Getting started](/guides/financial-planning/overview/getting-started).
Before you build, open the household, select the **Data** tab and choose **Review** in the left list. The badge counts the records with something still to confirm, and each record carries a **Verify** button that settles all of its waiting items at once.
## What the profile does not carry
The profile records what was stated, never what can be worked out from it. It holds "about two million across the IRAs, as of the March meeting" because a client said so, but never a total net worth or a probability of success. [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from) shows how to follow a figure back to the fact or assumption behind it.
## Keeping the profile current
Goals, constraints and family context reach the profile from conversation. When a meeting goes unrecorded, write up what was said as a note on the household's **Notes** tab: the goals and preferences in it land settled, and any figures wait to be confirmed.
A statement or a tax return uploaded to the household's **Documents** tab becomes profile facts, available to every plan for that household. Scans and photos are fine.
In an **Ask Hazel** chat, name the household and say what happened: "The Hendersons paid off the HELOC and opened a 529 for Maya." Hazel comes back with the exact changes, you approve them, and they land as confirmed facts. Do this in an ordinary chat, not in a plan's own conversation: a plan chat reads the profile and never writes to it.
A thin profile does not stop you building. On a household's **Financial Planning** tab, the card headed **Begin right away** names what is missing and still offers **Draft financial plan**, with a line saying that parts of the plan will use labeled assumptions.
## Where to go next
* [Documents and CRM data](/guides/financial-planning/your-data/documents-and-crm): where the profile's facts come from.
* [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief): what to tell Hazel that the profile cannot know.
* [Where numbers come from](/guides/financial-planning/your-data/where-numbers-come-from): how to check a figure before you present it.
* [Reviewing and verifying profile data](/guides/intelligent-client-profiles/managing-client-data/reviewing-and-verifying): the Review list in full.
# Where every number comes from
Source: https://docs.hazel.ai/guides/financial-planning/your-data/where-numbers-come-from
Trace any figure back to its sources, calculation and assumptions, and audit a plan before you present it.
You are going to put these numbers in front of a client, so you need to know where each came from. Most figures carry their own audit trail.
## Selecting a figure
A figure with something recorded behind it carries a dotted underline. Select it and the **Provenance and calculations** panel slides in from the right. Hover first for a tooltip: "How this was calculated" where there is math behind it, "Where this figure comes from" where there is only a source.
No underline means nothing was recorded, or another control owns the click. Where one you need will not open, the **Appendix** lists every recorded figure.
Tiles in a **Key numbers** row carry a small source icon, opening a short version with confidence and source rows and a **View calculations** link to the full panel.
## What the panel contains
| Section | What it shows |
| --------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Sources** | A **Source / Confidence / Value** table, one row per input. A linkable document becomes a link; where the household said it, the row carries their words |
| **Methodology** | The calculation step by step: the formula in words, then the same formula with this household's figures and the result. Traced inputs name their source row |
| **Assumptions** | What the figure rests on, numbered A1, A2 and so on, each a sentence rather than a value |
Methodology and Assumptions appear only where Hazel recorded them, and where nothing was recorded the panel says so. **View full breakdown**, at the top, opens the **Appendix**.
## The three confidence labels
| Label | What it means |
| -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Verified** | A document established it, you entered or corrected it yourself, or the household said it about themselves and what they want, which Hazel treats as settled once said |
| **Unverified** | Hazel captured it, but nobody confirmed it. Mostly money: a balance someone mentioned, an income figure in an email. Check it before you present it |
| **Modeled** | Nobody stated it. The plan worked it out from figures it does have, or fell back on a standard assumption because the household had none; Methodology and Assumptions tell you which. Fine on the future; on today, find the document |
Unknown is a further case: nothing could source the figure at all. Confidence is not printed beside figures in the plan's prose.
Confirm a fact in the household's profile, not in the plan: settle it in the [profile review queue](/guides/intelligent-client-profiles/managing-client-data/reviewing-and-verifying) and rebuild, and the source reads verified next time.
## Changing an assumption from the panel
A figure cannot be typed over. Change what produced it instead.
An assumption you can change for this plan carries an **Edit** button: selecting it opens the sentence for rewriting, and **Queue change** adds it to the recalculation queue. Nothing recalculates yet: the bar at the bottom of the plan counts what is waiting, and **Recalculate** sends the lot to Hazel in one pass. See [Editing a plan directly](/guides/financial-planning/the-plan/editing-a-plan-directly).
An assumption with no **Edit** button is fixed for this plan. One that will not open says why on hover: "Requires plan recalculation" where the plan cannot be recalculated from where you are reading it, "Past plan versions are read-only" on an earlier version.
## Auditing the whole plan
**Appendix.** The last entry in the plan's left rail. Near the top, **Applied settings** names the Client persona Hazel matched and the market scenario the projections drew from, then tables of **Solver assumptions**, **Planning assumptions** and **Market assumptions** under **Setting / Value / Set by**. A footnote says market rows nobody changed follow the scenario's published values. Below that, a card per figure.
**Plan settings.** The sliders icon in the plan's top bar when you open a plan from its chat. It shows this plan's effective **Solver assumptions** and **Market assumptions**, each labeled with the layer that set it, and lets you change those that can vary by household. See [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).
Plan settings and the figure panel share the same dock, so opening one closes the other. The household **Financial Planning** tab has no sliders icon: select **Open chat** in its top bar to reach the panel.
## Which layer set a setting
| Label | What it means |
| ------------- | ----------------------------------------------------------------------------------------------- |
| **Firm** | Your firm's default, set in Settings, then Planning |
| **Persona** | The matched Client persona's override of the firm default. The label carries the persona's name |
| **This plan** | An override set on this plan, for this household only |
In Plan settings, a reset plan-level override reads **Inherited** until you apply.
Three inputs cannot vary by household: **Advisory fee**, **Terminal heir tax rate** and **Monte Carlo simulations**. They are not in the Plan settings panel, but Applied settings lists all three, marked **Firm**.
## Display rules clients ask about
* **Displayed probability of success is capped.** Above the ceiling, the figure reads as greater than the ceiling, and a line beneath names the ceiling and says the cap applied. The ceiling starts at 99% and is a firm-level setting, so read that line rather than assuming 99%. At 100% the cap is off and a plan prints a flat 100%.
* **A range of outcomes stays a range.** Where the plan models market paths, the **Range of outcomes** chart draws an outer band for the 10th to 90th percentile, an inner band for the 25th to 75th, and a median line. Point a client at the bands, not the median.
## Before you present a plan
The info button on the plan's top bar. Its closing **Start here** bullet names what to resolve with the client first. The panel stamps the version the brief was written for, which can be earlier than the one on screen; with no brief, there is no info button.
Open the **Appendix** and read the **Persona** row at the top of Applied settings. "None" means nothing fitted, so Hazel built on your base template alone. Where it is wrong, do not present the plan: ask Hazel to rebuild against the persona you want, by name. In chat, **Edit** in the top bar opens that conversation; on the household **Financial Planning** tab, select **Open chat** first. There is no persona picker on the plan itself.
The figure on the plan's success gauge, one balance-sheet figure and one income figure. Read the methodology and assumptions, and confirm the source rows name something you trust and read verified. A document row names the source, not the date of the figure inside, so open the document yourself where currency matters.
Every figure's card in the Appendix carries the same Confidence column. Confirm an unverified figure with the client and settle it in the profile, or drop the claim.
## Where to go next
* [Glossary](/guides/financial-planning/glossary): the vocabulary on this page.
* [Anatomy of a plan](/guides/financial-planning/the-plan/anatomy-of-a-plan).
* [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions): what each input drives.
* [How planning uses the client profile](/guides/financial-planning/your-data/how-hazel-uses-the-profile).
* [Troubleshooting and FAQ](/guides/financial-planning/troubleshooting-and-faq): when a figure looks wrong.
# Intro to Hazel
Source: https://docs.hazel.ai/guides/getting-started/introduction
Hazel is a trustworthy, intelligent AI engine built by Altruist. Hazel connects to your data sources to deliver deeper reasoning and analysis grounded in your practice’s most relevant context.
Powered by the latest generation of models:
**ChatGPT GPT-5.6** - ChatGPT's most capable and efficient frontier model for professional work. [Learn more](https://openai.com/index/gpt-5-6/).
**Claude Opus 4.8** - Claude's most intelligent model, and an industry-leader for coding, agents, and computer use. [Learn more](https://www.anthropic.com/news/claude-opus-4-5).
**Claude Sonnet 4.6** - Claude's fast, highly capable AI model for automating tasks, navigating complex software via computer use, and analyzing data. [Learn more](https://www.anthropic.com/news/claude-sonnet-4-6).
## Get started
If you’re new to Hazel, start here to learn the essentials.
### Ask Hazel
Built to answer questions, surface insights, and help you get things done. Instead of clicking through endless pages, ask and access in seconds. It’s your intelligent partner that uses your data as the context to analyze and surface what matters most.
[Learn more about Ask Hazel.](/guides/ask-hazel-chat/overview)
### Meetings
Upcoming meetings are displayed with an agenda summary accessible in one click. That way, you’re always prepared to walk in (or log in) confident, informed, and ready to engage.
[Learn more about our Meeting Assistant.](/guides/meeting-assistant/overview/how-it-works)
### Emails
Your most important client emails are highlighted so you never miss a critical message. Only emails from contacts in your CRM are pulled into Hazel, eliminating unwanted distractions.
Hazel will then pull relevant contact details to help you draft intelligent responses, ensuring every reply is timely and informed.
[Learn more about our Email Assistant.](/guides/email-assistant/overview)
## Privacy & Security
Underneath Hazel's remarkable assistance, there is a powerful network of cybersecurity features engineered to keep client data safe, private, and within your control.
[Learn more about our privacy and security.](/guides/help/privacy-and-security/privacy-and-security)
## Features
Dive into what Hazel can do across your practice.
## Support
# Onboarding
Source: https://docs.hazel.ai/guides/getting-started/onboarding
# Welcome to Hazel!
You're about two minutes away from having an AI partner that knows your clients, preps your meetings, drafts your emails, and keeps your day organized.
Here's the idea: Hazel pulls together everything scattered across your **CRM, inbox, calendar, and custodial accounts**, and turns it into one clear, conversational view of what matters right now. Ask a question in plain English—*"What should I know before my 2pm with the Garcias?"*—and Hazel answers with the full picture.
This guide walks you through setup and your first few minutes in the app. Let's get you running.
**Were you invited by your firm?** A couple of these steps (choosing a plan and adding payment) are handled by your firm's Owner, so you'll breeze right past them. Everything else is the same.
## Setting up Hazel
When you first sign in—with your work email (Google or Microsoft) or Altruist single sign-on—Hazel greets you and walks you through a short setup. You can move **Back** and **Next** at any point, and skip the optional steps.
Hazel says hello and lets you know setup takes about two minutes. Click **Get started** to begin.
On **"How can Hazel help you?"**, pick the experience that fits your firm:
* **Admin AI** — draft emails, summarize documents, answer questions, and tackle daily tasks, all informed by your CRM, custodial data, and notes.
* **Tax Planning + Admin AI** — everything in Admin AI, plus AI-powered tax analysis. Upload 1040s, paystubs, and statements to generate personalized tax strategies.
Don't overthink it—you can change your plan later.
This is where Hazel comes alive. Connect your **email & calendar**, your **CRM**, and (for Altruist firms) your **custodian**. The more you connect, the more Hazel can do. See [Connecting your tools](#connecting-your-tools) below for the details—or click **Skip** and connect later.
Choose **Admin AI** or **Admin AI + Tax Planning**, and **Monthly** or **Annually** (annual saves the most). You won't be charged until your trial ends, and you can cancel anytime. See current rates on the [pricing page](https://hazel.ai/pricing).
Add a payment method to start your **14-day free trial**. You won't be charged until the trial ends—and your card is handled securely.
Bring in your colleagues by email and give each person a role (**Advisor** or **Owner**). You can also turn on the **Tax Planning** add-on per person. Not ready? Skip it—you can invite anyone later from Settings.
Hazel takes a moment to set up your workspace, then drops you straight into the app. That's it—you're in.
## Connecting your tools
Connecting your tools is the single most valuable thing you'll do—it's what lets Hazel work from *your* clients and *your* day instead of starting from a blank page.
**Your data is safe.** Hazel does not train AI models on your data. When data is shared with AI providers to do work for you, it's covered by strict no-training and data-retention controls, and Hazel is AICPA/SOC certified. [Learn more about Hazel's security](https://hazel.ai/security).
**Google** or **Microsoft (Outlook)**
**Wealthbox, Salesforce, Redtail,** or **HubSpot**
**Altruist** (Schwab & Fidelity coming soon)
### Email & calendar
Connect Google or Microsoft and Hazel can:
* Prep you for every meeting with agendas and the relevant client context.
* Draft replies in your voice, ready for your review.
* Keep your daily priorities current.
You'll be asked to grant permission to see your calendar and to read, compose, and send email on your behalf—that's what powers meeting prep and draft replies. [More on email & calendar →](/guides/integrations/calendar-and-emails/google)
**Signed in with Google or Microsoft SSO?** The first time you log in, Hazel automatically prompts you to connect the calendar for the same account you signed in with — no need to navigate to Settings. The consent screen is pre-filled with your SSO email, so you can approve it in one click.
If you dismiss the prompt, Hazel won't ask again. You can connect your calendar later from **Settings → Integrations**, but features like meeting prep, agendas, and daily priorities will be limited until you do.
### CRM
Your CRM is your client memory—connecting it lets Hazel surface client history in seconds and sync notes and tasks back automatically.
* Click your CRM to connect it. Most use a quick, secure sign-in; **Redtail** asks for your Redtail username and password.
* Hazel imports your contacts, notes, and tasks. The first import can take a little while (you'll see a **"Syncing… may take 10-15 min"** note), then Hazel keeps up automatically.
You can connect **one CRM at a time**. Once you connect one, the others are disabled—your CRM is your single source of truth for client data. [More on CRM integrations →](/guides/integrations/overview)
### Custodian (Altruist)
If your firm uses Altruist, connect it to ask Hazel about balances, holdings, recent changes, transactions, and cost basis—answered from the latest account data. You'll review a short **Data Access Consent** screen and click **Agree & Connect**. Access is **read-only** (Hazel can never trade or move money), nothing is stored, and you can disconnect anytime from Settings.
## Your first few minutes in Hazel
When you land in Hazel, you'll see your **Account strength**—a simple progress tracker with one friendly goal: *the more you set up, the more Hazel can do for your firm.* You already get credit for creating your account; here's the rest of the checklist.
Every client's notes, tasks, and history—right where Hazel needs them. (Skipped it during setup? Connect it here.)
Walk into every meeting fully prepped—Hazel pulls the client picture for you.
Hazel can join your meetings automatically and produce notes and follow-ups. Make sure your calendar is connected and **auto-join** is on in [Meeting Settings](/guides/settings/meeting-settings), and Hazel will hop into your next call. [How the notetaker works →](/guides/meeting-assistant/overview/how-it-works)
If you added Tax Planning, upload a client's tax documents and let Hazel turn them into a clear, interactive summary. [Explore Tax Planning →](/guides/modes/tax-mode)
When you complete the checklist, Hazel celebrates and nudges you to **invite your team**—because Hazel gets better when your whole firm is in it together.
**Try Ask Hazel right away.** Open a new chat and ask something real:
* *"Give me a recap of the Thompson household."*
* *"What follow-ups do I owe clients from this week?"*
* *"Draft a reply to this email in my voice."*
[Get the most out of Ask Hazel →](/guides/ask-hazel-chat/best-practices)
## Good to know
* **Your free trial is 14 days.** It starts when the first person at your firm gets going, and you won't be charged until it ends. It rolls into a paid plan automatically unless you cancel beforehand—cancel anytime.
* **Billing is Owner-only.** Only firm Owners choose the plan and manage payment. Invited teammates jump straight to using Hazel.
* **Roles, in one line.** **Owners** run the firm and see everything; **Advisors** work with the clients shared with them; **Assistants** are support seats. [More on roles →](/guides/settings/members/roles-and-permissions)
* **You can skip and come back.** Connecting tools and inviting your team are optional during setup—do them anytime from **Settings**.
Once you're set up, Hazel becomes your daily partner—prepping you for meetings, drafting communications, surfacing client insights, and helping you prioritize your day. Welcome aboard.
## Where to next
Begin with a full demo video.
Your conversational command center.
Let Hazel capture and summarize your meetings.
Connect everything Hazel works from.
# FAQ
Source: https://docs.hazel.ai/guides/help/faq
### Ask Hazel FAQ
Simply click on the plus icon within the Ask Hazel search bar, type in your prompt, and press enter! Hazel will analyze the file and provide you with a response.
The Ideas page is how we track feedback, through Canny's feedback portal. If it is not working, try clearing your cookies/data on the Hazel page for the Canny connection. Then try again by refreshing and re-clicking the Ideas button within Hazel.
No, you cannot edit your Altruist data within Hazel. Hazel has read-only access to your custodial data in Altruist.
In chat, no. Chat only relies on individual account data, meaning you have to connect to your Altruist account connected to view any custodial data.
In meeting agendas, perhaps. Meeting prep and agendas rely on the calendar event owner's Altruist connection, so your Altruist account data could be shared if other Hazel team members in your firm have access to the same meeting via the household participant permission. More on household permissions here.
### Meetings FAQ
To edit speaker labels in a Meeting transcript:
1. Open the Transcript tab in the Meeting you would like to update.
2. Click on the name of a Speaker label. You will be presented with the option to:
* Rename Manually
* Rename as Client
* Swap Labels
3. Click Save
### Email FAQ
On the follow up emails, you can tweak the prompt and give Hazel instructions and example quotes of things you'd likely say. By instructing Hazel in the prompt, you can set rules for tones to follow.
On the the email assistant that auto drafts inbound emails, you cannot set rules or give Hazel instructions on how to formulate emails.
Yes, you can edit the screening criteria and adjust the prompt Hazel uses to filter inbound emails within Settings / Account / Email Preferences.
### Altruist Integation FAQ
No. Hazel has read-only access to your custodial data in Altruist.
In chat, no. Chat only relies on individual account data, meaning you have to connect to your Altruist account to view any custodial data.
In meeting agendas, perhaps. Meeting prep and agendas rely on the calendar event owner's altruist connection, so your Altruist account data could be shared if other Hazel team members in your firm have access to the same meeting via the household participant permission. [More on household permissions here.](/guides/settings/members/roles-and-permissions)
We can only pull contacts from your CRM at this time, not Altruist.
The Altruist integration is completely separate from contacts. Meaning if you go into chat today and ask it for custodial data, it would fetch the appropriate Altruist contact and the relevant details.
### Billing FAQ
Each firm has one free trial period. It’s advised that all users in the firm test Hazel through that free trial. Reach out to [support@hazel.ai](mailto:support@hazel.ai) if you have any questions on your free trial.
Each firm has one free 14 day trial period.
All users are billed on the 1st of the month in which their subscription is due.
Monthly users are billed on the 1st of the month, each month they are subscribed.
Annual users are billed on the 1st of the month after their annual subscription is due.
You can update your payment method in by editing the existing Hazel subscription in Settings → Subscriptions.
Only Owners can invite new members and manage roles.
Hazel charges will appear on your statement as `ALTRUIST HAZEL.AI`.
To unsubscribe manage billing, go to Settings -→ Subscriptions > Manage Billing.
Only Owners can invite new members and manage roles.
Yes. Reach out to [support@hazel.ai](mailto:support@hazel.ai) to enroll in automatic invoices.
### Intelligent Client Profiles FAQ
In chat, in meeting prep, in financial planning, and on the household Summary tab. Ask Hazel (including the chat rail on the household page) reads the profile first when a question concerns a client; meeting prep and financial planning start from it. The Summary tab is written from it, sentence by sentence, with citations. It works in both directions: you can also update the profile from chat by telling Hazel what changed and confirming. When Hazel uses a fact you haven't verified, the fact is flagged as unverified rather than presented as settled. Archived records are left out entirely.
Tell Hazel, in plain English: "Update Bob's 401(k) balance to $612,000," "they opened
a 529 for Maya at Fidelity, $25,000 to start," or "fix the owner on the term policy, it's Sarah's." Hazel shows you the exact updates planned; confirm, and they're applied as your entries: verified, and visible in each value's History. Nothing changes without your confirmation, and you need edit access to the client, just like on the Data tab.
Your CRM is where you *put* information; the profile is what Hazel *learns* from meetings, client emails, documents, and notes, including things CRMs rarely hold (goals, concerns, family context, "we're thinking about selling the rental"), each tied to the conversation or document it came from. It complements your CRM and never writes to it.
No. Hazel can use unverified facts; they're just labeled as unverified whenever they come up. Verifying is how you promote what Hazel *heard* into what your firm has *confirmed*, so a practical habit is a one-minute pass through **Review** after each meeting: verify what's right, correct what isn't. Facts from documents, and anything you enter yourself, arrive already verified.
Yes. The profile always keeps the most recent statement of a fact, by source date. If a client gives a new number next month, it replaces the one you verified, and the old value moves into History. Verification marks a value as confirmed; it doesn't freeze it. The same goes for archived records: a later meeting or document showing something is active again brings it back. What *can't* happen is an older source overwriting newer information. An out-of-date statement or an old document never displaces fresher information.
It's a hard reset: Hazel erases everything currently known about the household, then relearns it from the recorded meetings and uploaded documents, oldest to newest. Three things to understand before using it: your manual edits and verification marks are erased; the Archive is emptied; and facts that came only from **emails or notes** are not relearned. It usually takes a few minutes (longer for a long history) and you can keep working while it runs. Use it when a profile has gone wrong beyond spot-fixing; for one wrong value, **Edit** is the right tool.
PDF, Word (DOCX), Excel (XLSX), and images, up to 10 MB per file, on the client's **Documents** tab. Legacy .doc and .xls need re-saving in the modern format first. Password-protected PDFs prompt for the password, which is used on your device and never leaves your browser.
The facts Hazel extracted from it are removed from the profile too; the delete confirmation warns you. If a document's fact had replaced an earlier value, the earlier value comes back. The upload itself is gone; re-upload if you deleted by mistake.
Yes, through chat. Tell Hazel about the new account, goal, or policy ("add a term life policy for Sarah, $2M through Banner, $180 a month"), review what Hazel proposes, and confirm; it takes its place in the right section as your verified entry. The Data tab itself has no create-a-record form (it's for correcting and managing what's already captured), so for new facts, use chat, write a note, upload a document, or bring it up in a recorded meeting.
If something ended (account closed, goal reached), **Archive** the record. It stops populating the profile, stays readable with its history, and can be unarchived. If a value is wrong, **Edit** it. If a whole record was never true (misheard, or entered in error), **remove** it: Hazel asks you to confirm, and removal is permanent. When in doubt, archive first; you can always remove later, but not the reverse.
Yes, two ways. Uploading documents brings in the past instantly: statements and tax returns carry their own dates, so history files correctly. And on a client whose profile is still new, **Rebuild profile** relearns from the household's full recorded meeting history plus every uploaded document, oldest to newest. On an empty profile there's nothing to lose, which makes it a quick way to bootstrap a long-standing client. (On an established profile, rebuild is a last resort because it erases manual edits; see the Rebuild question above.)
Anyone with access to that client, following the same rules as the rest of the client's record. There's one shared profile per client, so teammates see each other's verifications and corrections. Owners can see every client. Clients themselves never see it.
Hazel records it internally, but it isn't shown on the page today. If you need to know for a specific record, email [support@hazel.ai](mailto:support@hazel.ai) and we can check.
No. The profile reads emails your clients send you; your outgoing messages aren't captured. If you shared something important in an outgoing email, tell Hazel in chat, upload the underlying document, or mention it in your next recorded meeting.
Probably not. A joint fact is one record with two owners, shown under each owner's view. Expand it and check the Owner row: if it lists both spouses, it's one shared record. A true duplicate can be cleaned up by archiving the extra record.
# Pricing
Source: https://docs.hazel.ai/guides/help/hazel-pricing-page
# Go to Trust Center
Source: https://docs.hazel.ai/guides/help/privacy-and-security/go-to-trust-center
# Go to Security Page
Source: https://docs.hazel.ai/guides/help/privacy-and-security/hazel-security-page
# Privacy & Security
Source: https://docs.hazel.ai/guides/help/privacy-and-security/privacy-and-security
Hazel AI’s backend is built on distributed microservices running in Amazon EKS (Kubernetes), orchestrating workflow, AI reasoning, and data services. When user data is processed:
* Only authorized, permissioned data is included, leveraging role-based access controls and pre-processing filters that validate and sanitize all inputs.
* Retrieval-augmented generation (RAG) ensures only permissible information enters analytic processes, with post-processing checks ensuring security protocols are respected so AI responses adhere to user roles and organizational security policies.
Data Storage Architecture
* **Databases**: Data is stored in PostgreSQL databases with forced SSL connections, restricted to private subnets and accessed via secure VPCs. All connections require TLS and access is managed through granular security groups and a proxy layer that enforces encryption and limits exposure.
* **AI Knowledge Base**: Hazel uses AWS Bedrock with OpenSearch Serverless for vector-based search and knowledge management. Data in OpenSearch and S3 is encrypted at rest using AWS KMS keys and protected with IAM-integrated permissions.
* **Cache Layer**: Redis cache is encrypted both at rest and in transit.
* **Backup & Edge Services**: Backups reside in encrypted S3 buckets, and edge services (via Fastly) enforce non-caching for APIs to guarantee data freshness.
Security Practices
* **Encryption**: Hazel enforces encryption at rest for all services, uses TLS 1.2+ for transit, and manages secrets via Vault and AWS Secrets Manager.
* **Access Controls:** VPC-only admin access, RBAC for Kubernetes, granular security groups, VPN access for administration, and strict workload identity management.
* **Monitoring & Logging:** Constant monitoring with Datadog and AWS CloudWatch ensures visibility for threat detection, operational health, and compliance.
* **Network Security:** Protection with network segmentation, web application firewalls (WAF), and private subnets limits external exposure.
* **Application Security:** Kubernetes pod security contexts and service accounts restrict and audit workloads.
# Zero Data Retention
Source: https://docs.hazel.ai/guides/help/privacy-and-security/zero-data-retention
### What is Zero Data Retention?
Zero Data Retention (ZDR) is a privacy and security approach where a service **does not store your content** after it has been processed to complete your request. Your content might include things like text you submit, documents you upload, or information you provide in a workflow. With ZDR, that content is used to generate the result you asked for and then **is not kept** by the provider for later viewing, reuse, or analysis.
### What does Zero Data Retention mean for your data?
When ZDR is enabled, your content is handled **only long enough to deliver the requested output**. After processing, the service is designed so that your content is **not saved to long-term storage** (for example, it isn’t stored in a database or retained as a historical record tied to your account). This reduces the amount of sensitive information that exists within a vendor’s environment over time.
### Why is Zero Data Retention important?
ZDR reduces risk by limiting how much data is available to be exposed later. If content is not stored, there is **less sensitive information “at rest”** in the provider’s systems—meaning less data that could be accessed inappropriately, leaked due to misconfiguration, or obtained in a security incident. It also helps align with confidentiality expectations when you’re sharing personal, financial, or proprietary information.
### Hazel's stance on Zero Data Retention
Hazel is designed with meeting retention controls that let you decide whether to record, store, or delete data such as video, audio, transcripts, and meeting summaries. Hazel also supports a privacy-first approach to AI: your data is not used to train AI models, and you can request deletion at any time.
For AI processing, Hazel has partnered with leading AI providers (including OpenAI and Anthropic) under zero data retention. Hazel uses zero data retention handling where AI subprocessors delete data immediately after first use, keep it only for the minimum time needed to complete processing, and do not use it for model training. For meeting recordings and transcripts, Hazel aims to avoid long-term storage where it isn’t needed, while acknowledging that audio/video files may be briefly stored by transcription partners only to ensure reliable processing, and then promptly deleted.
# Troubleshooting
Source: https://docs.hazel.ai/guides/help/troubleshooting
### Ask Hazel Troubleshooting
Hazel may fail to respond if she reaches chat limits, which happens if the chat thread is extensive. The longer the chat thread and the more files you add to a single thread over time, the higher likelihood it is to max out its token limits. Once a chat fails, it will no longer accept responses.
It's best to switch chats when you are working through different topics. You can do so by clicking on `New Chat` in the top right corner of Ask Hazel.
The Ideas page is how we track feedback, through Canny's feedback portal.
If you are getting the following message "Hazel's feedback is private. Log in to see if you have access," try clearing your cookies/data on the Hazel page for the Canny connection. Then try again by refreshing and re-clicking the Ideas button within Hazel.
Hazel relies on your CRM connection to pull contacts, not Altruist.
The Altruist integration is completely separate from contacts. Meaning if you go into chat today and ask it for custodial data, it would fetch the appropriate Altruist contact and the relevant details, but they would not be listed as a Contact within Hazel.
### Meetings Troubleshooting
You may need to reset Hazel's note taker on various occasions:
* Hazel was not admitted and left the waiting room.
* Hazel was denied / kicked out of meeting, whether on purpose or on accident.
* Hazel experienced connectivity issues.
**How to Reset Hazel from Overview page:** Click on the menu icon for the meeting and click 'Reset Notetaker.' This will clear out any previous data and treat the meeting as if it were brand new.
**How to Reset Hazel from Meeting page:** Click on the menu icon and click **Reset Notetaker**. This will clear out any previous data and treat the meeting as if it were brand new.
If Hazel joined a Zoom call but afterward there was no transcript, it is likely that the host has their **global** local recording setting enabled, but they can't grant participants permission to record locally, which is needed for a bot to record.
**How to fix:** Host should review and enable the following Zoom settings:
1. In Recording tab, verify that `Internal meeting participants` and `External meeting participants` are checked.
2. In Recording tab, verify that Record to computer files is toggled on.
3. While in a Zoom Meeting, verify that `Request Local Recording` Permission is turned on.
If Hazel did not join my Zoom call, nor was she in the waiting room, it is likely that the host has their **global** permissions set to not allow anonymous users, or users not invited to the meeting, to join the meeting. This is needed for a bot to record.
**How to fix:** Host should enable this setting by allowing Anonymous users to join.
If Hazel did not join my Google Meet call, nor was she in the waiting room, it is likely that the host has their **global** permissions set to not allow anonymous users, or users not invited to the meeting, to join the meeting. This is needed for a bot to record.
**How to fix:** Host should review and enable the following Google settings:
1. Under **Meeting Access > Meeting Access type**: verify that the `Anyone with the meeting link can ask to join` checkbox is checked.
If this box is unchecked, then **all anonymous participants and participants not on the calendar invite** will be automatically declined when attempting to join the meeting.
This will appear as the "You can't join this meeting" screen from the bot's perspective:
To manage meeting access permissions for your entire organization, the host (admin) follow these steps:
1. Go to the Google Meet admin center.
2. Expand **Host Controls > Meeting Access**.
3. Under **Who needs to ask to join**, select **Open**.
If Hazel did not join my Microsoft Teams call, nor was she in the waiting room, it is likely that the host has their **global** permissions set to not allow anonymous users, or users not invited to the meeting, to join the meeting. This is needed for a bot to record.
**How to fix:** To manage anonymous meeting join for your entire organization, the host (admin) should follow these steps:
1. Go to the Teams admin center.
2. Expand **Meetings** > **Meeting settings**.
3. Under **Meeting Join and Lobby:** verify that the `Anonymous users can join a meeting` setting is toggled to On.
4. Save your updates.
If you accidentally synced notes to the wrong Household/Contact, here's what you need to do:
1. Within the Meetings page, remove the incorrect profile from the attendees and then add the correct profile.
2. Once the correct profile is selected, click on 'Sync to CRM' and re-sync the notes to the correct profile. Be sure to review the title, household/contact/tags section before you sync to CRM.
3. You'll need to delete the notes from the incorrect profile within your CRM so the incorrect meeting data is not associate the the incorrect household. Head to the Household, then within the Activity section, find the note, and click Delete.
To upload a meeting that failed to upload, navigate to the meeting your iPhone Settings / Apps / Hazel.
Verify there is a stored file associated with the Hazel app. If there is, you can recover a copy by heading to the Files app / search for the Hazel folder / then accessing any saved files.
* If there is a file here, download and upload within the Hazel web app.
* If there is no file here, we do not have any meeting data on file and the recording was lost.
Our visibility into manually recorded meetings is minimal due to streaming related issues (internet connection issues, iOSversion updates, recording settings, mic settings, etc). If there's a failure with the browser, the internet, or unforseen events, then there's a high chance of loss.
To upload a meeting that failed to upload, navigate to the meeting in Hazel, click `Record` and select `Upload Saved Recording`. If we have any reording data on file, we will upload it.
If you do not see an `Upload Saved Recording`, then we do not have any meeting data on file and the recording was lost. [*Reach out to support@hazel.ai for further support.*](mailto:support@hazel.ai)
Our visibility into browser recorded meetings is minimal due to browser related issues (different browsers, version updates, popup blocker settings, mic settings, etc). If there's a failure with the browser, the internet, or unforseen events, then there's a high chance of loss. The best option for in person meetings is to use our iOS app for recording.
Our visibility into browser recorded meetings is minimal due to browser related issues (different browsers, version updates, popup blocker settings, mic settings, etc). If there's a failure with the browser, the internet, or unforseen events, then there's a high chance of loss.
[*Reach out to support@hazel.ai for further support.*](mailto:support@hazel.ai)
**One very important thing to keep in mind if you use browser recording:**
Do not switch tabs while the browser is processing. If you do, your browser may fail to fully cache the recording due to their handling/refreshing/throttling on 'inactive tabs', which may corrupt the file and make it unprocessable.
Best thing to do would be to stay on the Hazel screen for a few seconds until the recording finishes processing and moves to the uploading screen before switching tabs. It's this processing period that is key for the browser to cache your recording.
Once the browser has processed the recording, it still needs to upload the recording to Hazel. Do not close the tab until the recording has finished uploading. You may switch tabs at this stage. Once the file has finished uploading to Hazel, then we have a file on record to process, reprocess, and download as needed.
The best option for recording in person meetings is to use the iOS app for recording.
The most common reason is that the meeting isn’t linked to a Household/Contact (or it has no attendees listed). If a meeting has no attendees, it will only be visible to users who are directly on the calendar event invite.
**How to fix it**
1. Ask the meeting owner to open the meeting in Hazel and review **Meeting Details → Attendees**.
2. Confirm there’s at least one attendee listed and that the meeting is associated with the correct Household/Contact.
* This links the meeting so other team members can view it, and allows Ask Hazel to reference it.
If there *is* an attendee listed: have the meeting owner check the Household/Contact permissions. If you don’t have permission to view the Household/Contact, you won’t be able to view meetings associated with it.
Still not working? Email [support@hazel.ai](mailto:support@hazel.ai) with the meeting title and meeting date/time so we can investigate.
### Phone Call Recording Troubleshooting
The most common cause is calling from a number that isn't on your **Allowed callers** list. Check **Settings → Meeting Settings**. Also check that **Record Phone Calls** is still on, that your caller ID isn't blocked or withheld, and that the call lasted more than about 15 seconds. Very short calls are discarded rather than turned into meetings.
The client was never merged into the call. Hazel records the merged conference, so both parties have to be on it. Check that your phone shows a single merged call, not two separate ones, before you start talking.
Turning **Record Phone Calls** off pauses recording but keeps your number. Turn it back on and you get the same number. If it stays off for an extended period, Hazel releases the number, and turning the setting back on issues a new one. Update your saved contact if that happens.
Check the number, including the country code (`+1` for US numbers), and try again. If your phone declined the call or it went to voicemail, Hazel reports that and you can retry.
Phone call recording isn't turned on for every account yet. Reach out to your Hazel contact if you'd like access.
### Email Troubleshooting
Leverage Hazel’s chat to turn your non-HTML signature into HTML text by:
1. Copying your email signature into a word document
2. Attach the document + the following prompt into Ask Hazel `Hazel, turn the email signature in the document attached into HTML text`
3. Copy the HTML text into Settings → HTML Email Signature
It’s most likely that your HTML email signature includes an image that is not public. Images in your HTML email signature need to be on a publicly accessible url link that points to an image source.
Within your HTML email signature, verify by:
1. CTRL+F `img` and find the `img src` section of text
2. Replace everything in “`double quotes`” with a publicly stored image
Ex. img src="/41235/file\_attachments/5952192” → img src="[https://hazel.ai/images/logo-footer.svg](https://hazel.ai/images/logo-footer.svg)"
If you don't have a publicly stored email, you can upload a logo in Settings → Account → Firm Information, right click on the logo once its loaded, and ‘Copy Image Address’.
Historical email backfill depends on having contacts/households in Hazel (typically from a connected CRM). Hazel uses those contacts to match and sync past email history.
**If you have NOT connected a CRM (or you have no contacts in Hazel)**
* Hazel can’t backfill historical emails because there are no contacts/households to match emails to.
* You *should* still see new inbound screened emails going forward in the Inbox (left sidebar).
* If you’re not seeing new emails either, check **Settings → Account → Email Preferences**. Emails that aren’t associated with a contact/household may be filtered out based on your screening preferences.
Connecting your CRM or adding contacts to Hazel will maximize Hazel's functionality, especially with the Email Assistant.
**If you HAVE connected a CRM (and contacts exist in Hazel)**
Try re-triggering the email sync:
1. Go to **Settings → Integrations**
2. Disconnect your Email integration
3. Refresh the page
4. Reconnect your Email integration to trigger a resync
Email resyncs typically take 5–15 minutes to ingest emails.
Still not working? Email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
Try re-triggering the email sync:
1. Go to **Settings → Integrations**
2. Disconnect your Email integration
3. Refresh the page
4. Reconnect your Email integration to trigger a resync
Email resyncs typically take 5–15 minutes to ingest emails.
Still not working? Email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
### Account Troubleshooting
Email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
If you and a team member are in *two different firm accounts*, you won’t be able to collaborate or see each other’s meetings. Hazel can’t “merge” two firm accounts. To get you on the same firm, Support must deactivate one user’s account, then the remaining firm owner/admin invites that user into the correct firm.
**Example** John and Henry are on separate firm accounts. If Henry’s firm is the one you’re keeping, Hazel Support deactivates John. John won’t be able to log in again until Henry invites him (Settings → Members) and John accepts the invite email.
**How to fix it**
1. Decide which firm account you want to keep.
2. Hazel Support deactivates the user who will be moving.
3. The admin of the firm you’re keeping goes to Settings → Members and sends an invite.
4. The deactivated user must accept the invite from the email to regain access.
**IMPORTANT:** Once deactivated, the user can’t log in until the invite is accepted. If the user signs up again before being invited, they may end up back in a separate firm and the process may need to be repeated.
# Google
Source: https://docs.hazel.ai/guides/integrations/calendar-and-emails/google
Connect your Google account to give Hazel access to your calendar events, emails, and the ability to draft responses to your contacts.
Hazel does not send email on your behalf. Every email draft that Hazel creates requires a human in the loop to review prior to sending
***
### **Email**
Hazel provides pre-drafted responses for emails that are received from contacts in your CRM. Hazel also uses these emails as context to help you prepare for meetings or leverage in Ask Hazel.
### Calendar
Hazel views your calendar and automatically scheduled note takers to join. Hazel reviews past and future events and provides you with agendas for each event.
### Permissions to review and accept:
* View events on all your calendars
* Read, compose, and send emails from your Google account
* See your primary Google Account email address
* Associate you with your personal info on Google
Have questions about the Google integration? Reach out to [support@hazel.ai](mailto:support@hazel.ai) for more information.
# Outlook
Source: https://docs.hazel.ai/guides/integrations/calendar-and-emails/outlook
Connect your Outlook account to give Hazel access to your calendar events, emails, and the ability to draft responses to your contacts.
Hazel does not send email on your behalf. Every email draft that Hazel creates requires a human in the loop to review prior to sending
***
### **Supported Instances**
Microsoft Office 365
Microsoft Enterprise
Microsoft Exchange
Hazel does not support Microsoft Exchange on premise today as it requires custom configuration and setup per-firm.
***
### **Email**
Hazel provides pre-drafted responses for emails that are received from contacts in your CRM. Hazel also uses these emails as context to help you prepare for meetings or leverage in Ask Hazel.
### Calendar
Hazel views your calendar and automatically scheduled note takers to join. Hazel reviews past and future events and provides you with agendas for each event.
### Permissions to review and accept:
* View events on all your calendars
* Read, compose, and send emails from your Outlook account
* See your primary Outlook Account email address
* Associate you with your personal info on Outlook
Have questions about the Outlook integration? Reach out to [support@hazel.ai](mailto:support@hazel.ai) for more information.
# HubSpot
Source: https://docs.hazel.ai/guides/integrations/crms/hub-spot
Connect your HubSpot account to give Hazel access to your contacts, households, tasks, and notes. With a bidirectional sync, updates made in Hazel appear in your CRM, and changes in your CRM are reflected back in Hazel, keeping your data consistent and up to date across both platforms.
Hazel does not import existing roles and permissions from your CRM. It is up to the Owner to edit household permissions for the organization. Learn more about editing household permissions.
***
### **HubSpot Sync Summary**
| **Integration** | **Backfill (History Pulled In)** | **Sync Speed** | **Filters Applied** | **Can Resync?** |
| :-------------- | :------------------------------- | :------------------------------- | :------------------------------------ | :-------------- |
| **HubSpot** | Full history | Every hour (incremental updates) | None (all contact data comes through) | ✅ Yes |
***
The HubSpot integration provides a bi-directional sync for: Tasks, Notes, Contacts, and Households. Hazel instantaneously syncs data to HubSpot, and HubSpot syncs data every 60 minutes to Hazel. Upon the initial sync, Hazel aims to provide a complete backfill of historical data based on HubSpot sync guidelines and imports all contacts with no filters applied.
***
Have questions about the HubSpot integration? Reach out to [support@hazel.ai](mailto:support@hazel.ai) for more information.
# Quivr
Source: https://docs.hazel.ai/guides/integrations/crms/quivr
Connect your Quivr account to give Hazel access to your contacts, households, tasks, and notes. With a bidirectional sync, updates made in Hazel appear in your CRM, and changes in your CRM are reflected back in Hazel, keeping your data consistent and up to date across both platforms.
Hazel does not import existing roles and permissions from your CRM. It is up to the Owner to edit household permissions for the organization. Learn more about editing household permissions.
***
| **Integration** | **Backfill (History Pulled In)** | **Sync Speed** | **Filters Applied** | **Can Resync?** |
| :-------------- | :------------------------------- | :------------------------------- | :-------------------------------------------------------------------------------------------------------------------------------- | :-------------- |
| **Quivr** | Full history | Every hour (incremental updates) | Must have: Client Type = **Household**, and Record Type = **Client Household**, **On-Boarding Client Household**, or **Prospect** | ✅ Yes |
The Quivr integration provides a bi-directional sync for: Tasks, Notes, Contacts, and Households. Hazel instantaneously syncs data to Quivr, and Quivr syncs data every 60 minutes to Hazel. Upon the initial sync, Hazel aims to provide a complete backfill of historical data based on Quivr sync guidelines and imports all contacts based on the following filters: Client Type = **Household**, and Record Type = **Client Household**, **On-Boarding Client Household**, or **Prospect.**
Have questions about the Quivr integration? Reach out to [support@hazel.ai](mailto:support@hazel.ai) for more information.
# Redtail
Source: https://docs.hazel.ai/guides/integrations/crms/redtail
Connect your Redtail account to give Hazel access to your contacts, households, tasks, and notes. With a bidirectional sync, updates made in Hazel appear in your CRM, and changes in your CRM are reflected back in Hazel, keeping your data consistent and up to date across both platforms.
Hazel does not import existing roles and permissions from your CRM. It is up to the Owner to edit household permissions for the organization. Learn more about editing household permissions.
***
| **Integration** | **Backfill (History Pulled In)** | **Sync Speed** | **Filters Applied** | **Can Resync?** |
| :-------------- | :------------------------------- | :------------- | :------------------------------------ | :-------------- |
| **Redtail** | Partial (not full history) | Every 24 hours | None (all contact data comes through) | ✅ Yes |
The Redtail integration provides a bi-directional sync for: Tasks, Notes, Contacts, and Households. Hazel instantaneously syncs data to Redtail, and Redtail syncs data every 24 hours to Hazel. Upon the initial sync, Hazel partially backfills historical data based on Redtail’s sync guidelines and imports all contacts with no filters applied.
***
### Ask Hazel about Redtail Activities
In addition to the synced data above, Hazel can read your Redtail **Activities** directly when you ask in chat. Activities are the structured workflow records behind your client onboarding pipeline, service tickets, follow-ups, and any other categorized work — so they’re the right source for workflow-status questions (use Activities, not notes, for these).
You can ask by category, status, assignee, and date range. For example:
* *“Who is currently in Client Onboarding?”*
* *“List my open Service activities.”*
* *“Show activities assigned to me starting in the next 30 days.”*
* *“List completed activities between 2026-01-01 and 2026-03-31.”*
Activity lookups are available to firms whose CRM is Redtail. If your firm uses a different CRM, Hazel will let you know and fall back to notes, tasks, or meetings as appropriate.
### Scope chat questions by advisor role
Hazel can also filter households by the advisor roles assigned in Redtail (Lead Advisor, Senior Advisor, Paraplanner, CSA, etc.) so you can scope chat questions to your own book or to a colleague’s. See [Best Practices](/guides/ask-hazel-chat/best-practices) for examples.
***
Have questions about the Redtail integration? Reach out to [support@hazel.ai](mailto:support@hazel.ai) for more information.
# Salesforce
Source: https://docs.hazel.ai/guides/integrations/crms/salesforce
Connect your Salesforce account to give Hazel access to your contacts, households, tasks, and notes. With a bidirectional sync, updates made in Hazel appear in your CRM, and changes in your CRM are reflected back in Hazel, keeping your data consistent and up to date across both platforms.
Hazel does not import existing roles and permissions from your CRM. It is up to the Owner to edit household permissions for the organization. Learn more about editing household permissions.
***
### **Salesforce Sync Summary**
| **Integration** | **Backfill (History Pulled In)** | **Sync Speed** | **Filters Applied** | **Can Resync?** |
| :-------------- | :------------------------------- | :------------------------------- | :------------------------------------ | :-------------- |
| **Salesforce** | Full history | Every hour (incremental updates) | None (all contact data comes through) | ✅ Yes |
***
The Salesforce integration provides a bi-directional sync for: Tasks, Notes, Contacts, and Households. Hazel instantaneously syncs data to Salesforce, and Salesforce syncs data every 60 minutes to Hazel. Upon the initial sync, Hazel aims to provide a complete backfill of historical data based on Salesforce sync guidelines and imports all contacts with no filters applied.
### **Salesforce instances accepted by Hazel:**
✅ Salesforce Classic
✅ Salesforce Lightning
✅ Salesforce FSC (Financial Services Cloud)
***
### **Salesforce Object Query Language**
Hazel’s chat supports **Salesforce Object Query Language (SOQL)**, Salesforce’s proprietary SQL-like query language for retrieving data from Salesforce. SOQL enables precise, structured queries—filtering by field values, combining criteria, sorting results, and pulling related records—so you can quickly find exactly the Salesforce records you need (based on your organization’s configured access and permissions).
***
Have questions about the Salesforce integration? Reach out to [support@hazel.ai](mailto:support@hazel.ai) for more information.
# Wealthbox
Source: https://docs.hazel.ai/guides/integrations/crms/wealthbox
Connect your Wealthbox account to give Hazel access to your contacts, households, tasks, and notes. With a bidirectional sync, updates made in Hazel appear in your CRM, and changes in your CRM are reflected back in Hazel, keeping your data consistent and up to date across both platforms.
Hazel does not import existing roles and permissions from your CRM. It is up to the Owner to edit household permissions for the organization. Learn more about editing household permissions.
***
| **Integration** | **Backfill (History Pulled In)** | **Sync Speed** | **Filters Applied** | **Can Resync?** |
| :-------------- | :------------------------------- | :------------------------------- | :-------------------------------------------------------------------------------------------------------- | :-------------- |
| **Wealthbox** | Full history | Every hour (incremental updates) | Must have: Type = **Person** or **Household,Active = true**, andContact Type = **Prospect** or **Client** | ✅ Yes |
The Wealthbox integration provides a bi-directional sync for: Tasks, Notes, Contacts, and Households. Hazel instantaneously syncs data to Wealthbox, and Wealthbox syncs data every 60 minutes to Hazel. Upon the initial sync, Hazel aims to provide a complete backfill of historical data based on Wealtbox sync guidelines and imports all contacts based on the following filters: Type = **Person** or **Household,Active = true**, andContact Type = **Prospect** or **Client.**
Have questions about the Wealthbox integration? Reach out to [support@hazel.ai](mailto:support@hazel.ai) for more information.
# Altruist
Source: https://docs.hazel.ai/guides/integrations/custodians/altruist
## Setting up the Hazel x Altruist Custodial Integration
Log in to Hazel. Head to Settings / Integrations and connect to Altruist.
Enter your Altruist login credentials and you'll be connected. Once you've done so, you can head into Ask Hazel and ask questions about your custodial data as it pertains to a household.
## Use Cases
Hazel has read-only access to your custodial data in Altruist.
At the moment, Hazel works best when you ask questions about a particular account, or household, as opposed to questions that need aggregation across households.
Firm-wide awareness on your custodial data is coming in Q1 2026.
### Examples you can try today
* Show me the Brito household accounts, balances, last transactions, and beneficiaries.
* Based on my last meeting with the Johnson, should I rebalance their portfolio?
* "List the beneficiaries for the Smith household."
* "What are their accounts and balances?"
* "Build a performance summary for this household"
* "What’s the RMD status for their accounts?"
## Authorization, Data Access, and Security
### **Data Access Consent**
By connecting your Altruist account, you authorize Hazel to:
* Access account and holdings information
* Retrieve transaction history and performance data
* Fetch cost basis and tax lot information
* Read portfolio composition and allocations
### **Security & Privacy**
* **No Storage:** Hazel does not store any Altruist data
* **Read-Only Access:** Hazel can only read data and cannot execute trades or transfers
* **Secure Authentication:** Uses OAuth2 with PKCE for maximum security
* **Revocable Access:** You can disconnect at any time from this settings page
# Overview
Source: https://docs.hazel.ai/guides/integrations/overview
## Integration Sync Summary
| **Integration** | **Backfill (History Pulled In)** | **Sync Speed** | **Filters Applied** | **Can Resync?** |
| :-------------- | :------------------------------- | :------------------------------- | :------------------------------------------------------------------------------------------------------------------------------------- | :-------------- |
| **Gmail** | 3 years of past emails | Near-instant (event-driven) | Only emails linked to CRM contacts | ❌ No |
| **Outlook** | 3 years of past emails | Near-instant (event-driven) | Only emails linked to CRM contacts | ❌ No |
| **Redtail** | Partial (not full history) | Every 24 hours | None (all contact data comes through) | ✅ Yes |
| **Wealthbox** | Full history | Every hour (incremental updates) | Must have: Type = **Person** or \*\*Household, Active = true\*\*, and Contact Type = **Prospect** or **Client** | ✅ Yes |
| **Salesforce** | Full history | Every hour (incremental updates) | None (all contact data comes through) | ✅ Yes |
| **Quivr** | Full history | Every hour (incremental updates) | Must have: Client Type = **Household**, and Record Type = **Client Household**, **On-Boarding Client Household**, or **Prospect** | ✅ Yes |
| **HubSpot** | Full history | Every hour (incremental updates) | None (contact data is selected upon connection) | ✅ Yes |
| **Altruist** | No history (live feed) | Near-instant (event-driven) | Access: Accounts, balances, transactions, and beneficiary information. | ❌ No |
# Introduction
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/Overview/Introduction
## What is an Intelligent Client Profile?
Hazel builds an Intelligent Client Profile for every household and contact: a living record of the lasting facts about each client. Who's in the family, what they earn and own, how they're protected, what they're working toward, all captured automatically from your recorded meetings, the emails clients you send, the documents you upload, and the notes you write.
Every fact carries its source. Click any value and Hazel shows you where it came from: the meeting, email, or document, and the date. Facts heard in a meeting or read in an email come with the client's own words; facts from a document link to the file itself. Nothing is a black box.
## Why it matters
* **Nothing gets lost between meetings.** A detail mentioned once in March is still there in November, filed under the right person, with the date it was said.
* **Hazel actually uses it.** Ask Hazel about a client and the profile is the first place Hazel looks: on the household page, in chat, in meeting prep, and in financial planning. The household Summary tab is written from it, sentence by sentence, with clickable citations. A better profile makes everything Hazel does for that client better.
* **Update it by talking to Hazel.** "The Hendersons paid off the HELOC and opened a 529 for Maya." Tell Hazel what changed, review the updates Hazel lays out, confirm, and the profile is current. No forms, no fields.
* **You stay in charge of quality.** New information lands in a Review queue where you can confirm it in one click. Verified facts are treated as confirmed; anything unverified is clearly labeled, and gets called out as unverified whenever Hazel uses it, rather than presented as settled.
* **Your whole team works from one profile.** Everyone with access to the client sees the same profile. When one of you confirms a fact or fixes a number, it's fixed for everyone.
## How it works, in one minute
1. You meet with a client (with the Hazel notetaker recording), a client emails you, you write a note, or you upload a statement or tax return to the **Documents** tab.
2. Shortly after, Hazel adds any lasting facts to the client's **Data** tab, each with its source and, for conversations, a verbatim quote.
3. New information is flagged in **Review**, where you confirm what's correct and what isn't, and tell Hazel who a fact belongs to when that wasn't clear.
4. Between meetings, you can keep profiles current by simply telling Hazel in plain English what changed, and confirming the updates.
5. Over time, each client builds a complete, source-backed profile without you typing it into forms, and everything Hazel does for that client gets smarter.
Hazel records only what was actually said or written, never a guess, an estimate, or a calculation. If a client says "about \$2 million across the IRAs," that's what gets recorded, not a computed total or a projection.
***
## Where to go next
* New to it? Start with [Getting Started](/guides/intelligent-client-profiles/overview/getting-started).
* Need a Demo? Watch the [ICP Webinar Replay](https://www.loom.com/share/e3cab34f053b4343b16355e04da1c09e).
* Ready to work with the data? See [Managing Client Data](/guides/intelligent-client-profiles/managing-client-data).
* Want the habits that make profiles great? See [Best Practices](/guides/intelligent-client-profiles/best-practices).
* Stuck or curious? See [Troubleshooting & FAQ](/guides/intelligent-client-profiles/troubleshooting-and-faq).
# Getting Started
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/Overview/getting-started
## Before you start
Intelligent Client Profiles build themselves from your everyday workflows, so there's nothing to configure. Two things determine how quickly they fill in:
1. **Your sources of client information.** Profiles learn from meetings recorded by the [Hazel notetaker](/guides/meeting-assistant/overview/how-it-works), email and CRM connected in **Settings → Integrations**, client documents you upload, and the notes you write in Hazel. Any one of these sources help build your profiles and the more sources you include, the richer the profile.
2. **Your clients in Hazel.** Facts are filed under your Households and Contacts, so the people you meet with need to exist as contacts (Households and Contacts typically sync from a connected CRM, but you can also create them in Hazel).
## Finding the profile
Click **Contacts** in the Hazel sidebar, pick a household from the **Households** list (or a person under **People**), and open it.
A household page shows **Summary · Meetings · Tasks · Emails · Notes · Documents · Data**. The profile lives across three of these: **Summary** (the narrative view), **Documents** (files Hazel has read), and **Data** (every fact Hazel has collected, with sources). Individual client pages have Documents and Data tabs, but no Summary tab.
A useful way to think about the three: **Summary** is where you *read* the profile, **Documents** is where you *feed* it, and **Data** is where you *maintain* it.
## The Summary tab: your household at a glance
The Summary tab is the profile turned into prose, designed to get you current on a household in about a minute.
* **Summary** is a narrative Hazel writes for each the profile. Phrases with citations are underlined: Click one to see exactly which meeting, email, or document the statement came from. The summary narrative refreshes itself as new facts arrive.
* **Tasks** and **Meetings** list the household's open work and recent meetings, in place.
* **About Household** covers key dates, a **Goals** card (each goal as it was recorded), and a card for each member with a short bio Hazel maintains.
* **Ask Hazel**, the chat rail on the right of each profile, come prepared to help answer specific questions about each household. Ask anything, and Hazel will draw on data from the profile to answer. It works in both directions: If you want to manually add an update to a profile, you can tell Hazel what's changed ("they sold the rental in June"), review the updates Hazel lays out, and confirm.
Next to the household's name you'll also see a small **Data Verification** ring showing how much of the profile you've confirmed. Click it to review new information without leaving the page.
## Your first visit
If Hazel hasn't processed anything for a client yet, the Data tab shows:
> **No household data yet.** "Hazel builds this page from the household's meetings, emails, and uploaded documents. Once a meeting is recorded or an email is synced, the facts Hazel extracts land here — every value with its source."
That's normal. Here are four ways to start building data into a profile, in order of speed:
1. **Upload a document.** Drop a statement, tax return, or account document on the **Documents** tab. Facts usually land within minutes.
2. **Tell Hazel.** In Ask Hazel, describe what you already know about the client (accounts, goals, family), review what Hazel proposes to add, and confirm.
3. **Record your next meeting** with the Hazel's notetaker. Facts appear in the corresponding household profile shortly after your meeting notes are ready.
4. **Connect email**, and facts arrive as clients write to you.
Then spend five minutes in **Data → Review** confirming what Hazel found. See [Managing Client Data](/guides/intelligent-client-profiles/managing-client-data).
## Who can see and change a profile
Profiles follow the same access rules as the rest of a client's record in Hazel:
* **If you can open the client, you can read their profile.** The household Data tab covers every member, so it requires access to the whole household.
* **If you can edit the client, you can maintain their profile**: verify, correct, archive, upload and delete documents.
* **Owners** can see and manage every client's profile, as they can everywhere in Hazel. The Admin role by itself manages firm settings, not client data.
* **There's one shared profile per client**, no private copies. A teammate's verifications and corrections are visible to you, and yours to them, so shared clients get one clean record instead of competing versions. Clients never see their profile; it's internal to your firm.
***
## Where to go next
* New to it? Start with [Getting Started](/guides/intelligent-client-profiles/overview/getting-started).
* Need a Demo? Watch the [ICP Webinar Replay](https://www.loom.com/share/e3cab34f053b4343b16355e04da1c09e).
* Ready to work with the data? See [Managing Client Data](/guides/intelligent-client-profiles/managing-client-data).
* Want the habits that make profiles great? See [Best Practices](/guides/intelligent-client-profiles/best-practices).
* Stuck or curious? See [Troubleshooting & FAQ](/guides/intelligent-client-profiles/troubleshooting-and-faq).
# Best Practices
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/best-practices
A great profile doesn't require extra work. It requires doing the work you already do in a way Hazel can capture. The habits below are ranked by payoff.
## Seed every client with documents on day one
The single highest-leverage move. Facts from documents arrive **already verified**, so ten minutes of uploading gives a new client a strong, trustworthy profile before your first meeting, instead of one that accumulates over months of conversations.
The most recent custodial or brokerage statements, the latest tax return, insurance policy declaration pages, and estate documents (will, trust). A pay stub or equity award statement rounds out the picture for working clients.
Drop everything on the **household's** Documents tab so facts are shared across the family. Use an individual's page only for documents that belong to them alone.
Upload in any order. Hazel dates facts by the document's own date when the document shows one; otherwise the upload day is used, and the date on the document row shows which. Newer information wins, so a dated old statement won't overwrite a fresher number. On older uploads, glance at the row date rather than trusting the order to sort itself out.
When documents read **In profile**, open the **Data** tab and give it two minutes. Assign any facts waiting under **Needs owner**, verify them as you file them, and you're done.
Export documents fresh from the source (custodian portal, tax software) rather than scanning paper. Clean, text-based PDFs read far more reliably than photographed or scanned pages, and they stay under the 10 MB per-file limit.
## Talk so the profile writes itself
Hazel records what was actually said, so *how* you and your clients talk in recorded meetings directly shapes the profile. Four habits make capture dramatically cleaner:
* **Say numbers and institutions out loud.** "So the Schwab brokerage is at about \$840,000 as of this month" refreshes a balance for free. A number that appears on screen but is never spoken can't be captured.
* **Name the owner.** Ownership is filed from the words used: "*my* 401(k)", "*Sarah's* Roth", and "*our* joint account" all file correctly. "The 401(k)" leaves Hazel unwilling to guess, and the fact waits in **Review → Needs owner** for you. Speaking possessively saves you the triage.
* **Get the "no" on record.** "So just to confirm, no long-term care coverage today?" turns a silence into a recorded answer ("None — confirmed with client"), which is often as valuable in planning as the assets themselves.
* **Recap to capture.** A 60-second verbal summary at the end of a meeting ("so we're opening the 529 at \$450,000, not \$415,000" said aloud updates the profile naturally.
Hypotheticals stay out by design. "Clients like you typically save around \$50k" won't be recorded as a fact about this client. Speak plainly about what *is* true and let illustrations be illustrations.
## Update the profile straight from chat
Between meetings, the fastest way to keep a profile current is to just **tell Hazel**. In Ask Hazel, including the chat rail on the household page, describe what changed in plain English:
> "The Hendersons paid off the HELOC and opened a 529 for Maya at Fidelity: \$500 a month."
> "Update Bob's 401(k) balance to \$612,000 as of yesterday."
> "Add a concern: Marianne is worried about her mother needing memory care."
Hazel shows you exactly what will be added or changed; confirm, and the updates land in the profile as **your entries, verified immediately**, just like edits on the Data tab. Nothing changes without your confirmation, and everything stays source-linked.
This is also the way to **add something brand new**: an account, a goal, a policy the client mentioned in a hallway conversation, without waiting for a meeting or a document. Make it a habit: hang up the phone, tell Hazel, move on.
Prefer writing things down? Notes work too: a quick note in Hazel after a call feeds the profile just like a meeting would. Chat gives you a confirmation step; a note is capture-and-go. Use whichever fits the moment. The point is that nothing you learn should live only in your head.
## Build a review rhythm
Verification is what turns "what Hazel heard" into "what your firm knows." It works best as a small, regular habit, not a project:
* **After each meeting: two minutes in Review.** The Data tab opens on **Review** with exactly what's new: the quote, the source, a **Verify all** button. Confirm what's right, **Edit** what's off, assign owners. Done.
* **Verify what you'd act on.** You don't need 100%. Prioritize the financial figures that feed decisions. Hazel will still use unverified facts, but flags them as unverified until someone confirms them, and the flag is only useful if verified actually means checked.
* **Watch the ring.** The **Data Verification** ring on the household page shows how much of the profile is confirmed at a glance: a quick health check before a client meeting or a planning session.
* **Check "as of" dates before acting.** The profile shows the most recent *stated* value, not a live feed. A balance "as of Jan 2026" is telling you to ask, upload, or update before you rely on it.
## Work as a team
There's one shared profile per client, so a little coordination goes a long way:
* **The meeting owner does the review pass.** Whoever ran the meeting spends the two minutes in Review afterwards. Their memory of the conversation is freshest, and one person's pass covers the whole team.
* **Delegate maintenance freely.** Anyone with edit access to the client can verify, correct, and upload, including assistants. Verification carries firm-wide, so a well-run operations seat can keep every profile clean.
* **Onboard the profile, not just the tool.** When a new team member takes over a household, the Summary tab plus five minutes on the Data tab is the fastest handover in the firm: every fact sourced, dated, and attributed.
## Habits that pay off
**Do:** upload the document packet on day one · say numbers, owners, and institutions out loud · confirm what clients *don't* have · recap decisions at the end of meetings · tell Hazel about changes the moment you learn them (chat or a quick note) · spend two minutes in Review after each meeting · check "as of" dates before acting on a number.
**Skip:** retyping what a document already proves (upload it instead) · verifying everything indiscriminately (verify what you'd act on) · using **Rebuild profile** for a single wrong value (that's what **Edit** is for) · scanning paper when a clean PDF export exists.
***
Something not behaving as expected? See [Troubleshooting & FAQ](/guides/intelligent-client-profiles/troubleshooting-and-faq).
# Adding Client Data
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/managing-client-data/adding-client-data
## Where facts come from
An Intelligent Client Profile learns from five places:
1. **Recorded meetings.** When the Hazel notetaker records a client meeting, Hazel reviews the conversation shortly after your meeting notes are ready and captures any lasting facts, each backed by a verbatim quote.
2. **Emails from your clients.** With email connected, Hazel captures facts from emails your clients send you. Only substantive messages count; scheduling back-and-forth and confirmations are ignored. Hazel matches emails by the address they're sent from, so it needs to match the email on the client's contact record. (Emails *you* send aren't captured.)
3. **CRM notes and data fields.** With CRM connected, Hazel captures facts from notes and data fields you store in your CRM. A quick note added in your CRM ("they closed on the lake house, \$1.1M") is captured like a meeting would be.
4. **Documents you upload.** Statements, tax returns, and account documents added to the **Documents** tab are read page by page and filed into the profile. Because a document is hard evidence, facts from documents arrive already **verified**.
5. **You, directly.** Correct any value on the Data tab, or tell Hazel what changed in chat and confirm the updates. Either way, your entries become part of the profile immediately, marked as yours and verified.
A few guardrails keep the profile honest:
* **No math, no guesses.** Hazel never sums balances, projects growth, or infers a tax bracket. A stated age becomes an approximate birth year (marked "\~"), never a computed age that drifts out of date. Hypotheticals ("clients like you typically save…") aren't recorded.
* **Ownership comes from what was said.** "*His* 401(k)" and "*our* joint account" file correctly; who happened to be speaking is never used. When ownership isn't clear, Hazel asks you instead of guessing.
* **Your CRM stays in charge of contact basics.** For clients synced from your CRM, identity and profile fields (names, birth dates, addresses, risk tolerance, and so on) are managed in the CRM. Hazel won't change them, even when the CRM field is empty. For clients that live only in Hazel, it's the reverse: the profile keeps the contact record up to date automatically.
* **"No" is an answer.** If a client confirms they don't have something (no will, no umbrella policy), the profile records "None — confirmed with client," which is often as valuable as the facts themselves.
Documents uploaded in an Ask Hazel **chat** help Hazel answer in that conversation, but only documents added to a client's **Documents** tab become part of the profile.
## Uploading documents
Open the client's **Documents** tab ("Statements, tax returns, and account documents Hazel has read into this profile.") and drop files anywhere on the tab, or click the dashed **"Upload your client documents here"** strip.
* **Accepted:** PDF, Word (DOCX), Excel (XLSX), and images, up to **10 MB** each. (Legacy .doc/.xls files need to be re-saved as .docx/.xlsx first.)
* **Password-protected PDFs** prompt you for the password; it's used on your device to unlock the file and never leaves your browser.
* **Uploading to a household** shares the document (and its facts) across all members; uploading on a person's page links it just to them.
Each document shows its progress: **Queued → Reading → Adding to profile → In profile**. When it reads **In profile** (green), click it to jump to the Data tab and see what Hazel learned. Documents are added one at a time per household to keep the profile consistent, so a busy household may briefly show **"Waiting for profile"**. A **Failed** pill means Hazel couldn't process the file; upload a fresh copy or a clearer version.
Two details worth knowing:
* **Dates matter.** Facts from a document carry the document's own date when Hazel can find one; otherwise they're dated to the upload day. The date on the document row shows which was used, and you can set it yourself. That's worth doing for older statements whose date Hazel couldn't read, because newer information wins in the profile and the date decides what counts as newer. See [how updates work](/guides/intelligent-client-profiles/managing-client-data/reviewing-and-verifying#how-updates-and-corrections-work).
* **Deleting a document removes what Hazel learned from it.** The delete confirmation says it plainly: "Deleting this document will also remove the information Hazel extracted from it." Values it had replaced come back.
## Updating the profile from chat
The fastest way to keep a profile current is to tell Hazel directly. In Ask Hazel, including the chat rail on the household page, describe what changed in plain English:
> "Update Bob's 401(k) balance to \$612,000 as of yesterday."
> "The Hendersons paid off the HELOC and opened a 529 for Maya at Fidelity: \$500 a month."
Hazel works out what that means for the profile and shows you the exact updates planned, down to which records and which values. Confirm, and they're applied. Three things to know:
* **Nothing changes without your confirmation.** Hazel proposes; you approve. Decline and the profile is untouched.
* **Chat updates are your entries.** They land verified and marked as entered by you, exactly like an **Edit** on the Data tab, and they appear in each value's History like any other change.
* **You can add brand-new records this way.** An account the client mentioned on a phone call, a new goal, a policy: describe it, confirm it, and it takes its place in the right section. Chat is the way to hand-add facts that no meeting or document has captured yet.
This works for cleanup, too: "archive the old Vanguard IRA, it was rolled over," "the mortgage rate is 5.875, not 5.375," or "fix the owner on the term policy, it's Sarah's" are often quicker in chat than clicking through the Data tab.
***
Once data is flowing in, the next step is checking it. See [Reviewing & Verifying](/guides/intelligent-client-profiles/managing-client-data/reviewing-and-verifying).
# Arching & Rebuilding
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/managing-client-data/archiving-and-rebuilding
## Archiving: when something ends
Accounts close, goals get reached, policies lapse. Hover a record and click the archive icon: **Archive** works on any record ("Mark completed" appears on goals). Archived records move to the **Archive** view ("Archived data does not populate your client's profile."), grouped by when they were archived, still fully readable with their sources and history. **Unarchive** brings one back.
Hazel can archive too, when a client says something ended; click the badge on an archived record to see why. The reverse also happens: if a later meeting or document shows something is active again, Hazel brings the record back. Archives follow the most recent statement, just like values.
## Removing: when something was never true
If a record was never true at all (misheard, or a hypothetical that slipped through), **remove** it instead of archiving. Removal is for things that never existed: Hazel asks you to confirm, and unlike archiving it's permanent. The record disappears entirely rather than moving to the Archive.
Rule of thumb: **ended → Archive. Never existed → Remove.** When in doubt, archive; you can always remove later, but not the reverse.
## Rebuilding a profile
If a profile has gone wrong in a way that spot-fixes can't repair, **Rebuild profile** (at the bottom of the Data tab's left rail) erases everything Hazel currently knows and relearns it from the client's meetings and uploaded documents, oldest to newest. Because the replay runs in order, each pass builds on the last, the same way the profile would have grown originally.
Rebuilding is a hard reset: it erases your manual edits and verifications, empties the Archive, and facts that came only from **emails or notes** are not relearned. The confirmation asks you to acknowledge this before it starts. Use it as a last resort; for one wrong value, **Edit** is the right tool.
One good use for it: a long-standing client whose profile is still new. On an empty profile there's nothing to lose, and a rebuild replays the household's full recorded meeting history plus every uploaded document, so years of context file in at once.
***
Hit something that doesn't behave the way you expect? See [Troubleshooting & FAQ](/guides/intelligent-client-profiles/troubleshooting-and-faq).
# Reviewing & Verifying
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/managing-client-data/reviewing-and-verifying
## The Data tab: the full client profile
The **Data** tab holds every fact with its source. On the left you'll see **Review**, six sections, and **Archive**; the orange badge next to a section counts its unverified facts.
| Section | What's inside |
| --------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **People** | The household's members, plus income, expenses, goals, concerns, preferences, life events, charitable giving, and other professionals like the client's CPA |
| **Financial Accounts** | Investment accounts and cash, the holdings inside them, and equity compensation |
| **Property & Other Assets** | Real estate, private and alternative investments, and other assets |
| **Liabilities** | Mortgages, consumer debt, and other obligations |
| **Insurance** | Life, disability, long-term care, property, and health coverage, plus annuities |
| **Estate & Entities** | Businesses, trusts and legal entities, wills, directives, and beneficiary designations |
Inside a section, each row is one record: an account, a property, a goal, a policy. Click a row to expand it. Every value has a small toolbar:
* **Source** shows where the value came from: a **Meeting**, **Email**, **Document**, or **Manual** badge and the date. Meeting and email facts include the client's words, with a **Show context** option that opens the surrounding conversation; meeting sources link to the meeting, and document sources link to the file itself.
* **History** shows every value this fact has ever had, with dates and sources, and a **Restore** option for earlier values.
* **Edit** corrects the value in place. "Saving marks this value verified."
* **Verify** / **Unverify** is your trust mark; more below.
Use the **Household / member** switch at the top right to view one person's slice of the profile; a joint fact appears under each owner. Balances and amounts show an "as of" date so you always know how fresh a number is. Values with a dotted underline aren't ordinary captured facts; hover one to see why. Some are implied from other facts (like an account's tax treatment implied by the account type), and some come straight from the contact record.
## Reviewing new information
The Data tab opens on **Review** ("Review new information before adding it to the client profile."). It collects two kinds of items:
* **Needs owner**: facts Hazel captured but couldn't attribute, e.g. "3 facts need an owner assigned so they're filed under the right person." Click **Assign owner**, pick the member(s), done. Assigning is about filing the fact under the right person; the fact is part of the profile either way.
* **Unverified facts**, grouped by section. Each card shows the new values, the quote they came from, and the source. Click **Verify all** to confirm a record's new values (holdings inside an account are verified on their own rows), or expand to edit individual values first.
When the queue is empty you'll see **"All caught up"**. A quick pass after each meeting keeps the profile trustworthy, and because the profile is shared, one person's review covers the whole team.
## What verifying means
Every value is either **Verified** or **Unverified** (captured from conversation, not yet confirmed). Records with a mix show "2 / 5 verified."
Hazel uses both, but differently. When Hazel draws on the profile in chat or planning, verified facts are presented as confirmed, while unverified facts are explicitly flagged as unverified so you never mistake a half-heard number for a checked one. Verifying is how you (and Hazel) tell them apart.
Some facts arrive already verified: goals, concerns, preferences, life events, and identity details (the client's own words about themselves), facts from documents, and anything you enter yourself. Financial details heard in conversation start unverified until someone confirms them.
Made a mistake? **Unverify** puts a value back in the unverified pool, **Edit** replaces it with the correct one, and **History → Restore** brings back an earlier value (it returns with the verification status it had at the time).
## How updates and corrections work
Clients update things, and the profile keeps up: **the most recent statement of a fact wins, by source date.** If a client gives a new 401(k) balance in July, it replaces the March number, and the March value moves into History, never thrown away. This applies to values you've verified too: verification marks trust, it doesn't freeze the value. Meanwhile, out-of-date sources can't do damage. A January statement uploaded in August won't overwrite anything fresher, and re-statements of what's already known don't churn the profile.
If Hazel filed something under the wrong person, expand the record's **Owner** row and click **Fix owner**. Ownership always goes through this chooser, and a fact always keeps at least one owner.
***
When something ends (or was never true to begin with), see [Archiving & Rebuilding](/guides/intelligent-client-profiles/managing-client-data/archiving-and-rebuilding).
# What ICP Can & Can't Do
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/managing-client-data/what-icp-can-and-cant-do
A few items to keep in mind while you build and maintain ICP:
* **It records; it doesn't calculate.** No net worth totals, projections, or tax estimates. Every number was stated by someone, so every number is auditable.
* **One current value per fact**, with prior values in History. It's not a time-series or performance tool, and it doesn't pull live balances from custodians.
* **The Data tab has no create-a-record form.** You can correct any value there, but adding something brand new happens through conversation: tell Hazel in chat, write a note, upload a document that shows it, or bring it up in a recorded meeting.
* **The profile builds from what Hazel can see.** A meeting without the notetaker, or a conversation that never makes it into a note, email, or document, leaves no trace to capture. For a client's pre-Hazel history, documents are the fastest way in.
* **Dollar amounts are recorded in USD.**
* **The structure is fixed.** No custom sections or fields, so every client's profile is organized the same way.
***
For tips and tricks, see [Best Practices](/guides/intelligent-client-profiles/best-practices).
# Troubleshooting and FAQ
Source: https://docs.hazel.ai/guides/intelligent-client-profiles/troubleshooting-and-faq
### Intelligent Client Profiles FAQ
In chat, in meeting prep, in financial planning, and on the household Summary tab. Ask Hazel (including the chat rail on the household page) reads the profile first when a question concerns a client; meeting prep and financial planning start from it. The Summary tab is written from it, sentence by sentence, with citations. It works in both directions: you can also update the profile from chat by telling Hazel what changed and confirming. When Hazel uses a fact you haven't verified, the fact is flagged as unverified rather than presented as settled. Archived records are left out entirely.
Tell Hazel, in plain English: "Update Bob's 401(k) balance to $612,000," "they opened
a 529 for Maya at Fidelity, $25,000 to start," or "fix the owner on the term policy, it's Sarah's." Hazel shows you the exact updates planned; confirm, and they're applied as your entries: verified, and visible in each value's History. Nothing changes without your confirmation, and you need edit access to the client, just like on the Data tab.
Your CRM is where you *put* information; the profile is what Hazel *learns* from meetings, client emails, documents, and notes, including things CRMs rarely hold (goals, concerns, family context, "we're thinking about selling the rental"), each tied to the conversation or document it came from. It complements your CRM and never writes to it.
No. Hazel can use unverified facts; they're just labeled as unverified whenever they come up. Verifying is how you promote what Hazel *heard* into what your firm has *confirmed*, so a practical habit is a one-minute pass through **Review** after each meeting: verify what's right, correct what isn't. Facts from documents, and anything you enter yourself, arrive already verified.
Yes. The profile always keeps the most recent statement of a fact, by source date. If a client gives a new number next month, it replaces the one you verified, and the old value moves into History. Verification marks a value as confirmed; it doesn't freeze it. The same goes for archived records: a later meeting or document showing something is active again brings it back. What *can't* happen is an older source overwriting newer information. An out-of-date statement or an old document never displaces fresher information.
It's a hard reset: Hazel erases everything currently known about the household, then relearns it from the recorded meetings and uploaded documents, oldest to newest. Three things to understand before using it: your manual edits and verification marks are erased; the Archive is emptied; and facts that came only from **emails or notes** are not relearned. It usually takes a few minutes (longer for a long history) and you can keep working while it runs. Use it when a profile has gone wrong beyond spot-fixing; for one wrong value, **Edit** is the right tool.
PDF, Word (DOCX), Excel (XLSX), and images, up to 10 MB per file, on the client's **Documents** tab. Legacy .doc and .xls need re-saving in the modern format first. Password-protected PDFs prompt for the password, which is used on your device and never leaves your browser.
The facts Hazel extracted from it are removed from the profile too; the delete confirmation warns you. If a document's fact had replaced an earlier value, the earlier value comes back. The upload itself is gone; re-upload if you deleted by mistake.
Yes, through chat. Tell Hazel about the new account, goal, or policy ("add a term life policy for Sarah, $2M through Banner, $180 a month"), review what Hazel proposes, and confirm; it takes its place in the right section as your verified entry. The Data tab itself has no create-a-record form (it's for correcting and managing what's already captured), so for new facts, use chat, write a note, upload a document, or bring it up in a recorded meeting.
If something ended (account closed, goal reached), **Archive** the record. It stops populating the profile, stays readable with its history, and can be unarchived. If a value is wrong, **Edit** it. If a whole record was never true (misheard, or entered in error), **remove** it: Hazel asks you to confirm, and removal is permanent. When in doubt, archive first; you can always remove later, but not the reverse.
Yes, two ways. Uploading documents brings in the past instantly: statements and tax returns carry their own dates, so history files correctly. And on a client whose profile is still new, **Rebuild profile** relearns from the household's full recorded meeting history plus every uploaded document, oldest to newest. On an empty profile there's nothing to lose, which makes it a quick way to bootstrap a long-standing client. (On an established profile, rebuild is a last resort because it erases manual edits; see the Rebuild question above.)
Anyone with access to that client, following the same rules as the rest of the client's record. There's one shared profile per client, so teammates see each other's verifications and corrections. Owners can see every client. Clients themselves never see it.
Hazel records it internally, but it isn't shown on the page today. If you need to know for a specific record, email [support@hazel.ai](mailto:support@hazel.ai) and we can check.
No. The profile reads emails your clients send you; your outgoing messages aren't captured. If you shared something important in an outgoing email, tell Hazel in chat, upload the underlying document, or mention it in your next recorded meeting.
Probably not. A joint fact is one record with two owners, shown under each owner's view. Expand it and check the Owner row: if it lists both spouses, it's one shared record. A true duplicate can be cleaned up by archiving the extra record.
### Intelligent Client Profiles Troubleshooting
The tabs appear on every Household and Contact you have access to, so if they're missing, it's almost always an access question. Confirm you can open the client at all; if you can't, or you can open them but see fewer tabs than a teammate does, ask your firm's Owner about your access to that client.
Still not working? Email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
**How to fix:**
1. Check the client's **Meetings** tab. Was a recent meeting recorded by the notetaker, and does it have notes and a transcript? Meetings without a transcript can't feed the profile.
2. **Check the client is on the meeting.** Facts are filed for a meeting's participants. Email works the same way: the address your client writes from must match the email on their contact record.
3. If a meeting just ended, give it a few minutes; facts land shortly after meeting notes are ready.
4. Nothing recent to draw from? The fastest way to build the profile right now is to upload a statement or tax return on the **Documents** tab, or tell Hazel what you know in chat.
Still not working? Email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
Hazel couldn't process the file. Try **Upload a fresh copy**; a clearer scan or an export straight from the source works best. Scanned image-only PDFs with unreadable text are the most common cause. If a clean copy still fails, email [support@hazel.ai](mailto:support@hazel.ai) with the file name so we can investigate.
Large documents take longer to read, and documents are added to a profile one at a time per household, so a batch upload will show some files **"Waiting for profile"** while earlier ones finish. If a single document has shown **Reading** or **Adding to profile** for more than a few hours, email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
Match the fix to the problem:
* **Wrong value**: click **Edit** under it (saving marks it verified).
* **Right value, wrong person**: expand the record's Owner row and click **Fix owner**.
* **An older value was right**: open **History** and **Restore** it.
* **Verified by mistake**: click **Unverify**, or just fix the value.
* **The whole record was never true**: **remove** it. Hazel asks you to confirm, and removal is permanent, so archive instead if it merely ended.
The profile shows the most recent *stated* value with its "as of" date; it doesn't pull live balances from custodians. Upload a recent statement, tell Hazel the new number in chat, restate it in your next recorded meeting, or click **Edit** and update it now.
Check the **Archive** view on the Data tab. Records that ended (archived by you, a teammate, or by Hazel when the client said something ended) move there, grouped by date. **Unarchive** brings one back. Records also disappear when a document that was their only source is deleted. Email [support@hazel.ai](mailto:support@hazel.ai) if you need to confirm what happened.
The fact and its quote are fine; Hazel just can't open the surrounding conversation. It happens when the meeting's transcript isn't available in Hazel (for example, consent wasn't given), or when the fact came from an email in a teammate's mailbox, since email context opens only from your own. Document facts link to the document instead, and manually entered values have no conversation behind them.
Chat updates apply only after you confirm the changes Hazel lays out, so check the conversation for a proposal you haven't approved yet. Beyond that: make sure the conversation is clearly about one client (name the household, or use the chat rail on the household's own page), and that you have edit access to that client, since updating from chat follows the same access rules as the Data tab. If Hazel proposed the wrong change, decline it and rephrase with specifics ("Bob's 401(k), not the IRA").
Still not working? Email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
Chat draws on the profile first, but it can also pull directly from your CRM, meetings, and emails, so a freshly corrected value and an older mention can briefly disagree. The Data tab is the reviewed record. If a chat answer keeps citing information you've corrected, email [support@hazel.ai](mailto:support@hazel.ai) with the chat and the client so we can investigate.
Usually a momentary hiccup. Retry, and refresh the page if it persists. One special case: if Hazel is adding new information to this profile at the same moment, your change may be held briefly and asked to retry. Persistent failures on one specific record are worth reporting.
Still not working? Email [support@hazel.ai](mailto:support@hazel.ai) so we can investigate.
# Debriefing a Meeting
Source: https://docs.hazel.ai/guides/meeting-assistant/debriefing-a-meeting
After a meeting, Hazel helps you turn any conversation into a clear, actionable debrief. Hazel pulls the important takeaways from the conversation based on the **Meeting Type** that was selected, capturing the desired details you need from a call. You don’t have to replay the call or build a recap from scratch. In a few minutes, you can review everything in one place and finish the meeting with clarity for your team and your client.
If a Household/Contact is not assigned, we cannot sync notes and to the proper Household/Contact in your CRM. **Click on the ‘+’ icon in the top right and manually associate a Household/Contact.**
[Please review the ‘Setting yourself up for success’ section above.](/guides/meeting-assistant/meetings-overview/set-up-for-success)
1. **Review and sync notes to the CRM** – After each meeting, Hazel generates notes for you. Review these notes and sync them to your CRM to keep it up to date. These notes also power important features like the pre-meeting agenda and chat.
*Need to refine the output? You can switch the Meeting Type or *[*edit the meeting notes prompt*](/guides/meeting-assistant/meeting-details/meeting-types/customizing-meeting-types)* right from the notes view to regenerate the notes in the format and level of detail you want.*
2. **Send meeting summary email to clients** – Hazel creates a meeting summary email for you. Simply review it, make any adjustments if needed, and send it to your clients to provide them with a clear recap.
3. **Review and sync tasks to the CRM** – Hazel generates relevant tasks based on the meeting discussion. Sync these tasks into your CRM so you can stay on top of the most important priorities moving forward.
# Performance Dashboard
Source: https://docs.hazel.ai/guides/meeting-assistant/meeting-details/performance-dashboard
The Performance dashboard gives firm owners a single view of how their advisors run client meetings. It surfaces the same per-meeting analytics shown on each meeting page — talk/listen ratio, longest monologue, longest client story, interactivity, patience, and advisor and client question rates — rolled up across every meeting in the firm.
The Performance tab is only visible to **Owners**. Advisors continue to see their own meeting analytics on each meeting page, but do not see firm-wide rollups.
### Opening the dashboard
Sign in as an Owner and select **Performance** in the left sidebar. The page loads at `/advisor/performance` and shows an empty state if your firm has no completed meetings with computed metrics yet.
### Filtering the dashboard
Two filters scope every chart and table on the page:
* **Meeting type** – Restrict to a single meeting type (for example, *Discovery* or *Annual Review*) or leave on **All meeting types**.
* **Timeframe** – **All time**, **Last 90 days**, or **Last 30 days**.
Charts and tables that compare advisors also have an **Advisors** multi-select. Open the picker, search by name, and check the advisors you want included. Clearing the selection (or leaving it empty) shows every advisor.
### What the dashboard shows
A baseline row per meeting type, showing the firm’s average for each metric across every advisor. This is the benchmark every other view is measured against.
One row per advisor with their averages across the selected scope. Use this to spot advisors whose numbers stand apart from the firm — flagged cells (●) deviate from the firm’s typical range for that meeting type by more than 30%. Hover any cell to see the firm benchmark and the percent delta.
A monthly line chart of a single metric, with one line for the firm and one per selected advisor. Use the metric selector above the chart to switch between talk/listen, longest monologue, interactivity, patience, question rates, and longest client story. The trend is bounded to the most recent 24 months.
The primary table: every (advisor, meeting type) combination with all seven metrics side by side, ordered to make outliers easy to scan. Cells outside the firm’s typical range for that meeting type are flagged.
Pick an advisor to load their most recent meetings (newest first) with the per-meeting metric values. Use this to jump from a flagged cell straight to the meeting that drove it.
### Correcting speakers to fix metrics
Performance metrics are only useful when the transcript correctly identifies who is the advisor and who is the client. If diarization labelled a speaker incorrectly — or split one person across multiple labels — an Owner or the meeting’s advisor can fix it from the meeting detail page.
Open the meeting from the **Meetings** page and switch to the **Transcript** tab.
Click **Edit speakers** above the transcript. The dialog lists every speaker the recording detected, with their talk time, utterance count, and a snippet of their first line.
Rename any speakers that need correcting. Giving two rows the same name **merges** those speakers into one — useful when a single person was split across multiple labels.
For each speaker, mark them as **Advisor** or **Client**. Speakers without a side (for example, a one-time interruption) are excluded from advisor- and client-specific metrics.
Click **Save**. Hazel updates the transcript and recomputes that meeting’s performance metrics in the background. The new values flow into the firm Performance dashboard the next time you load it.
Merged speakers must agree on side — you can’t merge an Advisor row with a Client row. Resolve the conflict before saving.
### Tips for accurate analytics
* Set advisor/client sides on every meaningful speaker. Metrics like **talk/listen ratio** and **longest advisor monologue** depend on the side assignment.
* Re-check **Edit speakers** the first time you review a new client — fixing it once is enough; the meeting’s metrics update automatically.
* If a chart or row looks empty, widen the **Timeframe** or remove the **Meeting type** filter. Metrics only appear once a meeting has been processed and has a non-zero number of utterances on each side.
# Adding Meeting Types
Source: https://docs.hazel.ai/guides/meeting-assistant/meeting-types/adding-meeting-types
1. **Navigate to Settings → Meeting Settings**
* From your Hazel dashboard, click on **Settings** and then select **Meeting Settings**.
* This is where all your existing meeting types are stored and where you can manage or create new ones.
2. **Add a New Meeting Type**
* Click the **“Add Meeting Type”** button.
* This is where you can name your new meeting type and begin customizing it to fit your workflow.
If you don’t want to start from scratch, you can duplicate the ‘General Meeting’ type prompts by clicking on the Add Meeting Type button, naming it, then clicking Save. Find the newly created meeting type, click ‘Edit’, and it’ll have default prompts ready for editing.
3. **Customize Your Prompts**
* Each meeting type can be tailored by editing its prompts. Hazel uses these prompts to generate structured outputs. You can set up prompts for:
* **Agenda** – Define what should be prepared before the meeting. For example, highlight goals, discussion topics, or referral opportunities.
* **Reminder Email** – Customize the tone and content of reminders sent to attendees before the meeting.
* **Meeting Notes** – Guide Hazel on what information to capture during the meeting (e.g., goals, action items, key decisions).
* **Follow-Up Email** – Personalize the post-meeting summary email that Hazel generates for clients or team members.
1) Head to any meeting that is ‘Ready for Review’
2) At the top of the meeting notes, click the button that says ‘Customize Prompt’
3) Edit the prompt
4) Click ‘Save and Regenerate’
1. Head to Settings / Meeting Settings
2. You can either add a new Meeting Type or edit existing Meeting Types.
4. Select the Meeting Type you would like to edit, or click ‘Add Meeting Type’ to create a new meeting type.
5. Click ‘Save and Regenerate’
If you don’t want to start from scratch, you can duplicate the ‘General Meeting’ type prompts by creating a new meeting type and only filling in the Name, then clicking save.
Need help with prompting? Visit [Customizing Meeting Types](/guides/meeting-assistant/meeting-details/meeting-types/customizing-meeting-types) to get started, or the [Prompt Library](/resources/prompt-library/prompting-tips-and-tricks) for pre-built templates!
4. **Match on Key Phrases**
* You can include keywords or phrases that Hazel will use to automatically identify and match the correct meeting type.
* For example, adding “annual review” as a key phrase ensures that when those words appear in a calendar invite, Hazel automatically categorizes the meeting as an **Annual Review**.
# Customizing Meeting Types
Source: https://docs.hazel.ai/guides/meeting-assistant/meeting-types/customizing-meeting-types
### Prompting for Meeting Types
Meeting Type prompts let you define the instructions Hazel will follow to generate the agenda, reminder email, meeting notes, and follow-up email for that meeting type.
You can keep it simple by using a **single prompt** to create an all-in-one summary, or get more specific by adding **multiple prompts** separated by three dashes (`---`). Each prompt section runs as its **own separate query** on the transcript, allowing you to control *exactly* what Hazel returns—such as a narrative recap, a Q\&A section, action items, or a financial data table.
In general, prompts work best when they’re focused on one job—one prompt for decisions, one for next steps, one for client updates—rather than asking for everything at once. Using multiple prompts keeps each request clear and targeted, which helps Hazel return more complete, organized, and accurate results.
Think of prompts the same way you’d train a paraplanner: by telling them what details matter and how you want notes structured. Editing meeting note templates or creating new meeting types isn’t required, but it helps align Hazel with your firm’s processes and ensures your notes consistently reflect what’s most important to you.
For single prompts, only one prompt is run to generate the meeting notes.
### Example Single Prompts
**Default Prompt**
Generate a detailed summary of the meeting using sections & bullet points covering all topics discussed. Use a narrative style, be detailed, and include any outcomes or conclusions made. Be thorough, but consolidate notes under a few major topics.
**Paragraph-Based Narrative Style Prompt**
Generate detailed notes of the meeting using paragraphs and a narrative style. Be detailed, and include any outcomes or conclusions made.
To run multiple prompts for detailed summaries, in the editor, simply add three dashes (`-` `-` `-`) and a divider should appear. Below it, you can add a new prompt.
Each prompt will run independently, and be added to the end of the notes generated before it. Think of each prompt as an separate instruction for Hazel.
### Example Multi Prompts
**Basic Notes, Q\&A, and a Financial Data Table Prompt**
Generate a detailed summary of the meeting using sections & bullet points covering all topics discussed. Use a narrative style, be detailed, and include any outcomes or conclusions made. Be thorough, but consolidate notes under a few major topics.
***
Add a Q\&A section which covers the most important questions asked in the meeting, and the corresponding responses. If the same question is asked multiple times, only include it once. Format the Q\&A section using bold text for the question, and regular text for the answer.
\[EXAMPLE FORMAT]
**Q: What is the current interest rate on my savings account?**
A: The current interest rate on your savings account is 0.25%.
**Q: What is the current interest rate on my checking account?**
A: The current interest rate on your savings account is 0.25%.
***
Add an html table that includes any numeric financial data discussed
Using a template in prompts can help standardize the output, while allowing AI to insert relevant context. Template-based prompts are great anytime you want **repeatable, structured output** from meetings. You define the exact sections, fields, and structure you need, and Hazel pulls the right details from the transcript to populate each part consistently, meeting after meeting.
**Example Template-based Prompt**
Generate financial advisor meeting notes using the following template. Fill in each section using the context of the meeting:
**Client Meeting Notes**
**Date:** \[Meeting Date]
**Time:** \[Start Time] – \[End Time]
**Advisor:** \[Advisor Name]
**Client(s):** \[Client Name(s)]
**Meeting Type:** \[Annual Review / Planning Session / Check-in / Onboarding / Other]
**Agenda / Objectives:**
1. \[Objective 1 – e.g., Review financial goals]
2. \[Objective 2 – e.g., Discuss investment performance]
3. \[Objective 3 – e.g., Update on tax/estate planning]
**Key Discussion Points:**
* **Financial Goals Update:** \[Notes on progress, changes, or new goals]
* **Cash Flow & Budgeting:** \[Income changes, expenses, savings rate, etc.]
* **Investments & Portfolio:** \[Performance summary, adjustments discussed, risk tolerance]
* **Tax Planning:** \[Tax strategies, upcoming deadlines, coordination with CPA]
* **Estate Planning:** \[Wills, trusts, beneficiary updates, etc.]
* **Insurance Review:** \[Coverage updates, gaps identified, recommendations]
* **Other Topics:** \[Education planning, retirement updates, life events, etc.]
**Decisions Made / Recommendations:**
* \[Recommendation 1 – e.g., Increase monthly savings to \$X]
* \[Recommendation 2 – e.g., Rebalance portfolio toward X% bonds]
* \[Decision 3 – e.g., Schedule estate planning attorney meeting]
**Action Items:**
| Task | Responsible Party | Due Date | Notes |
| :---------------------------------------- | :---------------- | :------- | :------- |
| \[e.g., Upload recent paystub to portal] | \[Client/Advisor] | \[Date] | \[Notes] |
| \[e.g., Submit Roth conversion paperwork] | \[Advisor] | \[Date] | \[Notes] |
**Next Steps / Follow-Up:**
* \[Summary of what happens next, and when]
# Default Meeting Types
Source: https://docs.hazel.ai/guides/meeting-assistant/meeting-types/default-meeting-types
Hazel is preloaded with five meeting types to help you get started. Each meeting type has its own unique prompts which capture your notes from different perspectives.
Expand the following default Meeting Types to view and copy the templates:
**Pre-meeting Notes/Agenda Prompt**
> Generate pre-meeting notes for a meeting between a financial advisor and one or more clients. Include any relevant information that may be helpful to the advisor, and a brief summary of the previous meeting, if applicable. Include any potential conversation starters.
**Pre-meeting Email Prompt**
> Using the meeting agenda and any context provided, generate a professional email to send to meeting participants. The email should:
>
> 1. Begin with a warm, friendly comment based on previous meetings context. It should be informal though personal and friendly.
> 2. Briefly state the purpose of the meeting
> 3. List the key topics to be discussed (in bullet points) by stating "Here's what we'll be covering"
> 4. Include any specific preparation required from participants
> 5. Mention any relevant documents or materials that will be reviewed
>
> Write in a professional but warm tone, and keep the email concise yet informative.
**Meeting Notes Prompt**
> Generate a detailed summary of the meeting using sections & bullet points covering all topics discussed. Use a narrative style, be detailed, and include any outcomes or conclusions made. Be thorough, but consolidate notes under a few major topics.
>
> ***
>
> Add a Q\&A section which covers the most important questions asked in the meeting, and the corresponding responses. If the same question is asked multiple times, only include it once. Format the Q\&A section using bold text for the question, and regular text for the answer.
>
> \[EXAMPLE FORMAT]
>
> **Q: What is the current interest rate on my savings account?**
>
> A: The current interest rate on your savings account is 0.25%.
>
> **Q: What is the current interest rate on my checking account?**
>
> A: The current interest rate on your savings account is 0.25%.
>
> ***
>
> Add an html table that includes any numeric financial data discussed
**Follow-up Email Prompt**
> This is a meeting in between an advisor, and one or more of their clients. Using the transcript, generate a follow up email for the client with the important topics discussed. Pay particular attention to the action items, ensuring they are included in the follow up. And pay also particular attention to numerical figures and percentages, bolding these in the email. Omit the subject, and just write the body of the email. Use a write tone that's formal, but human.
**Pre-meeting Notes/Agenda Prompt**
> Generate pre-meeting notes for a meeting between a financial advisor and one or more clients. Include any relevant information that may be helpful to the advisor, and a brief summary of the previous meeting, if applicable. Include any potential conversation starters.
**Pre-meeting Email Prompt**
> The meeting is a prospective client meeting, and is intended to give the financial advisor and a prospect space to understand if it's a mutually beneficial relationship. Using the meeting agenda and any context provided, generate a professional email to send to meeting participants. The email should:
>
> 1. Open with a warm, personable greeting and a quick reminder about the upcoming meeting (include date, time, and method of meeting).
> 2. Use a mix of bullet points where it's more beneficial to summarize, and a couple sentence paragraphs where emphasis needs to be made. Keep the overall email concise, between 50-100 words.
> 3. Briefly state the purpose of the meeting and what the prospect can expect.
> 4. Highlight the key topics we'll cover in simple bullet points (so the client sees structure and clarity).
> 5. Note any specific preparation the client should do beforehand (if applicable).
> 6. Mention any relevant documents or materials that will be reviewed.
> 7. Close with a friendly, welcoming tone that conveys excitement and professionalism (e.g., "looking forward to our conversation").
>
> Write in a professional but friendly and warm tone. Keep the email concise yet engaging.
**Meeting Notes Prompt**
> This is a prospective conversation between a financial advisor and a prospect, intended to provide a space for them to get to know each other on a high level. Generate a detailed summary of the meeting using sections & bullet points covering all topics discussed. Use a narrative style, be detailed, and include any outcomes or conclusions made. Be thorough, but consolidate notes under a few major topics.
>
> ***
>
> From the transcript, extract both quantitative and qualitative information. Highlight prospect insights that reveal motivation, values, or decision drivers. Note specific language or cues that show:
>
> **1. Key Data Points**
>
> * Client/prospect names and ages
> * Names of family members (spouse and children)
> * Important dates or timelines mentioned
> * Financial figures (account balances, income, expenses, assets, liabilities, etc.)
> * Any other quantifiable details shared
>
> **2. Client Goals & Context**
>
> * Summarize client goals, values, and concerns.
> * Capture relevant background or personal context that shapes their financial priorities
>
> **3. Motivations & Decision Drivers**
>
> * Highlight what matters most to them financially (e.g., family, retirement, growth, security, legacy)
> * Note decision-making style (analytical, emotional, relationship-driven, cautious, etc.)
> * Identify language or cues that reveal motivation, priorities, or hesitations
>
> ***
>
> Add an html table that includes any numeric financial data discussed
>
> ***
>
> Add any objections posed by the client and analyze how the advisor responded to overcome the objection. For each objection, provide a thorough response with all relevant context to the objection. Be sure to provide context on both the objection made by the client, and how the advisor responded to the objection. Provide a list based on the example format below, only provide objections and responses, no need for titles.
>
> \[EXAMPLE FORMAT]
>
> \*\*Objection: \[objection made by client]
>
> Response: \[response by advisor]
**Follow-up Email Prompt**
> This is a meeting between an advisor and a prospective client. Using the transcript of the meeting and the distilled notes, generate a follow-up email to the prospect. The email should:
>
> * Open with a warm thank-you and genuine appreciation for their time.
> * Use a mix of bullet points where it's more beneficial to summarize, and a couple sentence paragraphs where emphasis needs to be made. Keep the overall email concise, between 100-250 words.
> * Reference specific goals, motivations, or values the prospect mentioned (e.g., family security, retirement lifestyle, growth aspirations) to show understanding.
> * Address any objections, concerns, or hesitations raised in the meeting in a thoughtful and empathetic way (not defensive, but reassuring).
> * Reconnect those motivations and concerns to how we can help — highlighting solutions or next steps that align directly with their needs.
> * End with a compelling call-to-action that encourages response (e.g., scheduling the next conversation, providing documents, simply sharing their thoughts, or a reminder of the next meeting time if discussed on the call).
> * Keep the tone professional yet personable, confident yet approachable. Make it clear we're looking to build a mutually beneficial relationship.
**Pre-meeting Notes/Agenda Prompt**
> Generate pre-meeting notes for a meeting between a financial advisor and one or more clients. Include any relevant information that may be helpful to the advisor, and a brief summary of the previous meeting, if applicable. Include any potential conversation starters. Include age as of meeting and age as of end of year for the attendees listed.
**Pre-meeting Email Prompt**
> The meeting is an discovery call, and is intended to give the client space to share their personal and financial short, medium, and long term goals. Using the meeting agenda and any context provided, generate a professional, but friendly email to send to meeting participants. The email should:
>
> 1. Briefly state the purpose of the meeting
> 2. Use a mix of bullet points where it's more beneficial to summarize, and a couple sentence paragraphs where emphasis needs to be made. Keep the overall email concise, between 50-75 words.
> 3. List the key topics to be discussed. (in bullet points)
> 4. Include any specific preparation required from participants
> 5. Mention any relevant documents or materials that will be reviewed
>
> Write in a professional but friendly and warm tone. Keep the email concise yet engaging.
**Meeting Notes Prompt**
> This is an exploratory conversation between a financial advisor and their client, intended to provide a space for them to get to know each other on a deeper level. Generate a detailed summary of the meeting using sections & bullet points covering all topics discussed. Use a narrative style, be detailed, and include any outcomes or conclusions made. Be thorough, but consolidate notes under a few major topics.
>
> ***
>
> Extract specific data points, including client names, dates, financial figures (account balances, income, expenses), ages, and any other quantifiable information. Summarize qualitative information, such as client goals, values, concerns, and any relevant context. Use direct quotes where appropriate to capture the client's voice.
>
> ***
>
> Add an html table that includes the clients short, medium, and long term goals
>
> ***
>
> Add an html table that includes any numeric financial data discussed
**Follow-up Email Prompt**
> This is a meeting between an advisor and a new client. Using the transcript of the meeting and the distilled notes, generate a follow-up email to the client. The email should:
>
> * Open with a warm thank-you and genuine appreciation for their time.
> * Use a mix of bullet points where it's more beneficial to summarize, and a couple sentence paragraphs where emphasis needs to be made. Keep the overall email concise, between 50-100 words.
> * Reference specific goals, motivations, or values the prospect mentioned (e.g., family security, retirement lifestyle, growth aspirations) to show understanding.
> * Address any objections, concerns, or hesitations raised in the meeting in a thoughtful and empathetic way (not defensive, but reassuring).
> * Reconnect those motivations and concerns to how we can help — highlighting solutions or next steps that align directly with their needs.
> * End with a compelling call-to-action that encourages response (e.g., scheduling the next conversation, providing documents, simply sharing their thoughts, or a reminder of the next meeting time if discussed on the call).
> * Keep the tone professional yet personable, confident yet approachable. Make it clear we're looking to build a mutually beneficial relationship.
**Pre-meeting Notes/Agenda Prompt**
> Generate pre-meeting notes for a meeting between a financial advisor and one or more clients. Include any relevant information that may be helpful to the advisor, and a brief summary of the previous meeting, if applicable. Include any potential conversation starters.
**Pre-meeting Email Prompt**
> The meeting is an annual review meeting, and is intended to give the financial advisor and their client the opportunity to review the clients financial plan. Using the meeting agenda and any context provided, generate a professional email to send to meeting participants. The email should:
>
> 1. Open with a warm, personable greeting and a quick reminder about the upcoming meeting (include date, time, and method of meeting).
> 2. Use a mix of bullet points where it's more beneficial to summarize, and a couple sentence paragraphs where emphasis needs to be made. Keep the overall email concise, between 50-100 words.
> 3. Briefly state the purpose of the meeting and what the prospect can expect.
> 4. Highlight the key topics we'll cover in simple bullet points (so the client sees structure and clarity).
> 5. Note any specific preparation the client should do beforehand (if applicable).
> 6. Mention any relevant documents or materials that will be reviewed.
> 7. Close with a friendly, welcoming tone that conveys excitement and professionalism.
>
> Write in a professional but friendly and warm tone. Keep the email concise yet engaging.
**Meeting Notes Prompt**
> Generate a detailed summary of the meeting using sections & bullet points covering all topics discussed. Use a narrative style, be detailed, and include any outcomes or conclusions made. Be thorough, but consolidate notes under a few major topics.
>
> Exclude any financial advisor's personal updates or unrelated small talk from advisors.
>
> ***
>
> For each topic **that was discussed**, write a **couple short, but detailed paragraphs**. Organize by key financial areas (examples: Retirement Planning, Investments, Pension & Social Security, Cash Flow, Estate Planning, Tax Planning, Risk Management). **Only include** sections if they were discussed. Here's what you should and should not include:
>
> * **Include** what the client shared (concerns, goals, data)
> * **Include** what the advisor explained or suggested
> * **Include** any tentative or final conclusions
> * **Do not include** sections if they weren't discussed.
>
> ***
>
> Add an html table that includes any numeric financial data discussed. The table should have the following columns: Financial Data (e.g. label), Category (e.g. Income, Account, Expense), Data from Meeting (e.g. number). Include values like account balances, income, contribution %, retirement income needs.
>
> ***
>
> Add an html table that includes any Fiduciary / Suitability Impacting Data discussed. The table is for compliance and should have all of the relevant suitability and compliance worthy information discussed in the meeting. The table columns should be: As of \[Today's Date], Comment.
>
> * Populate if any of the following were **explicitly discussed**:
> * Demographics (age, marital status, family structure)
> * Occupation / employment status
> * Net Worth / Liquid Net Worth
> * Annual Income
> * Liquidity Needs
> * Investment Objective
> * Risk Tolerance
> * Time Horizon
> * Material changes in financial situation
> * If not mentioned, leave blank.
>
> ***
>
> Please review the meeting transcript and evaluate the following areas to help the advisor improve future meetings:
>
> * Provide one piece of feedback on how the advisor could have deepened rapport or shown more understanding. If not mentioned, leave blank.
> * Highlight any key quotes or moments where the client expressed hesitancy that the advisor disregarded. If not mentioned, leave blank.
> * Highlight one area after the client expressed positive praise that could have been a natural opening to ask for a referral. If not mentioned, leave blank.
**Follow-up Email Prompt**
> This is a meeting between an advisor and their existing client. Using the transcript of the meeting and the distilled notes, generate a follow-up email to the prospect. The email should:
>
> * Open with a warm thank-you and genuine appreciation for their time and trust.
> * Use a mix of bullet points where it's more beneficial to summarize, and a couple sentence paragraphs where emphasis needs to be made. Keep the overall email concise, between 100-250 words.
> * Reference specific goals, motivations, or values the client mentioned (e.g., family security, retirement lifestyle, growth aspirations) to show understanding.
> * Summarizes key observations from the review (progress since last year, wins, challenges).
> * Reiterates recommendations made during the call, in clear and simple language.
> * Closes on a relational note using a comment the client mentioned in the transcript. End the email on a light, friendly note
> * End with a compelling call-to-action that encourages response (e.g., scheduling the next conversation, providing documents, or simply sharing their thoughts).
> * Keep the tone professional yet personable, confident yet approachable. Make it clear we're looking to build a mutually beneficial relationship.
**Pre-meeting Notes/Agenda Prompt**
> Generate pre-meeting notes for a meeting between a financial advisor and one or more clients. Include any relevant information that may be helpful to the advisor, and a brief summary of the previous meeting, if applicable. Include any potential conversation starters.
**Pre-meeting Email Prompt**
> Using the meeting agenda and any context provided, generate a professional email to send to meeting participants. The email should:
>
> 1. Briefly state the purpose of the meeting
> 2. List the key topics to be discussed (in bullet points)
> 3. Include any specific preparation required from participants
> 4. Mention any relevant documents or materials that will be reviewed
>
> Write in a professional but warm tone, and keep the email concise yet informative.
**Meeting Notes Prompt**
> Generate a detailed summary of the meeting using sections & bullet points covering all topics discussed. Use a narrative style, be detailed, and include any outcomes or conclusions made. Be thorough, but consolidate notes under a few major topics.
>
> ***
>
> Summarize this meeting in a structured format using the EOS/Level 10 style. Include the following sections:
>
> 1. **Attendees & Roles**
> * List each participant, their role, and any relevant context.
> 2. **Team Objectives**
> * Capture the team's top priorities, quarterly objectives, or 'Rocks' mentioned.
> 3. **Metrics**
> * List key metrics, KPIs, or numbers reviewed (e.g., new accounts, AUM changes, client touchpoints, pipeline updates).
> 4. **To-Dos & Assignments**
> * Document all action items, who is responsible, and deadlines.
> 5. **Issues / IDS (Identify, Discuss, Solve)**
> * Record issues raised, discussion points, and agreed-upon solutions.
> 6. **Decisions Made**
> * Capture final decisions, commitments, or strategic shifts.
> 7. **Notes & Other Important Details**
> * Any additional context, client insights, compliance reminders, or operational updates.
**Follow-up Email Prompt**
> This is a meeting between an advisor and their team members. Using the transcript of the meeting and the distilled notes, generate a follow-up email to the prospect. The email should:
>
> * Use a mix of bullet points where it's more beneficial to summarize, and a couple sentence paragraphs where emphasis needs to be made. Keep the overall email concise, between 100-200 words.
> * Reference the most important KPIs, performance numbers, trends, or topics discussed.
> * Summarizes key observations and strategic decisions agreed upon.
> * Reiterates commitments made during the call, in clear and simple language.
> * Capture major takeaways or important context relevant to the team's objectives.
# How Hazel’s notetaker works
Source: https://docs.hazel.ai/guides/meeting-assistant/overview/how-it-works
Hazel’s notetaker supports the full meeting workflow—**prep, capture, and debrief**—so advisors can walk into every client conversation informed and walk out with clear next steps. Before the meeting, Hazel surfaces an agenda-style overview of what’s most relevant to review going in (recent activity, key details, and open items), helping you focus the conversation. During the meeting, Hazel captures the discussion in the background. After the call, Hazel produces a structured debrief with the meeting summary, decisions, action items, and follow-ups based on what matters most to you.
1. Hazel can record via the following methods:
* Sending a notetaker to your Zoom, Microsoft Teams, Google Meet meeting
* Capturing the audio through the Hazel web app or any voice recorder app
* Phone call recording
2. Assign a Household/Contact to the Meeting to pair the data to the client.
* Once you connect your CRM and Email, Hazel will auto-assign the appropriate Household/Contact.
* If a Household/Contact is not auto-assigned, it’s likely because we could not associate the email to an existing relationship in your CRM. **Click on the ‘+’ icon in the top right and manually associate a Household/Contact.**
3. Assign a Meeting Type
* You can instruct Hazel through prompting, similar to how you train a Paraplanner on what notes to capture. Simply edit the meeting note templates or create new meeting types based on the information you’d like to capture. This is not a requirement, but helps train Hazel on your firms processes.
# Set Yourself Up For Success
Source: https://docs.hazel.ai/guides/meeting-assistant/overview/set-up-for-success
Follow these tips to get to most out of Hazel's Meeting Assistant.
* **Check attendees** – This ensures the meeting is associated with the correct household, unlocking the ability for other team members to view the meeting and associating that meeting to a Household so Ask Hazel can access it. This connection is super powerful for keeping conversations organized and accessible.
* **Select the meeting type** – There are five [default meeting types](/guides/meeting-assistant/meeting-details/meeting-types/default-meeting-types) available, each designed to provide different styles of notes. Choosing the right one ensures the notes are tailored to the type of meeting you’re running.
* **Meeting Status** – You can update the meeting status from **Ready For Review, Complete, or Reset Meeting Data.**
* **Ready For Review → Complete**: Stay organized by clearly showing what’s been finished and what’s still left to review.
* **Reset Meeting Data**: Doing this will reset the meeting complete. All data will be lost.
# Preparing for a Meeting
Source: https://docs.hazel.ai/guides/meeting-assistant/preparing-for-a-meeting
It is important to assign Attendees so that other team members that are not on the calendar invite can view the meeting. If a Household/Contact is not auto-assigned, it’s likely because we could not associate the email to an existing relationship in your CRM.
**Click on the ‘+’ icon in the top right and manually associate a Household/Contact.**
[Please review the ‘Setting yourself up for success’ section above.](/guides/meeting-assistant/meetings-overview/set-up-for-success)
Once you connect your CRM and Email, Hazel will auto-assign the appropriate Household/Contact, and prepare an agenda and a reminder email.
Hazel will analyze previous meetings, tasks, emails, and CRM data to present you with information that is relevant for your upcoming meeting.
The agenda is prepared based on the **Meeting Type** that is selected for the meeting. A meeting type is a prompt template that instructs Hazel on what to prep and what she should capture based on the specific kind of client conversation. You can fully [customize a meeting type](/guides/meeting-assistant/meeting-details/meeting-types/customizing-meeting-types)’s agenda by editing the sections, prompts, and fields Hazel pulls into the pre-meeting overview, creating a consistent, repeatable prep flow that matches how your firm runs each conversation.
Hazel will also create a **Reminder Email** for you to review and send your client.
# Recording a Meeting
Source: https://docs.hazel.ai/guides/meeting-assistant/recording-a-meeting
## Hazel can record via the following methods
#### Video Meetings
Once you connect your Calendar, Hazel will be automatically scheduled to join calendar events with meeting links available. At the time of the event, Hazel will be in the waiting room waiting for you to admit it.
We support online video recording with Zoom, Microsoft Teams, and Google Meet.
***
You can verify Hazel will join your online meetings by clicking on the menu icon for the particular meeting you're curious about. If ‘Take Notes’ is on, then Hazel will join your meeting.
You can also verify by clicking on an upcoming meeting and verifying that it is ‘Scheduled’.
Navigate to the **Meetings page**, and select ‘New’ on the top right. Upon clicking it, you’ll have the option to enter the **Meeting Name**, and select whether it’s a **Virtual Meeting** or a **Manual Recording**.
* If it’s a **Virtual Meeting**, enter your meeting url link and Hazel will join your call in less than a minute.
* If it’s an **Manual Recording**, proceed and start recording directly from your computer, or upload a media file (audio and video are both accepted).
To record manually, click ‘Start Recording.’ This will open a new tab and an indicator that recording is in progress:
Once done recording, click ‘Stop Recording’. The tab should close, and upon refreshing the meeting page, you’ll see a processing indicator. Once the notes are ready, the page will refresh with the notes of the meeting.
Navigate to the upcoming meeting, and select ‘Record’ on the top right. Upon clicking it, you’ll have the option to start recording directly from your computer, or upload a media file (audio and video are both accepted). Click on `Upload Audio/Video File`
#### In Person Meetings
Navigate to the upcoming meeting, and select ‘Record’ on the top right. Upon clicking it, you’ll have the option to start recording directly from your computer, or upload a media file (audio and video are both accepted)
To record manually, click ‘Start Recording.’ This will open a new tab and an indicator that recording is in progress:
Once done recording, click ‘Stop Recording’. The tab should close, and upon refreshing the meeting page, you’ll see a processing indicator. Once the notes are ready, the page will refresh with the notes of the meeting.
1. Download the [hazel.ai](https://apps.apple.com/us/app/hazel-ai/id6752910685) app from the App Store.
2. Sign in to your Hazel account
3. Select the meeting you would like to start recording or create a new meeting
#### Phone Calls
Using your Hazel line takes a one-time setup:
1. **Turn on Record Phone Calls.** Go to **Settings → Meeting Settings** and turn on **Record Phone Calls**. Your personal Hazel number appears under the toggle after a few seconds, with a copy button. Save it to your phone's contacts as "Hazel."
2. **Add the numbers you'll call from.** Under **Allowed callers**, add every phone you might dial Hazel from: your cell, your desk phone. Up to 5.
> Add the numbers **you** call from, not your client's. Until you add at least one number, calls to your Hazel line won't record. Hazel only records calls placed from a number on this list.
Hazel can record a phone call two ways. If the call isn't already a meeting in Hazel, [call your own Hazel line](/guides/meeting-assistant/recording-a-meeting#recording-on-your-hazel-line) and merge your client in. If the meeting already exists in Hazel, use [Join my call](/guides/meeting-assistant/recording-a-meeting#recording-a-phone-call-with-join-my-call) and Hazel will call you.
Hazel gives you a personal phone number. Call it, merge your client in, and Hazel records both sides. No calendar event needed.
**Recording a call**
1. **Dial your Hazel number.** From one of your allowed numbers, call your Hazel line. Hazel answers and plays a short announcement that the call is being recorded.
2. **Merge your client into the call.** Use your phone's **Add Call** and \*\*Merge Calls \*\*controls to dial your client and merge them in. Both sides are now captured. If you're already on with your client, add Hazel to that call instead and merge.
Your client does not hear Hazel's announcement when they're merged in afterward. It plays only to the phone that dialed Hazel. Tell them the call is being recorded yourself.
3. **Hang up when you're done.** Hazel stops recording when the call ends, so there's no button to press. A new meeting appears on your **Meetings** page named for the number you called from and the time, for example "Phone call: (415) 555-0123 on Aug 3, 2026 2:15 PM". Notes and tasks are generated the same way as any other Hazel recording.
**Tip:** Phone-call meetings start with no client attached. Open the meeting and use **Add Participants** to link the client so the meeting is tied to the right household and your notes sync to your CRM.
Use this when the meeting already exists in Hazel and you want the recording attached to it. Hazel calls your phone and you merge it into the call you're already on.
1. **Open the meeting and choose Join my call.** On the meeting page, click **Record** in the top right, then choose **Join my call**.
2. **Enter the number for Hazel to call.** Usually your own cell, the phone you're on the call with. Include the country code, for example `+14155550123`.
3. **Answer, then merge.** Hazel calls you within a few seconds and rings for about 30 seconds. Answer it and tap **Merge Calls** on your phone. The dialog reads "Recording" once Hazel is capturing the conversation.
4. **Hang up when you're done.** The recording attaches to this meeting, and notes appear on it shortly after.
Hazel plays no announcement on these calls. Nobody on the line is told a recording has started unless you tell them.
You can record phone calls by recording from your computer and having your phone on speaker so both voices are captured by Hazel. To do so:
Navigate to the upcoming meeting, and select ‘Record’ on the top right. Upon clicking it, you’ll have the option to start recording directly from your computer.
To record manually, click ‘Start Recording.’ This will open a new tab and an indicator that recording is in progress.
You can then initiate your phone call and put your phone on speaker so both voices are captured by Hazel.
Once done recording, click ‘Stop Recording’. The tab should close, and upon refreshing the meeting page, you’ll see a processing indicator. Once the notes are ready, the page will refresh with the notes of the meeting.
## Disabling / Enabling Auto-scheduled Notetakers
Hazel automatically adds notetakers to new calendar events by default. If you would like to disable this option, you can do so within Settings → Meeting Settings. Be aware that disabling will cancel all your existing auto-scheduled notetakers for future meetings.
When in the Overview or Meetings view, there is an option to disable the notetaker one a one-off basis. Click on the menu icon for the meeting and toggle off ‘Take notes’.
## Downloading a Video Recording
You can download a Video Recording within the Transcript section of a call if you have ‘Save Video’ turned on in Settings. If you would like to disable this option, you can do so within Settings → Meeting Settings. To download the Video Recording, go to the Meeting you would like to download, and select the Download icon. This will open the video in a new tab. In the new tab, select Download and wait for the video to download on to your device.
# Custom Modes
Source: https://docs.hazel.ai/guides/modes/custom-modes
## What Are Custom Modes?
Custom Modes let you package your own **instructions** and **background knowledge** into a reusable Mode that Hazel follows whenever you select it in Ask Hazel. Instead of re-explaining how you want a task done every time, you define it once—then start a focused chat in that Mode as often as you like.
Custom Modes are built on the same foundation as every Mode (see the [Modes Overview](/guides/modes/overview)). What makes them *custom* is that **you** create them: you write the instructions, attach the reference files, and decide whether the Mode stays private to you or is shared with your whole firm.
Anyone in your firm can create a Custom Mode. A Mode you create is **private** to you by default—it only becomes visible to teammates when you mark it **Available firm wide**.
## Creating a Custom Mode
You create Custom Modes from the **mode selector** in Ask Hazel—not from Settings (Settings is where you manage and archive existing Modes).
1. **Open the mode selector**
On the Ask Hazel page, click on 'Modes' (it shows **Ask** by default).
2. **Choose “New Mode”**
Select **New Mode** at the bottom of the dropdown. This opens the Mode setup panel.
3. **Give it a Title**
Enter a clear, recognizable **Title** (up to 255 characters). This is the name teammates and you will see in the mode selector—e.g. *“Prospect Follow-Up Email”* or *“Quarterly Review Prep.”*
4. **Add a Short description**
Write a one-line **Short description** of what the Mode is for. It appears under the Mode’s name in Settings and helps your team pick the right Mode.
5. **Write the Detailed instructions**
In the **Detailed instructions** editor, describe exactly how Hazel should behave in this Mode—its tone, the steps to follow, the format of the output, and anything it should always (or never) do. You can use basic formatting (bold, italic, underline, bullet lists).
Not sure how to phrase it? Write a rough draft and click **Improve**—Hazel will rewrite your instructions into clearer, more structured guidance. Select a portion of the text first to improve just that part, or leave nothing selected to improve the whole thing.
6. **Add knowledge files (optional)**
Under **Add files**, click **Upload files** to attach reference material that gives the Mode more context—templates, checklists, firm policies, sample outputs, and the like. See [Adding Knowledge Files](#adding-knowledge-files) for supported formats and limits.
7. **Choose who can use it**
Leave the **Available firm wide** toggle **off** to keep the Mode private to you, or turn it **on** to share it with everyone in your firm. See [Private vs. Firm-Wide](#private-vs-firm-wide).
8. **Save**
Click **Save**. A Mode needs at least a **Title** and **Detailed instructions** before it can be saved. If you attached files, they’ll finish processing in the background—your Mode is ready to use as soon as processing completes.
## Writing Effective Instructions
Custom Modes are only as good as their instructions. A few guidelines:
* **Be specific about the output.** State the format you want—e.g. *“Respond with a short summary, then a bulleted list of action items.”*
* **Describe the process, not just the goal.** Tell Hazel the steps to take and what to check.
* **Set the tone and audience.** *“Write in a warm, plain-English tone suitable for a client.”*
* **Call out constraints.** Note anything Hazel should avoid, omit, or always include.
* **Iterate.** Run the Mode, see how it responds, then refine the instructions.
Looking for a starting point? The [Modes prompt library](/resources/prompt-library/modes) has ready-made instruction sets you can paste into a new Custom Mode.
## Adding Knowledge Files
Attaching files gives a Mode reference material it can draw on in every conversation—without you re-uploading it each time.
* **Supported formats:** PDF, DOCX, XLSX, TXT, and CSV
* **Limits:** up to **3 files** per Mode, **25 MB** per file
* **Processing:** uploaded files are processed in the background. While a file is processing, the Mode shows a *processing* state; once it’s **ready**, the knowledge is available in chat.
You can download a file by clicking its name, or remove it with the **✕** next to it.
If a file fails to process, the Mode is flagged with a warning (“Some documents failed to process. Please review this Mode.”) and can’t be selected in chat until you remove or replace the file.
## Private vs. Firm-Wide
The **Available firm wide** toggle controls who can see and use the Mode:
* **Private** (toggle off) — only you can see and use the Mode. It appears under **PRIVATE** in Settings → Modes.
* **Firm wide** (toggle on) — everyone in your firm can select and use the Mode. It appears under **FIRM** in Settings → Modes, labeled with **“Created by: \[your name]”** for your teammates.
Only the person who created a Mode can edit, archive, or restore it. Teammates can **use** a firm-wide Mode and **view** its details, but not change it.
## Using a Custom Mode in Ask Hazel
1. On the Ask Hazel page, open the mode selector.
2. Pick your Custom Mode from the list (your private Modes and any firm-wide Modes both appear here).
3. Start chatting. The conversation stays in that Mode for its entire duration.
A few things to know:
* Modes are chosen **before** you start a conversation, and a chat keeps the same Mode throughout.
* Modes **don’t remember past conversations**—each chat starts fresh, but your instructions and knowledge are always applied, so results stay consistent.
* In any Mode, Hazel can still use **all the data you’ve connected** (CRM, email, calendar, meeting notes, custodial data).
* If a Mode is still processing or has a failed file, it appears but can’t be selected until it’s ready.
## Editing, Archiving & Restoring
Manage your Modes from **Settings → Modes**, where they’re grouped under **PRIVATE** and **FIRM**.
* **Edit** — for Modes you created, click **Edit** to update the title, description, instructions, files, or sharing setting. (For firm-wide Modes created by someone else, you’ll see **View** instead—details are read-only.)
* **Archive** — toggle a Mode off to archive it. Archived Modes are removed from the mode selector but aren’t deleted: *“This Mode will be archived and no longer available in chat. You can restore it at any time from Settings.”*
* **Restore** — toggle an archived Mode back on to make it available in chat again.
You can also edit your own Modes directly from the mode selector—hover a Mode you created and click the **pencil** icon. For firm-wide Modes created by teammates, hover and click the **info** icon to view details.
## Starting From a Pre-Built Mode
Some Modes come pre-built as templates. When you open one, you’ll see *“This is a pre-built Mode. Duplicate it to your account to customize and use it.”* Click **Use this Mode** to copy it into your account, then edit the instructions and files to fit your firm.
## What Custom Modes Can — and Cannot — Do
**Custom Modes can:**
* Apply your saved instructions to every chat in that Mode
* Use your connected data and the knowledge files you attach
* Analyze and visualize data, and draft emails, documents, and summaries
* Produce downloadable files, including PDFs, DOCs, CSVs, XLSX
* Be kept private or shared across your firm
Custom Modes help you think, analyze, and create—you stay in control of what happens next.
**Custom Modes cannot:**
* Remember previous conversations (each chat starts fresh)
* Take actions on your behalf or run automations
* Send emails or messages
* Schedule meetings, reminders, or tasks
* Write back to your CRM or other systems
## Related
* [Modes Overview](/guides/modes/overview) — how Modes work and when to use them
* [Tax Planning](/guides/modes/tax-mode) — Hazel’s specialized, built-in tax Mode
* [Modes prompt library](/resources/prompt-library/modes) — ready-made instructions to paste into a Custom Mode
# Overview
Source: https://docs.hazel.ai/guides/modes/overview
## What Are Modes?
Modes are a **specialized version of Ask Hazel**, designed to help you work through complex, focused tasks with greater consistency and clarity.
Ask Hazel is our general-purpose chat experience. It uses Hazel’s AI engine to access your connected data and advanced language models so you can analyze information, explore scenarios, and get work done through conversation.
Modes build on that foundation by adding **structure and intent**. Instead of starting from a blank chat every time, a Mode lets you define how Hazel should behave for a specific type of work. By combining clear instructions and background knowledge, Modes help Hazel stay focused on what you’re trying to accomplish—without requiring you to repeat or refine prompts over and over.
## How Modes Work
Modes use the same AI engine as Ask Hazel, but with additional guidance that shapes how Hazel thinks and responds.
Each Mode is defined by:
* **Description** – what the Mode is designed to help with
* **Instructions** – how Hazel should approach problems and respond
* **Knowledge** – files, links, or documents that provide background context
When you use a Mode, Hazel:
* Has access to **all data you’ve connected** (CRM, email, calendar, meeting notes, custodial data)
* Automatically uses relevant data when needed to answer your questions
* Applies the Mode’s instructions and knowledge to stay focused and consistent
Modes do **not** remember past conversations. Each chat starts fresh, but because the instructions and knowledge are always in place, you can expect more reliable and repeatable results than a standard chat.
## What Makes Modes Different from Ask Hazel
Ask Hazel is flexible and open-ended. Modes are **purpose-built**.
Modes are designed for deeper, more advanced work and unlock capabilities that aren’t available in a regular Ask Hazel chat, including:
* Multi-step reasoning that chains analysis together in a single conversation
* Generative interfaces that present information visually, not just as text
* Code execution for advanced calculations and analysis
* Structured outputs such as reports, summaries, and client-ready materials
If Ask Hazel is your starting point, Modes are what you use when the work requires precision, structure, and depth.
## Using Modes in Ask Hazel
Modes are selected **before** you start a chat.
* When starting a conversation, you choose the Mode you want to use
* Each chat remains in the selected Mode for its entire duration
* By default, no Mode is selected—you intentionally choose one each time
* By default, every chat starts in Ask anything mode (shown as 'Ask'). Switch to Tax Planning or a Custom Mode before you begin; the selected mode applies to that whole chat.
* Modes are available to all eligible users in your firm
* Ask anything is available to everyone; Custom Modes to all eligible firm users; Tax Planning is a paid add-on (or available during a trial) — without access you'll see an
'Upgrade' badge and selecting it prompts a request rather than switching.
This ensures that each conversation begins with the right context and expectations in place.
## What Modes Can — and Cannot — Do
Modes are powerful, but they are designed to support thinking and analysis—not automation.
### Modes can:
* Analyze and visualize complex data
* Chain multiple steps of reasoning in one conversation
* Draft client communications and internal documents
* Produce downloadable files, including PDFs, DOCX, CSV, XLSX
### Modes cannot:
* Take actions on your behalf
* Send emails or messages
* Schedule meetings, reminders, or tasks
* Update or write back to CRM or other systems
* Run workflows or background automations
* Remember past conversations
Modes help you **think, analyze, and create**. You remain in control of what happens next.
## When to Use a Mode
Modes are best used when the work requires focus and structure.
You should consider using a Mode when:
* You want consistent outputs without rewriting prompts
* You want to share prompts with team members
* You’re working within a specialized domain, such as tax or investments
For quick questions or open-ended exploration, Ask Hazel is often enough. For everything else, Modes are built to go deeper.
When you’re ready, choose a Mode and start exploring.
# Tax Planning
Source: https://docs.hazel.ai/guides/modes/tax-mode
## What Is Tax Mode?
Tax Mode is a specialized Hazel Mode designed to help financial advisors analyze tax returns, understand tax outcomes, and explore planning considerations through conversation.
Modes are purpose-built experiences within Ask Hazel. While Ask Hazel is flexible and open-ended, Modes provide structure, intent, and depth for complex workflows. Tax Mode applies this approach specifically to tax planning—bringing together advanced analysis, connected data, and generative reporting in a single guided experience.
Tax Mode is built for paraplanners and advisors who understand core tax concepts and want a faster, clearer, and more flexible way to work through tax analysis without relying on manual workflows.
## What Tax Mode Helps You Accomplish
Tax Mode helps you move from raw tax documents to clear understanding and actionable insight.
With Tax Mode, you can:
* Analyze tax returns and supplemental tax documents in context
* Understand how income, deductions, and credits contribute to total tax liability
* Review marginal and effective tax rates and how income flows through brackets
* Identify observations and planning considerations based on the client’s situation
* Explore “what-if” scenarios through conversation
* Generate interactive, client-ready tax reports
* Export reports as PDFs or data as CSVs when needed
Tax Mode is designed to support **analysis, exploration, and explanation**—helping you form your own recommendations with greater confidence.
## How Tax Mode Works
Tax Mode follows a structured but flexible workflow that adapts as you engage with it.
1. **Start a conversation in Tax Mode**
From the Ask Hazel page, select Tax Mode before beginning your chat.
2. **Upload tax documents**
A tax return or supplemental tax document is required to get started. You may upload multiple documents, including large files or multiple years of returns.
3. **Analysis and clarification**
Hazel analyzes the uploaded documents and uses relevant connected data—such as CRM information, emails, calendar entries, meeting notes, and custodial data—when available.
If additional information is needed, Hazel will ask clarifying questions before proceeding.
4. **Interactive tax report generation**
Hazel produces an interactive tax report based on the documents and context provided.
5. **Ongoing exploration**
You can ask follow-up questions, request changes, explore scenarios, or export outputs—all through conversation.
You remain in control throughout the process, deciding what to review, adjust, or explore further.
## Customizing Tax Mode
You can tailor how Tax Mode behaves without leaving the Ask Hazel page.
When Tax Mode is selected, an **Edit Tax Mode settings** (pencil) icon appears in the prompt footer on the Ask Hazel landing page. Selecting it opens a side drawer where you can add **custom instructions** that Hazel will follow in every Tax Mode conversation—for example, default assumptions, preferred terminology, or firm-specific guidance.
* Custom instructions apply to all of your future Tax Mode conversations until you update or remove them
* Changes save from the drawer; you can close the drawer and return to your conversation without navigating to the settings page
* Tax Mode instructions are only available when Tax Mode access is enabled for your account
## What’s Included in the Tax Analysis
By default, Tax Mode generates a comprehensive, interactive tax report focused on clarity and decision support. The report includes:
* **Tax Summary & Key Figures**
A complete snapshot of the client’s tax situation, including total income, AGI, deductions, taxable income, total tax, filing status, marginal and effective tax rates, and summaries of dividends and capital gains.
* **Ordinary Income Bracket Analysis**
A clear view of how income is taxed across brackets and how much room remains in the current marginal bracket.
* **Capital Gains & Dividend Analysis**
Long-term capital gains and qualified dividends analyzed against applicable thresholds.
* **Medicare & IRMAA Considerations**
Medicare Part B and Part D impacts, including IRMAA thresholds, when applicable.
* **Deductions Overview**
A comparison of standard versus itemized deductions and how deductions contribute to overall tax outcomes.
* **Income Sources Summary**
A breakdown of income sources to help contextualize where taxable income originates.
* **MAGI Planning Considerations**
A structured view of how different MAGI definitions affect eligibility and phaseouts for deductions, credits, and benefits, shown as under threshold, above threshold, or in phaseout.
These sections update as you ask Hazel to refine the analysis or explore additional questions.
## Source Attribution
Each figure in the tax dashboard can be traced back to where it came from—such as a specific line on an uploaded tax return or a derived calculation.
* Hover over a value in the dashboard to see a **source attribution tooltip** showing the underlying source or how the number was derived
* Use the quote icon in the dashboard header to **show or hide source attribution** across the entire dashboard for the session
* Source attribution is available on tax dashboards generated in Tax Mode and helps you verify figures, explain numbers to clients, or audit the analysis before sharing
## Exporting the Dashboard to PDF
You can export the full tax dashboard as a print-grade PDF directly from the dashboard header.
To export:
1. Open the tax dashboard for your conversation
2. Choose **portrait** or **landscape** orientation from the orientation toggle in the header
3. Select the **Export to PDF** (download) icon
The exported PDF is paginated for print, with each major section laid out on its own page where appropriate. Sections that aren't relevant for a printed report are omitted automatically so the PDF stays focused on the figures, charts, and analysis you'd share with a client or CPA.
The filename includes the dashboard title and the export date, and the export uses the orientation currently selected in the header.
## Observations and Scenario Exploration
Tax Mode proactively surfaces **observations** based on the tax return and available context. These observations highlight areas you may want to explore further—they are not recommendations.
Examples include:
* Bracket management and Roth conversion considerations
* IRMAA thresholds and Medicare-related planning
* Charitable strategies such as donor-advised funds or qualified charitable distributions
* Education credits and related phaseouts
* Social Security and income timing considerations
If you choose to continue, you can explore scenarios directly in chat. Scenario exploration is advisor-driven:
* You decide what assumptions or inputs to change
* Hazel recalculates the analysis based on your request
* Visuals and figures update when you instruct Hazel to update them
Scenarios are exploratory by design and help you understand tradeoffs and implications before taking next steps.
## Common Use Cases
Advisors commonly use Tax Mode to:
* Prepare for annual client tax reviews
* Quickly understand a client’s tax situation before a meeting
* Identify tax planning considerations worth discussing
* Explore the impact of potential changes through scenario analysis
* Collaborate more effectively with CPAs and tax professionals
* Explain complex tax concepts clearly using visual summaries
## What Tax Mode Can — and Cannot — Do
**Tax Mode can:**
* Analyze tax returns and supplemental tax documents
* Generate interactive tax reports and summaries
* Surface observations and planning considerations
* Support scenario-based exploration
* Produce PDF reports
**Tax Mode cannot:**
* File or prepare tax returns
* Submit documents to the IRS or state agencies
* Provide tax advice directly to end clients
* Replace a CPA or tax professional
* Take actions on your behalf
Tax Mode is designed to help you analyze and understand tax information—not to execute or automate outcomes.
## Important Things to Know
* Tax Mode works best when each conversation focuses on **one client or household**
* Multiple documents and multiple tax years are supported, with emphasis on the most recent return and forward-looking analysis
* Hazel uses connected data sources when available to add context and depth
* Each Tax Mode conversation starts fresh
(For more details, see the Modes Overview)
## Availability and Pricing
More details on our [Pricing page](https://hazel.ai/pricing).
Tax Planning will be offered as an add-on. Hazel Admin AI + Tax Planning is priced at:
* \*\*Pay monthly: \*\*\$150 per user, per month
* \*\*Pay annually: \*\*\$125 per user, per month
*Tax Planning is for informational purposes only, and is not tax advice.*
# Billing
Source: https://docs.hazel.ai/guides/settings/billing
Admin AI is the base seat and the Tax Planning add-on is billed per seat on top of the base seat. For more details on pricing, visit our [Pricing page](https://hazel.ai/pricing).
## Managing Subscriptions
Users with the ‘Owner’ or 'Admin' role can manage their firms Billing details and update credit card details within **Settings → Subscrip**. Here’s a breakdown on everything you need to manage your Hazel subscription:
### Add a Credit Card
To add a credit card, go to **Settings → Billing**, then enter your credit card details or connect your [**Link**](https://link.com/) account. Once you click save, your subscription will be added and set to a **Monthly** billing cadence.
### Add new members to Subscriptions
To add new subscriptions, simply go to the members page and add the desired members. Billing will be updated based on the role assigned to each active user.
### Unsubscribe or Removing existing subscriptions
To unsubscribe manage billing, go to Settings > Subscriptions.
Only ‘Owners’ and 'Admins' have the ability to manage existing members roles and permissions.
### Setting the Billing Schedule
To adjust your billing schedule, select between our Monthly and Yearly options. Pricing will adjust accordingly.
Users save two months by paying annually instead of monthly.
# Integrations
Source: https://docs.hazel.ai/guides/settings/integration-settings
## Managing Integrations
Connect Hazel with your CRM, email, calendar and more. If you didn’t connect during the onboarding flow or would like to check on your connections, simply head to **Settings → Integrations** to manage your integrations.
## Integration Status
Once you go through the permissions and set up your integrations, there will be a green dot indicating you are **Connected.**
You can disconnect your integrations by clicking on the menu icon, then **Disconnect.**
You can refresh your CRM connection by clicking on **Trigger Sync.**
# Meeting Settings
Source: https://docs.hazel.ai/guides/settings/meeting-settings
## Managing Save Videos and Transcripts
You can configure meeting recordings and transcripts in Settings / Meeting Settings. Videos and transcripts are saved by default. If Save Videos or Save Transcripts are toggled off, then updates will be applied to all future meetings.
## Auto-Schedule Notetakers
Notetakers are automatically added to new calendar events by default. You can disable Auto-schedule Notetakers by toggling off ‘Auto-schedule Notetakers’ in Settings / Meeting Settings. Disabling this will cancel all your existing auto-scheduled notetakers for future meetings.
## Updating the Notetaker’s Name
You can configure the name of the notetaker joining your meetings in Settings / Meeting Settings / Bot Name. By default, the notetaker will be names ‘Hazel Notetaker’. Any changes made will be applied to all future meetings.
## Adjusting the Scheduled Join Time
You can configure how many minutes before the meeting the bot should by adjusting in Settings / Meeting Settings / Bot Join Minutes Ahead. By default, the notetaker will join your meetings 1 minute ahead of the scheduled meeting time. You can adjust this as early as 10 minutes before the scheduled meeting time.
## Meeting visibility
Meeting visibility controls **who at your firm can see a recorded meeting** — its transcript, notes, tasks, and summary. You set a firm-wide default here, and you (or the meeting's owner) can change it on any individual meeting.
### The four options
From most private to most open:
| Option | Who can see the meeting |
| :----------------- | :------------------------------------------------------------------------------------------------------------------------------------- |
| **Private** | Only you (the meeting's owner). *Your firm's Owners can always see it, too.* |
| **Attendees only** | You and the teammates who attended the meeting. |
| **Contact access** | Anyone at your firm who can already see this client — i.e., teammates with access to the meeting's household or contact. **(Default)** |
| **Everyone** | Every user at your firm. |
**Firm Owners always have access** to every meeting, regardless of the setting. So "Private" really means *only you and your firm's Owners*.
### Set your firm's default
In **Meeting Settings**, choose the visibility that should apply to **every new meeting**. New meetings use this default automatically — most firms leave it on **Contact access** so meetings are visible to the same teammates who already work with that client. You can change the default at any time; it applies to meetings going forward.
### Change visibility for a single meeting
Open any meeting and use the **visibility selector** in the header to switch it to any of the four options — or set it back to your firm's default. The change takes effect immediately.
Only the meeting's **owner**, an **advisor who attended** the meeting, or a **firm Owner** can change a meeting's visibility. Everyone else sees the current setting as read-only.
Notes:
* The option names and the "Private (firm owners can access)" caveat match the exact in-app wording, so the doc lines up with what users see.
* I led the table from most-private to most-open and flagged Contact access as the default (it's the firm default in the product).
* If your firm has meeting visibility turned off, the selector doesn't appear — say the word if you want a one-line note covering that edge case, though it's probably more detail than a help-center reader needs.
# Managing Teams
Source: https://docs.hazel.ai/guides/settings/members/managing-teams
The **Team** tab in Settings is where you manage everyone in your firm: who's on the team, what each person can do, and—most importantly—**which clients each person can see**.
There are two separate ideas to keep in mind:
* **Roles** decide what someone can *do* in Hazel (and whether they can manage firm settings). More on [Roles and Permissions](/guides/settings/members/roles-and-permissions).
* **Access** decides which **Households** and **Contacts** someone can *see*. Access is controlled per client—by sharing records with a person directly, or with a **Group** they belong to.
**Who can manage the Team tab?** Only **Owners** and **Firm Admins** see the Team tab. A few actions—assigning the Owner role and creating automatic rules—are limited to **Owners**.
## How the Team tab is organized
Open **Settings → Team**. Across the top you'll find four views:
* **Members** — everyone in your firm (and anyone you've invited)
* **Groups** — reusable sets of people you can share clients with all at once
* **Households** — browse every household and manage who can see each one
* **Standalone contacts** — same, for contacts who aren't part of a household
## Roles
Every member has one role:
| Role | What it's for |
| :---------- | :---------------------------------------------------------------------------------------------------------------------------------------- |
| **Owner** | Runs the firm in Hazel. Can see **every** household and contact, manage all settings, and manage the team. Owners are always Firm Admins. |
| **Advisor** | The standard seat. Works with the clients shared with them and uses Hazel's full feature set. |
**Firm Admin** is a separate permission you can give an Advisor or Assistant so they can manage the team and billing—*without* giving them access to any additional client data. Owners are always Firm Admins.
Giving someone access to settings (Firm Admin) and giving them access to clients are two different things. Making someone a Firm Admin never changes which clients they can see.
## Managing members
### Invite a teammate
1. Go to **Settings → Team → Members**.
2. Click **Invite** (top-right).
3. Enter their **Email address**—or pick someone under **Suggested from your CRM** to fill it in.
4. Choose a **Role**: **Advisor** (default) or **Owner**.
5. Click **Send invite**.
The person appears in the list as **Invited** until they accept.
You can invite people as **Advisor** or **Owner**. Invited members get access to all CRM-imported data (contacts, notes, tasks), so set their client access after they join.
### Change a member's role or settings
1. On the **Members** view, click a person's row to open their panel.
2. The panel opens on a hub showing the member's avatar, role, status, **Member settings**, and a **Permissions** summary.
3. Under **Member settings**, change their **Role**, toggle **Tax Planning**, or toggle **Firm Admin**.
4. Click **Save**, then confirm in **Apply changes?**.
If a role needs a paid seat and none is free, you'll see a note that a paid seat will be billed before you confirm.
**Member settings save immediately.** Permissions changes are different — they're staged and don't apply until you press **Review & apply** (see [Share clients with one member](#share-clients-with-one-member)).
### Disable a member
To offboard someone, open their panel and, under **Account**, click **Disable account**, then **Save** and confirm.
Disabling a member is **permanent and cannot be undone.** It turns off their access and disconnects their integrations. To bring someone back, you'd have to invite them again. You can't disable your own account.
## Controlling who can see which clients
Access is set **per client**. Each time you share a household or contact, you choose a level:
| Level | What it allows |
| :--------- | :------------------------------------------------------------ |
| **Read** | Can view the client's details only. |
| **Write** | Can view and edit details. |
| **Manage** | Can view, edit, **and** control who else can see this client. |
There are three ways to share a client:
* **Directly with a person** — from a member's panel, or from the client's own **Who has access** section.
* **With a Group** — share once with a Group, and everyone in it gets access (see [Groups](#groups)).
* **When you create the client** — choose who can see it right in the new-household or new-contact form.
**Owners always see everything.** You never need to share a client with an Owner.
When you open the **Who has access** panel from a household or contact's detail page, the controls only let you make changes if you have **Manage** on that record. If you can see the client but don't manage it, the panel is read-only — you can review who has access without risk of changing it.
### Share clients with one member
The member panel opens on a **hub** with three rows under **Permissions**: **Households**, **Standalone contacts**, and **Groups**. Each row summarizes how much the member can reach today—e.g. *"42 total · 30 direct, 12 via groups"* for a resource type, or *"3 groups · 18 contacts, 27 households"* for groups. Click a row to drill in.
1. **Settings → Team → Members**, click the person's row.
2. Under **Permissions**, open **Households** or **Standalone contacts**.
3. Use the **filter chips** above the list to focus by source: **All**, **Manual** (added by hand), **CRM-synced** (granted by your CRM), or **Via group** (inherited from a group). Each chip shows its count.
4. To **add** clients, click **Add households** / **Add contacts**, search, and check the clients to share. Set the **Access level for new grants** (defaults to **Read**) — changing it after you've checked rows updates everything you've selected so far.
5. To change an existing share, use the level dropdown on a row (**Read**, **Write**, **Manage**, or **Remove access**).
6. Click **← Back** to return to the hub. Repeat for the other resource type or for **Groups** if needed.
7. When you're done, click **Review & apply** in the footer to see every staged change in one list, then **Apply**.
Rows granted **via group** are read-only here — they show the group name and aren't editable from the member's panel. Edit the group's access instead (see [Share clients with a group](#share-clients-with-a-group)).
Nothing is saved as you click—every change you make across Households, Standalone contacts, and Groups is staged into one set until you press **Review & apply**. The footer shows the running count (e.g. *"5 pending changes"*). If you close the panel with unsaved changes, Hazel asks before discarding them.
#### Bulk actions (Owners only)
Inside **Add households** or **Add contacts**, Owners see three bulk actions at the top of the flow. They're scoped to the resource type you're in (households *or* standalone contacts), and like everything else they stage into the pending set — nothing applies until **Review & apply**.
| Action | What it does |
| :-------------------- | :-------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Grant all access** | Stages a grant for **every** household (or contact) in the firm at the current **Access level for new grants**. |
| **Copy access from…** | Pick another member; stages grants for everything they can access of this resource type. The member's existing access is kept; only additions are staged. |
| **Revoke all access** | Stages removal of every direct grant the member has for this resource type. Group-granted access isn't touched. |
### Set access when creating a client
When you create a household or contact, use the **Who can access this household? / Who can access this contact?** section to add the people and groups who should see it. Owners are always included automatically. You can fine-tune levels afterward.
## Groups
A **Group** is a named set of people. Share a household or contact with a Group, and **everyone in that Group** can see it—so you set up access once instead of person-by-person. Groups are perfect for service teams, pods, or "everyone who works with retirement clients."
### Create a group
1. **Settings → Team → Groups**.
2. Click **New Group**, enter a **Group name**, and click **Create**.
### Add people to a group
1. On the **Groups** view, click the **+** in the group's **Members** column.
2. In **Manage members**, search for people and click **Add**.
3. Remove someone with the **✕** next to their name, then click **Done**.
Everyone you add immediately gains access to whatever the Group can see.
### Share clients with a group
The **Household access** and **Contact access** windows are split into three tabs that share a single pending set:
* **Add households** / **Add contacts** — pick clients by hand.
* **Automatic rules** — set rules that share matching clients (and future ones) for you. See [Automatic rules](#automatic-rules).
* **Has access** — review who the group can currently see, with a preview of what changes after you apply.
The **Has access** tab shows a count badge that previews the projected roster when you have pending changes — e.g. *"42 → 47"* means the group reaches 42 clients today and will reach 47 after you apply.
1. On the **Groups** view, click the **+** in the group's **Household** or **Contact** column.
2. On the **Add households** / **Add contacts** tab, search and check the clients to share (use **Select all shown** to add a whole list).
3. Optionally switch to **Automatic rules** to add a rule, or **Has access** to preview the result. The footer shows a running tally of pending contact and rule changes.
4. Click **Review & apply**, then **Apply**.
In a Group's **Contact access** window you can only add **standalone contacts**. Contacts who belong to a household automatically follow that household's access.
### Rename or delete a group
Use the **⋯** menu on a group's row to **Rename** or **Delete** it.
Deleting a group removes the access its members got **through that group**. People keep anything that was shared with them directly or through another group.
## Automatic rules
Instead of adding clients to a Group one by one, you can set up an **automatic rule** that shares matching clients with the Group for you—both clients that match today **and** any that match in the future.
A rule matches on:
* **Owner** — the advisor a client is assigned to, and/or
* **Tag** — clients carrying a tag you choose.
### Create a rule
1. Open a group's **Household access** or **Contact access** window.
2. Go to the **Automatic rules** tab and click **Create rule**.
3. Under **Where**, add a condition: choose an **Owner** and/or one or more **Tags**.
4. Check the live preview—e.g. *"24 contacts match today · future contacts that match will be added automatically."* Click **View list** on the preview to see the exact clients that will be matched before you stage the rule.
5. Click **Add rule**. The rule appears in the list with a **Pending** badge until you apply.
6. Click **Review & apply**, then **Apply**.
You can edit a pending or active rule from the **Automatic rules** tab — pencil icon to re-open the builder, trash icon to stage a deletion (the deletion's impact count is shown alongside it).
**How rules combine:**
* Within a single rule, a client must match **every** condition (for example, owned by *Jane* **and** tagged *Tax*).
* Across rules, a client that matches **any** rule gets access.
* Leaving a rule with no conditions shares **all** clients of that type.
### See who a group can reach
Open the **Has access** tab in any access window to review everyone the Group can currently see, split into **Added manually** and **Via rule:** sections (one section per rule, headed by the rule's sentence)—a quick way to audit access.
When you have pending changes, the tab previews the post-apply state: rows you've added show a **Will add** badge, rows you've removed show **Will remove**, and pending rules show *"N will gain access"* for the contacts they'll match. Use the `×` on a manual row to stage a removal; rule rows are display-only (the rule would re-add them on apply — edit or delete the rule instead).
A few things to know about rules:
* If you remove a client by hand but a rule still matches them, they **keep** access through the rule. To stop sharing, edit or delete the rule.
* Clients shared by a rule can't be removed one at a time on the **Has access** tab—change the rule instead.
* Deleting a rule removes the access that rule granted.
## Tips
* **Start with Groups.** For anything more than one or two people, share with a Group instead of individuals—it's far easier to maintain.
* **Use rules for "always" sharing.** If a whole category of clients should always be visible to a team (by owner or tag), a rule keeps it current automatically as clients are added.
* **Remember the two axes.** Role = what someone can do; access = which clients they can see. Promoting someone to Firm Admin doesn't reveal any new clients.
* **Owners see everything**—you don't need to share clients with them.
## Related
* [Roles and Permissions](/guides/settings/members/roles-and-permissions) — a closer look at what each role can do
* [Owner / Admin Capabilities](/guides/settings/members/owner-capabilties) — managing the team and firm settings
# Owner / Admin Capabilities
Source: https://docs.hazel.ai/guides/settings/members/owner-capabilties
### Inviting Team Members
Hazel does not import existing roles and permissions from external data sources. Please review this before adding team members. [Learn more about editing household permissions.](/guides/settings/members/roles-and-permissions)
Hazel makes it easy to bring your team onboard. To invite a new member:
1. Go to Settings → Members.
2. Enter the team member’s email address.
3. Select the appropriate role (Owner or Advisor).
4. Click Invite Member.
Your teammate will receive an email with instructions to join Hazel. Once they accept, they’ll have access according to the role you selected.
Only Owners can invite new members and manage roles.
### Editing Member Roles and Status
You can easily update roles, adjust member status, or manage pending invites from the Members table.
To make changes:
1. Go to Settings → Members.
2. Find the team member you’d like to update.
3. Use the action menu to:
* Change Role (Owner or Advisor)
* Set status to Active or Inactive
* Revoke Invite if the member has not yet joined
Changes take effect immediately and update the member’s permissions based on their role.
Only Owners can edit roles, manage status, or revoke invites.
### Editing Household Permissions
Hazel gives Owners control over who can see specific Households and their related client information.
To edit Household permissions:
Go to Contacts.
Select the Household you’d like to update.
Click the Edit icon next to the Household name.
Scroll down to the Household Permissions section.
Choose one of the following visibility options:
* All – Visible to all users in the firm
* Advisors – Visible only to Owners and Advisors
* Private – Visible only to Owners
# Roles and Permissions
Source: https://docs.hazel.ai/guides/settings/members/roles-and-permissions
Hazel supports three distinct roles within a firm: **Owner, Admin, Advisor.** Each role determines which firm-admin controls and features a member can use. Access to specific client records (Households & Contacts) is controlled separately, per record, by sharing — not by role — except Owners see every record." Add to the Advisor (and new Assistant) sections that they only see records shared with them.
## Role Comparison
| Role | Admin Access (Firm Settings) | Access to All Members Meetings | Record Meetings | Full Feature Access | Chat Access | Use Case |
| :---------- | :--------------------------- | :----------------------------- | :-------------- | :------------------ | :---------- | :--------------------------------- |
| **Owner** | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | ✅ Yes | Manage firm + full Hazel access |
| **Admin** | ✅ Yes | ❌ No - based on permissions | ✅ Yes | ✅ Yes | ✅ Yes | Manage firm + advisor level access |
| **Advisor** | ❌ No | ❌ No - based on permissions | ✅ Yes | ✅ Yes | ✅ Yes | Use Hazel for client management |
## Owner
Hazel does not import existing roles and permissions from external data sources. It is up to the Owner to edit household permissions for the organization. [Learn more about editing household permissions.](/guides/settings/members/owner-capabilties)
The **Owner** seat is designed to manage the Hazel account and control firm-wide permissions, with full visibility.
* **Admin capabilities:** Invite and remove members, manage firm settings, and control who can see which Households and Contacts.
* **Feature access:** Full access to all Hazel features, including meeting recording. The owner role has access to view all meetings associated to the firm account.
* **Use case:** Owners can use Hazel’s chat to instantly access any client data, prepare and debrief meetings, record meetings, send emails, and sync notes and tasks back to the CRM — while also administering the team’s overall Hazel setup.
## Admin
The **Admin** seat is designed to manage the Hazel account and control firm-wide permissions, with limited visibility.
* **Admin capabilities:** Invite and remove members, manage firm settings, and control who can see which Households and Contacts.
* **Feature access:** Full access to all Hazel features, including meeting recording. The admin role has access to view meetings they have permission to view within the firm account.
* **Use case:** Admins can use Hazel’s chat to instantly access any client data, prepare and debrief meetings, record meetings, send emails, and sync notes and tasks back to the CRM — while also administering the team’s overall Hazel setup.
## Advisor
The **Advisor** seat gives individual advisors full access to Hazel’s productivity tools without administrative controls.
* **Admin capabilities:** None — advisors cannot manage members or firm settings.
* **Feature access:** Full access to Hazel features, including recording meetings.
* **Use case:** Advisors can record meetings, prepare and debrief meetings, access client data in seconds using Hazel’s chat, send client emails directly, and ensure notes and follow-up tasks sync back to the CRM.
# Welcome to Hazel Documentation
Source: https://docs.hazel.ai/index
# Hazel Help Center
Everything you need to get started, master features, and get the most out of your AI assistant for wealth advisors.
### Quick Start
Learn what Hazel can do
Get set up in minutes
Invite your team members
Privacy and data protection
# Release Notes
Source: https://docs.hazel.ai/release-notes
## June 2026
### New features
#### Team management: Groups and automatic rules
Manage who on your team can see which clients without sharing one record at a time. Create **Groups** and share households or contacts with the whole Group at once, then add **automatic rules** to share clients that match by **Owner** or **Tag**—including any future clients that match. See [Managing Teams](/guides/settings/members/managing-teams).
#### Hazel Android app
Record your in-person client meetings right from your phone and sync them automatically to Hazel. The Hazel app is available on iOS, and now on Android too.
[Download from the Google Play Store](https://play.google.com/store/apps/details?id=ai.hazel\&hl=en_US)
#### Get insights on every client meeting
See how you ran the meeting, not just what was said. Hazel shows you your talk-to-listen ratio, the number of questions asked, and the sentiment of the conversation. Use it to give every client a better experience.
ghts on every client meeting
## May 2026
### New features
#### File generation and polished reports available in chat
We added file generation to Modes and Ask Hazel, so Hazel can produce downloadable files—like PDF reports—right from a Mode or a chat. See [Custom Modes](/guides/modes/custom-modes). Ask Hazel for a report—a tax analysis, a plan summary—and get back a clean, consistently formatted PDF with a uniform layout, charts, and disclaimers, ready to download right from the chat. See [Ask Hazel](/guides/ask-hazel-chat/overview).
#### See where every tax number comes from
Tax Mode dashboard cards now have a toggle that reveals the source and derivation of each figure—brackets, deductions, MAGI, Medicare, and more—so you can show clients exactly how a value was calculated. See [Tax Planning](/guides/modes/tax-mode).
#### Improvements to in-person meetings
Improvements to how the audio is captured, separated by speaker, and flows into the same notes, tasks, and analysis as your virtual meetings. See [Recording a meeting](/guides/meeting-assistant/recording-a-meeting).
#### Improvements to meeting agendas and pre-meeting prep
Hazel's meeting agendas are now more powerful—pulling in relevant context about the household, including account details from your custodian—and you can run pre-meeting prep with one click. See [Preparing for a meeting](/guides/meeting-assistant/preparing-for-a-meeting).
### Minor improvements
* **iPhone photos accepted.** HEIC and HEIF images convert automatically on upload, so you can attach iPhone photos without a format error.
* **Tasks filters that stick.** The Tasks view now remembers your filters between visits and lets you apply several at once.
* **More reliable meeting recording.** Fewer duplicate or missed notetaker bots, and steadier post-meeting processing.
## April 2026
### New features
#### HubSpot integration
Connect HubSpot to Hazel for a bidirectional sync of contacts, households, tasks, and notes. Updates flow instantly from Hazel to HubSpot, and HubSpot syncs back every 60 minutes—including a full historical backfill on first connect. See [HubSpot](/guides/integrations/crms/hub-spot).
#### Redesigned Meeting Details
A reworked meeting page brings the summary, agenda, and full transcript into one feed, with attendees up top, a cleaner tasks panel, and a one-click button to sync notes to your CRM. See [Recording a meeting](/guides/meeting-assistant/recording-a-meeting).
#### Rich, interactive answers in chat
Hazel can now answer with formatted, interactive components—tables, bar/line/pie charts, metric cards, and callouts—so the numbers and comparisons you ask for come back ready to read at a glance. See [Ask Hazel](/guides/ask-hazel-chat/overview).
#### Sync a tax analysis to Salesforce
When Hazel produces a tax analysis, push it into Salesforce with a single button—Hazel adds a note on the household and attaches the analysis PDF under the household's Files. Requires a connected Salesforce CRM. See [Tax Planning](/guides/modes/tax-mode).
#### Compliance archiving with Hadrius
Connect Hadrius from Settings to automatically archive your recorded meetings and transcripts for compliance, delivered after each meeting is analyzed. See [Integration settings](/guides/settings/integration-settings).
### Minor improvements
* **Sharper meeting notes and transcripts.** Upgraded language and speech models make summaries read more naturally and in-person transcripts more accurate.
* **Image documents supported in Tax Mode.** Drop in a photo or scan of a tax document and Hazel reads its figures into your analysis.
* **Filter meetings by household.** The meetings list now has a household filter and shows participants as a column.
* **Faster email lists.** Email views load noticeably faster and flag which messages have attachments.
## March 2026
### Modes
Create custom Modes that teach Hazel how you work. Set up your instructions, templates, and preferences once — then have them one click away in any new chat. Build a **Firm Coach Mode** so advisors can tap into firm best practices instantly or an **Estate Planning Specialist** **Mode** that produces a summary, asset distribution by scenario, and recommendations the moment you upload an estate plan. See the [overview](https://docs.hazel.ai/guides/modes/overview) or visit our [Modes prompt library](https://docs.hazel.ai/resources/prompt-library/modes) to see how other advisors are putting Modes to work.
### **Tasks**
You can now create, edit, and delete tasks from a dedicated Task view, or ask Hazel to create one for you in chat. For example: "Hey Hazel, create a high priority task due tomorrow to complete the ACAT for The Smith Family." This view is great for advisors that see a surge of clients per day or for firm managers that want a way to see what's been assigned and what's still pending for their team at a glance.
### **Tax Mode and Chat reliability improvements**
In Tax Mode, federal income tax brackets now include an improved visualization for ordinary income.
We've also made chat more resilient. Hazel now displays a clear indicator when a conversation is getting too long, and automatically retries if our infrastructure is unavailable.
# Ask Hazel
Source: https://docs.hazel.ai/resources/prompt-library/ask-hazel
### Client Engagement prompts
* Help me prepare for my upcoming meeting with `Client Name`
* I'm on the phone with `Client Name`. What are the 3 most relevant action items?
* Provide a recap of my interactions with the `Household Name` this year.
* Draft an email for `Client Name` explaining the current bond environment and why it may make sense to extend bond duration.
* Draft a reassuring market-volatility email personalized to `Client Name`’s plan and time horizon.
### Financial Planning prompts
* Summarize the `Client Name`’s top 3 stated goals and the gaps we still need to close.
* I want to see what we have accomplished with `Client Name`. Where are we at with the 5 areas of financial planning? (estate, tax, insurance, investments, cash flow).
* Let's come up with some financial planning recommendations for `Client Name`
* Provide me with an estate summary `[Attach Document]`
* Explain Medicare IRMAA Requirements
* What are the current interest rates and when is the Federal Reserve meeting again?
* Provide me with a social security analysis for `Client Name` `[Attach Document]`
* Explain ‘x’ (e.g., duration, sequence risk, Roth conversions) in plain English I can say out loud.
* Build a retirement readiness summary for `Client Name`: timeline, income sources, gaps, next actions.
### Custodial Data prompts
* List the beneficiaries for the `Client Name` household.
* What are `Client Name` 's accounts and balances?
* What changed since our last review for `Household Name` (accounts, cash, contributions, withdrawals, beneficiary changes)?
* Based on my last meeting with the Johnson, should I rebalance their portfolio?
* Build a performance summary for the `Client Name` household.
* Review `Client Name` 's unrealized gain/losses and propose a way to generate \$25,000 cash in the most tax neutral manner.
### Prospecting and Discovery prompts
* Give me the 6 highest-impact discovery questions for a prospect who is `[retiring in 3–5 yrs / business owner / HNW / tech equity]`. Include what each answer *unlocks* (strategy + next step)
* Prospect said: ‘`[insert objection]`'. Give me 3 responses: (1) empathetic (2) data-driven (3) story-based. End each with a soft next step.
* Based on my recent meeting with `Prospect Name`, what were my 2 biggest weaknesses (missed questions, weak framing, or objection handling) I should work, and what were my 2 biggest strengths I should keep doing?
* Given my recent meeting with `Prospect Name`, list the top 5 proactive moves we can evaluate in the first 30 days, and what data we need to confirm.
### **Tax Planning prompts**
* For `Client Name`, build a three-scenario plan: Base case (no changes), optimize for current tax bracket, and aggressively optimize (cross-bracket moves if rational)
* For `Client Name`, tell me the total RMD required for tax year YYYY, how much they have taken YTD, how much is still remaining, and how they’ve satisfied their RMD in the past (for example, cash distributions to a bank account vs. QCDs, including typical timing, charities used if applicable, and any withholding or delivery preferences)
* For `Client Name` and `Spouse Name`, provide a Social Security analysis (62, FRA, 70) that also optimizes Roth conversions, and summarize the plan as a year-by-year timeline with what we should do each year up until RMD age of 73? \[Attach Social Security Document]
* For `Client Name`, summarize realized gains and losses YTD, and identify opportunities to make the outcome tax neutral through loss harvesting, lot selection, and gain timing, while flagging wash sale risks \[Attach Investment Statement]
# Financial Planning
Source: https://docs.hazel.ai/resources/prompt-library/financial-planning
A client will see this plan, so the wording you send is most of the work.
## Where each of these goes
Send the requests under **Starting a plan** in **Ask Hazel**, with the segmented control under the message box set to **Financial Planning**. A chat left on **Ask** answers you but never builds a plan.
Every later request changes a plan you already have, so it belongs in that plan's conversation: press **Edit** in the top bar in the chat view, or **Open chat** on the household's **Financial Planning** tab. Both open the same conversation, and each request comes back as a new version with the earlier one still readable.
## Starting a plan
### The full brief
Fill in what applies, delete the rest.
```text theme={null}
Build a financial plan for [household].
Treat them as [persona name, if you want one].
The question this plan has to answer: [the one real question].
Goals: [target retirement year, purchases, education, giving, exit].
Constraints: [what is off the table].
Emphasize: [the topics that matter for this family].
Leave out: [what does not apply].
```
Leave the persona line out and Hazel matches the household to the closest Client persona your firm has, then names it in the **Persona** row of **Applied settings** in the plan's **Appendix**.
### Short briefs
```text Pre-retirement review theme={null}
Build a plan for [household], treat them as pre-retirees. The question
is whether [name] can retire at [age] without cutting current spending.
Model Social Security starting at [age] for each of them. Emphasize
income sequencing and the healthcare gap before Medicare. Audience: a
review meeting with a couple who want the trade-offs.
```
```text New client, first meeting theme={null}
Build a plan for [household], new clients. Keep it to the picture and
two decisions. Emphasize where they stand today and the biggest
risk they face.
```
```text Retiree income plan theme={null}
Build a plan for [household], treat them as retirees. The question is
whether their spending of [amount] a year holds. Lead with distribution
ordering, then required distributions.
```
```text Business owner pre-exit theme={null}
Build a plan for [household]. The business is their largest asset and
an exit is [timeframe] out. Emphasize liquidity, concentration, and
the pre-exit tax picture. Constraint: they will not sell earlier.
```
## Naming a different persona
```text theme={null}
Build a plan for [household] using our [persona] persona, not the
[other persona] one. [One sentence on why.]
```
```text theme={null}
Rebuild this against our [persona] persona. This household is really
[what is actually going on].
```
## Levers
```text theme={null}
Add a retirement-age lever from [low] to [high] in one-year steps,
starting at [current], and show its effect on the success gauge.
```
```text theme={null}
Add a savings lever to show what another [amount] a month buys.
```
```text theme={null}
Add a spending lever around their stated need of [amount], to show
what a [percentage] cut does.
```
```text theme={null}
Add a claiming-age lever for each of them, 62 through 70.
```
Moving a lever recomputes the figures on screen and saves nothing. To keep a setting, approve the recommendation the lever sits under, or say the values in the chat:
```text theme={null}
Rebuild the plan with retirement at [age] and spending at [amount].
```
If a line under the lever panel says the projections do not respond to those controls yet, that lever illustrates the trade-off. Ask in the chat for the numbers at a setting.
## Visuals
Say what you want the visual to prove and let Hazel pick the form. To point at one already in the plan, name the heading above it, or a name from the [visual reference](/guides/financial-planning/the-plan/visual-reference).
```text theme={null}
Show the allocation as a mix comparison, current versus target.
```
```text theme={null}
Replace the [chart] with the range of outcomes. They should see the
spread, not a single line.
```
```text theme={null}
Turn the [schedule] into a table, with a column for [what matters].
```
```text theme={null}
Add a before-and-after on the key figures, so the recommendations
show an effect.
```
```text theme={null}
Show where each next dollar of savings should go, in order.
```
## Structure
```text theme={null}
Add a [topic] tab to [section], covering [what it should cover].
```
```text theme={null}
Drop the [section]. [Why it does not apply.]
```
```text theme={null}
Move [section] ahead of [section]. They have seen the balance sheet
a hundred times.
```
```text theme={null}
Give the [section] a short narrative before the tables. It opens too
cold.
```
If Hazel declines a structure change, your firm's strictness is probably **Exact**, which a firm owner can loosen: [Strictness](/guides/financial-planning/settings/personas/strictness).
## Deeper analysis
```text theme={null}
Compare [option A] against [option B] side by side, with a verdict.
```
```text theme={null}
Stress-test the plan against two down years right at retirement.
```
```text theme={null}
Show the Roth conversion schedule through age [age], with the bracket
each year fills and the tax it costs.
```
## Tightening before a meeting
```text theme={null}
Cut the summary to three paragraphs and lead with whether they can
retire at [age].
```
```text theme={null}
Tighten the whole plan. It reads long. No section longer than a page.
```
```text theme={null}
Explain every figure in one sentence before showing it. No jargon
without a gloss.
```
## Checking a plan
```text theme={null}
Where does the [figure] in [section] come from, and what math
produced it?
```
```text theme={null}
Which figures here rest on facts nobody has confirmed?
```
```text theme={null}
Is there a section on [topic]? If not, where did that topic end up?
```
```text theme={null}
Which persona did you build this from, and what did it change from
our firm defaults?
```
```text theme={null}
What data were you missing that would have changed this plan?
```
Run the last one before every rebuild: it names the document to upload next.
Assumptions do not change through chat. Select the figure one sits behind, press **Edit** in the **Provenance and calculations** panel, **Queue change**, then **Recalculate** in the bar at the bottom of the plan. See [Solver assumptions](/guides/financial-planning/assumptions/solver-assumptions).
## Where to go next
* [Front-loading the brief](/guides/financial-planning/your-data/front-loading-the-brief)
* [Changing a plan with chat](/guides/financial-planning/the-plan/changing-a-plan-with-chat)
* [Levers and scenarios](/guides/financial-planning/the-plan/levers-and-scenarios)
* [Visual reference](/guides/financial-planning/the-plan/visual-reference)
# Agenda
Source: https://docs.hazel.ai/resources/prompt-library/meeting-types/agenda
Have an Agenda prompt you'd like to share with the community? Email [support@hazel.ai](mailto:support@hazel.ai) so we can upload it here.
# Follow-up Email
Source: https://docs.hazel.ai/resources/prompt-library/meeting-types/follow-up-email
> You are an expert financial communication assistant. Your task is to generate a client follow-up email for a financial advisor based on provided meeting notes.
>
> **Key Instructions:**
>
> 1. *Begin with a Personal, Conversational Greeting:* Start the email with a two-line subject-style intro using the following format:
> * "\[Client First Name & Co-Client First Name] – good to see you."
> * Or: "\[Client First Name & Co-Client First Name] – great to catch up with you."
> Optionally include a personalized comment tied to the meeting (e.g., a congratulations, or reference to a new job or life update).
> * After the greeting, and within the same paragraph line, include a short introductory sentence such as: "Below are notes from our meeting with **to-dos highlighted:**"
> * Please note, nothing should be bold besides "to-dos highlighted:"
> 2. *Organize Content into Clear, Topical Sections:* Segment the email into distinct financial planning topics based on the meeting notes. Each section should follow this format:
> * **Section subject-style header** (e.g., "Taxes:")
> * **Bullet point style with the content for each section head** (e.g., "You filed jointly for 2023."). Use sub-bullets within the content if necessary to add additional detailed content that is pertinent to a bullet point.
> * Write in short, clear, and conversational sentences.
> * All text in the email prompt should be the same font size.
> * Include only valuable, relevant information from the meeting. Avoid trivial points.
> 3. *Highlight Action Items in Bold:* Any action items or next steps for the client should be clearly marked with bold formatting:
> * *Bold the entire sentence or phrase that describes the action item.* Example: "**Send us a copy of your finalized estate documents.**"
> 4. *Use Specific Names and Details:* Refer to each client by name as needed and be as specific as possible about the context (e.g., names of children, account types, roles, specific financial goals).
> 5. *Keep the Tone Warm, Professional, and Clear:* Write in a tone that's professional yet conversational and easy to understand. Avoid overly technical language unless you provide a clear, simple explanation. Never use financial jargon without context or clarity.
> 6. *Close with a Meeting Reminder and a Friendly Sign-Off:* At the end, confirm any upcoming meeting (if applicable).
> * If a next meeting was scheduled on the call, please also include a sentence like: "Our next meeting is set for MM/DD at 00:00am/pm."
> * Close with a warm, friendly phrase like: "Talk soon," or "As always, please don't hesitate to reach out if anything comes up."
> * Sign the email with \[financial advisor name].
***
> You are an expert financial communication assistant. Your task is to generate a follow-up email to a financial advisor's client(s) based on provided meeting notes.
>
> **Key Instructions:**
>
> 1. **Begin with a Professional and Personalized Greeting:** Start the email with "Hi" followed by the client's first name and a comma (e.g., "Hi \[Client First Name],"). Immediately follow this with a sentence expressing gratitude for their time and acknowledging the productivity of the meeting.
> 2. **Provide a Clear Summary of Key Discussion Points:** Following the greeting, present a concise summary of the main topics discussed. Use bullet points to list these key takeaways for easy readability. Focus on explaining what was discussed and, where relevant, the reasoning behind it. Exclude trivial or insignificant topics. Focus on takeaways that provide genuine value and context for the client.
> 3. **Clearly Identify and List Action Items:** Create a distinct section dedicated to any actions the client(s) need to take. Label this section clearly (e.g., "Your action Items:", "Next Steps:"). Present these actions as a bulleted list, ensuring each item is a clear and actionable instruction directly related to the meeting's content.
> 4. **Address Specific Individuals or Situations Directly:** If the meeting involved multiple clients or specific individual circumstances were discussed, use names and specific details to tailor the information appropriately within the email.
> 5. **Include Forward-Looking Considerations (If Applicable):** If the discussion included topics that are not immediate action items but are important for future consideration, include a separate section for these. Frame these as points for the client to keep in mind or that will be revisited later.
> 6. **Maintain a Professional yet Approachable Tone:** Write in a formal and respectful manner, avoiding overly casual language. However, ensure the tone remains helpful, empathetic, and easy for the client to understand. Avoid financial jargon or explain it clearly if necessary.
> 7. **Focus on Essential Information:** Only include information directly relevant to the meeting and the follow-up. Avoid unnecessary pleasantries beyond the opening and closing, and do not repeat information unnecessarily.
> 8. **Structure for Clarity:** Organize the email logically with clear paragraphs and bullet points to enhance readability and ensure the client can easily identify key information and action items. Do not use bold or italic fonts in any of the text.
***
Have an Follow-up Email prompt you'd like to share with the community? Email [support@hazel.ai](mailto:support@hazel.ai) so we can upload it here.
# Meeting Notes
Source: https://docs.hazel.ai/resources/prompt-library/meeting-types/meeting-notes
> Generate a detailed client summary of the meeting using sections & bullet points covering all topics discussed. THIS IS FOR THE CLIENT AND SHOULD BE INFORMAL, lack jargon, and be concise. Include their aspirations, goals, and recommendations of future financial decisions and recommendations needed for them to achieve their goals. Make this in bullet point form under the following topics:
>
> **What are your goals?**
>
> \[paragraph here]
>
> **Where are we today?**
>
> \[paragraph here]
>
> **What the the top next steps?**
>
> \[paragraph here]
>
> **Implementations**
>
> \[paragraph here]
***
> Add any objections posed by the client and analyze how the advisor responded to overcome the objection. For each objection, provide a thorough response with all relevant context to the objection. Be sure to provide context on both the objection made by the client, and how the advisor responded to the objection. Provide a list based on the example format below, only provide objections and responses, no need for titles.
>
> \[EXAMPLE FORMAT]
>
> **Objection: \[objection made by client]**
>
> Response: \[response by advisor]
***
Have a Meeting Notes prompt you'd like to share with the community? Email [support@hazel.ai](mailto:support@hazel.ai) so we can upload it here.
# Reminder Email
Source: https://docs.hazel.ai/resources/prompt-library/meeting-types/reminder-email
> You are an expert financial communication assistant. Your task is to generate a client follow-up email for a financial advisor based on provided meeting notes.
>
> Using the meeting agenda and any context provided, generate a professional email to send to meeting participants. The email should:
>
> 1. Briefly state the purpose of the meeting
> 2. List the key topics to be discussed (in bullet points)
> 3. Include any specific preparation required from participants
> 4. Mention any relevant documents or materials that will be reviewed
>
> Write in a professional but warm tone, and keep the email concise yet informative.
>
> The email is being sent by a supporting staff member, not the advisor. Generate a professional meeting reminder to a client from supporting staff on behalf of financial advisor. The reminder should include embedded links to the meeting, reschedule, and cancel links if they are available. Maintain a warm but professional tone, avoid first-person plural pronouns, and refer to financial advisor and the clients as if you were a supporting staff member writing this email.
***
Have an Reminder Email prompt you'd like to share with the community? Email [support@hazel.ai](mailto:support@hazel.ai) so we can upload it here.
# Modes
Source: https://docs.hazel.ai/resources/prompt-library/modes
**Browse ready-made prompts** for Hazel Modes. Each prompt below is a set of instructions you can copy and paste directly into a new Mode. Find one that fits your workflow, click the **Prompt** tab to see the full instructions behind it, and paste them into Hazel — no customization required. Whether you need help drafting client emails, analyzing portfolios, or preparing meeting recaps, these prompts give Hazel the structure and focus to deliver consistent, repeatable results right away.
***
#### Financial Planning Checklist
> You are in **Client Profile Checklist** mode.
>
> Your task is to produce a single, continuous client profile checklist that an advisor can use as both a living reference document and a meeting preparation tool. The checklist must be organized into the sections listed below, in this exact order.
>
> For each item, mark it with a status: ✅ Confirmed (data exists and has been reviewed), ⚠️ In Progress (partially addressed or needs follow-up), ❌ Not Yet Addressed (no data found or never discussed), or ➖ Not Applicable (does not apply to this client's situation — state why briefly).
>
> After each status marker, include a short plain-language note summarizing what you found in the data, what was discussed, or what remains outstanding. If a date of discussion or review exists in the source data, include it. Do not fabricate dates, dollar amounts, or details that are not present in the source material — instead mark the item as needing follow-up and specify what the advisor should collect.
>
> Present this in a table format, where each secion header is its own table. Each table should have 4 columns: Description, Status, Date Gathered, Notes. Be concise but use complete, meaningful phrases — every line should make sense to an advisor reading it cold without context.
>
> **CLIENT IDENTITY AND HOUSEHOLD**
>
> * Full legal name(s) of client and spouse or partner
> * Date(s) of birth and planning age(s) used for projections
> * Current home address and state of residency for tax planning purposes
> * Citizenship and residency confirmation — confirm whether client or spouse has lived in or holds citizenship in any other country, as this affects estate, tax, and insurance planning
> * Marital status and date of marriage or domestic partnership if applicable
> * Dependent children — names, ages, and any special needs or guardianship considerations
> * Other dependents — aging parents, adult children, or others the client financially supports
> * Employment status for each household member — W-2 employee, 1099 contractor, business owner, retired, or not currently employed
> * Employer name(s) and general job stability outlook as noted in discussions
> **GOALS AND STATEMENT OF FINANCIAL PURPOSE**
> * Client's statement of financial purpose — what is important about money to them, in their own words as captured in notes
> * What the client said they would regret not accomplishing if discussed
> * What the client would prioritize if money were not a barrier
> * Retirement vision — what retirement looks like to the client, including desired age, lifestyle, location, and activity level as described
> * Debt freedom target — target date and strategy for becoming debt-free if discussed
> * Home purchase or relocation plans — primary residence purchase, sale, or move with approximate timeline
> * Vacation or second home plans — desired location, timeline, and estimated budget if available
> * Family flexibility goals — desire to have children, adopt, or expand family and any financial preparation discussed
> * Time with family — what this means to the client specifically and how it connects to financial decisions
> * Travel goals — near-term travel plans for the current year and longer-term travel aspirations
> * Parental support — desire or obligation to help parents financially, including caregiving considerations
> * Living environment goals — desire to control where and how they live, urban vs. rural, rental vs. ownership, proximity to family
> * Education funding goals — helping children or grandchildren with education, including amounts, timeline, and vehicles discussed
> * Charitable goals — what being charitable means to the client, desired donation levels, organizations, or time commitments discussed
> * Business ownership goals — interest in starting, growing, buying, or selling a business with any details discussed
> * Family protection goals — what it means to the client to make sure their family is taken care of in the event of an emergency, disability, or death, defined in their own words
>
> **SAVINGS AND ACCUMULATION**
>
> * Emergency fund status — current balance, target amount, and where it is held
> * Bucket funding status — progress toward funding designated savings buckets for specific goals, noting any shortfalls
> * Retirement account contributions — current year contributions to 401(k), 403(b), 457, IRA, and Roth IRA including whether maximizing employer match and catch-up contributions if eligible
> * Brokerage account savings — current taxable brokerage contributions and purpose of these funds
> * Annuity or pension status — any existing annuity contracts or pension entitlements with current values and projected income if available
> * Health savings — HSA or FSA contributions, current balances, and whether being used as a savings vehicle or spending account
> * College funding — 529 plan contributions, current balances, beneficiaries, and whether funding is on track for projected costs
> * Stock option or equity compensation review — any unvested options, RSUs, or ESPP participation with vesting schedules and exercise strategy if discussed
> * Large planned expenses — home purchases or renovations, auto purchases, weddings, family events, or other major outlays with estimated timing and amounts
>
> **DEBT MANAGEMENT**
>
> * Total debt-to-income ratio as most recently calculated
> * Debt reduction strategy — whether using snowball method, avalanche method, or another approach
> * Additional debt payments — any extra principal payments being made and to which obligations
> * Student loan status — balances, interest rates, repayment plan type, and whether pursuing public service loan forgiveness or other forgiveness programs with projected forgiveness date if applicable
> * Mortgage status — current balance, interest rate, term remaining, and whether refinancing has been reviewed or pursued
> * Co-signed loan review — confirmation that advisor has reviewed whether client has any co-signed obligations and associated risks
> * Credit card balances — current balances, interest rates, and payoff plan if carrying balances
> * Other liabilities — auto loans, personal loans, lines of credit, business debt, or any other obligations with balances and terms
>
> \
> **RISK MANAGEMENT AND INSURANCE**
>
> * Disability insurance — current coverage details, elimination period, benefit period, and whether self-insuring is appropriate given assets and income
> * Life insurance — current policies in force, coverage amounts, policy types, carriers, and whether coverage is adequate for family protection goals stated above
> * Health insurance — current plan type, annual out-of-pocket exposure, and whether FSA or HSA is being utilized optimally
> * Healthcare cost projection — whether long-term healthcare and caregiving costs have been estimated using a planning tool and discussed with the client
> * Care discussion and expression of wishes — whether the client has articulated their preferences for care in aging or incapacity scenarios
> * Home, auto, and umbrella insurance — whether policies have been reviewed for adequate coverage, deductibles, and liability limits
> * Whether insurance recommendations from prior reviews have been implemented
> * Periodic financial decision-making assessment — whether the firm's scheduled assessment for cognitive and decision-making capacity has been sent or completed per the firm's review cycle
> * Digital asset management — whether the client has documented access credentials and instructions for digital accounts, wallets, and online assets including cryptocurrency if applicable
>
> **TAX PLANNING**
>
> * State-specific credits or exemptions — whether applicable state credits have been applied for and implemented
> * Tax estimate and projection status — whether current-year tax projections have been run and reviewed
> * Savings maximization — whether all available tax-advantaged savings opportunities are being utilized
> * Income optimization — whether income structure has been reviewed for tax efficiency, including business income strategies if self-employed or business owner
> * Refund optimization — whether withholdings are calibrated to avoid large refunds or underpayment penalties
> * State-specific exemptions — whether the client is taking advantage of any available state tax exemptions relevant to their situation
> * Required Minimum Distribution planning — whether RMDs are applicable, current status, and whether Qualified Charitable Distributions are being used if charitable goals exist
> * Investment location optimization — whether tax-efficient asset location has been reviewed across account types with attention to dividend and capital gains placement
> * Withholding updates — whether W-4 has been reviewed and updated for current-year accuracy
> * Social Security tax planning — whether taxation of benefits has been projected and integrated into the broader plan
> * Pension tax planning — whether pension income withholdings and tax treatment have been reviewed
> * College planning tax considerations — whether 529 contributions, financial aid implications, and FAFSA strategies have been reviewed if applicable
>
> **ESTATE PLANNING**
>
> * Beneficiary designation review — whether beneficiaries on all accounts, policies, and plans have been reviewed for accuracy and alignment with the overall estate plan
> * Totten trust or POD/TOD designations — whether payable-on-death or transfer-on-death designations have been reviewed on bank and brokerage accounts
> * Will — current status, date last updated, and whether it reflects the client's current wishes
> * Guardianship designations — whether guardians have been named for minor children in the will
> * Durable Power of Attorney for Healthcare — whether executed, who is named, and date of document
> * Financial Power of Attorney — whether executed, whether durable or non-durable, who is named, and date of document
> * Living Will — whether executed and date of document
> * Health Care Proxy — whether executed, who is named as agent, and date of document
> * HIPAA Authorization — whether executed to allow designated individuals to access medical information
> * Five Wishes or equivalent advance directive — whether the client has documented their preferences for decision-maker designation, medical treatment preferences, comfort measures, how they want to be treated by caregivers, and what they want loved ones to know
> * Organ donor registration — whether the client has registered their preference
> * Pre-paid or pre-planned final arrangements — whether funeral or memorial services have been pre-planned or pre-funded
> * Trusted contact designation — whether a trusted contact person has been named on financial accounts per regulatory requirements
> * Digital wallet and cryptocurrency access — whether access instructions for any digital wallets or cryptocurrency holdings have been documented and stored securely, including use of a password manager if applicable
>
> **INVESTMENTS**
>
> * Portfolio allocation review — whether current allocations have been reviewed against target allocations for each goal
> * Fee review — whether investment fees including expense ratios, advisory fees, and any transaction costs have been reviewed and discussed
> * Holdings and allocation review — whether individual holdings have been reviewed for concentration risk, overlap, and alignment with the investment policy
> * Tax diversification review — whether the mix of pre-tax, Roth, and taxable accounts has been evaluated for tax flexibility in retirement
> * Portfolio longevity review — whether projections have been run to assess whether the portfolio can sustain planned withdrawals over the client's expected lifetime
> * Implementation status — whether the client has implemented the most recent investment recommendations
>
> **RETIREMENT READINESS**
>
> * Pension quote updates — whether the most recent pension benefit estimates have been obtained and incorporated into projections if applicable
> * Social Security planning updates — whether benefit estimates have been updated and claiming strategy has been reviewed, or if already receiving, whether this line item should be retired from the checklist
> * Retirement income projection — whether a comprehensive projection of all income sources in retirement has been run and reviewed with the client
>
> **ACTION ITEMS AND FOLLOW-UPS**
>
> * List every outstanding action item identified during this review, who owns it (advisor, client, attorney, CPA, insurance agent, or other professional), and the target completion date if one was discussed
> * List every missing input or data gap that prevents a complete assessment in any section above
>
> At the end of the checklist, provide exactly five **Best Next Steps** as bullet points — the five highest-priority actions the advisor should take or initiate before the next client meeting, drawn directly from the gaps and opportunities identified in the checklist above. Anchor each step to a specific section and item.
***
#### Financial Planning Statement
> You are a **financial planning analyst** writing a structured summary of a client’s financial planning software output and related discussion notes. Write in a professional, client-friendly tone (plain English, minimal jargon). Be prescriptive, personable, and respectful. Target 900-1,000 words (\~2–3 pages single-spaced). Use clear headings and short paragraphs. Use bullet points only where requested.\
> \
> You will be provided the \[Household Name], primary decision maker: \[head of household, first name], and secondary decision maker: \[spouse, first name] to aid in your summary. Use the first names and third person pronouns for both the primary decision maker and secondary decision maker. \
> \
> Break down your writing into 3 different sections with paragraphs: executive summary, financial planning, recommendations. Include assumptions or details needed to make this a well rounded briefing. Each of those sections will contain the following: Executive summary - two paragraph briefing on your findings. Make it concise and focus on the overall scope of the findings. Financial planning - provide two paragraphs on each of the areas of financial planning (estate, tax, investments, risk management, and cash flow). Each area should include current state of affairs, any conversations discussed, feelings or observations, and recommendations made with detailed reseasoning. Be prescriptive. Recommendations - consolidated list with each recommendation provided using one sentence. Provide the list in a table format with 2 columns: 1st will include the area of financial planning (estate, tax, investments, risk management, and cash flow), 2nd will include the recommendation stated in one sentence.
***
#### Estate Document Summary
> You are in **Estate Planning Specialist** mode.
>
> **Task:** Given the client scenario below, produce the final output in table format only, with exactly three tables in this order: 1) Estate Summary table 2) Distribution of Assets by Scenario table 3) Recommendations table (a list of opportunities to improve the estate plan).
>
> **Rules:** Use the firm's Estate Planning Process Guide for engagement workflow (phases, deliverables, timelines). Use the firm's Approved Strategies Menu for strategy recommendations and tiering. Determine strategy tier based on estate size and label it clearly.
>
> If any **federal** tax thresholds are mentioned or implied (estate tax exemption, annual gift exclusion, GST exemption, AFR/7520/other IRS rates), you MUST **verify current-year values via web search before stating numbers**. If the guide's value differs, explicitly note: "Our guide references \[X] but the current \[YEAR] value is \[Y]."
>
> If strategies are **Tier 3**, explicitly state: **"Senior Advisor co-review required per firm policy."** If a **non-citizen spouse** is involved, flag **QDOT** requirement per the guide.
>
> Avoid generic estate planning advice; anchor recommendations to the firm process + guide terminology. Keep recommendations actionable and specific.
>
> **Output Requirements:** Provide **no prose outside tables** besides the titles for the tables and a list of Best Next Steps (5 bullet points). Use concise, advisor-usable phrasing. In the Distribution of Assets by Scenario table, show how assets pass under different scenarios based on the order of death or incapacity of relevant individuals (e.g., Spouse 1 dies first, Spouse 2 dies first, simultaneous death, single client death). Each scenario should be its own row or group of rows showing who receives what and through which mechanism (will, trust, beneficiary designation, intestacy, etc.). If a required detail is missing (e.g., asset breakdown), include a row in Estate Summary labeled "Missing inputs" and list what's needed.
>
> **Table Schemas (use these columns):** Estate Summary table columns: Item | Details. Distribution of Assets by Scenario table columns: Scenario | Asset or Account | Passes To | Mechanism | Notes / Planning Implications. Recommendations table columns: Opportunity | What to do (actionable) | Who owns it | Notes / Escalations.
# Prompting Tips & Tricks
Source: https://docs.hazel.ai/resources/prompt-library/prompting-tips-and-tricks
Great results start with great prompts. The clearer your request, the smarter your AI assistant can be. We’ve gathered some quick tips and tricks for writing prompts that get right to what you’re trying to accomplish:
**Start with the end in mind.**
❌ Instead of: “Here’s an example...” within the prompt sections
✅ If you already have examples you’ve used before, drop them into Hazel’s chat. Hazel will look at your example, work backwards, and suggest a prompt that clearly tells it what you want. Take that prompt and drop it into the appropriate prompt section for better results.
**Provide concrete examples: Offer Hazel existing examples of what you want to see, illustrating ideal outcomes.**
❌ Instead of: “Write a reminder email with bullet points of topics we’ll cover.”
✅ Try: “Write a reminder email for an upcoming meeting. We’ll be discussing one or multiple, of the following topics: Investments, Tax, Estate, Insurance, Cash Flow. Based on outstanding client tasks, advisor tasks, and previous meeting notes, tailor the reminder email to address this/these topics to address in the upcoming meeting. For example, if the main focus is 'Investments,' the reminder should highlight any relevant investment-related items. Here’s an example of the tone and structure I want to follow: …’”
**Include relevant context: Help Hazel grasp the bigger picture to ensure responses match your broader goals.**
❌ Instead of: “Summarize the notes and provide next steps”
✅ Try: Ex. Discovery meeting: “This was a meeting where we discussed goals and observations of future financial decisions and recommendations. The next step in the process is to take the findings from this conversation and provide recommendations needed to achieve their goals. Generate a detailed summary of the discovery meeting using sections & bullet points covering all topics discussed. Include next steps based on the most relevant topics that will be addressed on the next meeting.”
**Define a clear role: Specify roles to ensure the AI’s tone, depth, and accuracy align with your expectations.**
❌ Instead of: “Avoid jargon and always use the clients name.”
✅ Try: “Key instructions:
1. **Use Specific Names and Details:** Refer to each client by name as needed and be as specific as possible about the context (e.g., names of children, account types, roles, specific financial goals).
2. **Keep the Tone Warm, Professional, and Clear:** Write in a tone that’s professional yet conversational and easy to understand.Avoid overly technical language unless you provide a clear, simple explanation. Never use financial jargon without context or clarity.
3. ………”
# Podcasts
Source: https://docs.hazel.ai/resources/webinars-and-podcasts/podcasts
January 9, 2026
In this episode of the Only Human Podcast, host Yohance Harrison is joined by the Hazel team to explore the evolution of Hazel, its integration with Altruist, and the future of AI-driven wealth management.
The discussion covers the birth and growth of Hazel, the importance of security and privacy in AI tools for advisors, and how technology can help financial professionals stay human while scaling their impact. The episode also features the signature "Marry, Divorce, Date" segment, offering candid insights into the fintech landscape for advisors.
November 26, 2025
Advisors spend nearly half their time on tasks that take them away from clients. Hazel was built to change that.
In this episode of The Advisor Journey, Dasarte sits down with Altruist’s Gokul Ramanathan and Fernando San Martín to share how Hazel, Altruist’s new AI-powered assistant, is helping advisors work smarter and focus on what matters most.
They discuss Hazel’s origins, its role in simplifying advisor workflows, and the standards that keep client data secure. From automating meeting prep to identifying follow-up opportunities, Hazel is designed to give advisors time back while improving the client experience.
Advisors will learn how this technology evolved from a startup idea into a core Altruist product—and why the future of advice should feel more efficient, personal, and human.
September 17, 2025
In this episode of Streamline My Practice, hose Dave Zoller is joined by the Hazel team to explore how Hazel is working on making advisors irreplaceable.
Hazel is making human advisors more valuable by amplifying they do best.\
\
Instead of replacing relationship-building, it enhances it:\\
* Remembers every client detail so you can focus on connection\\
* Prepares personalized talking points before every meeting\\
* Coaches you to improve after every interaction\\
* Automatically tracks and communicates your ongoing value
# Webinars
Source: https://docs.hazel.ai/resources/webinars-and-podcasts/webinars
September 1, 2026
In part two of our webinar series on financial planning, we're unveiling a complete demo of Hazel's Financial Planning Agent for the first time.
You’ll learn how to:
* Build plans automatically with Hazel instead of assembling them yourself
* Run financial planning scenarios and watch Hazel dynamically reshape plans in real time
* Get results you can trust, because every number in the plan maps back to the meeting, email, document, and Intelligent Client Profile behind it
You stay in control of every recommendation. Hazel does the heavy lifting.
July 28, 2026
A financial plan is only as good as what you know about the client. But your planning software holds only a fraction of the full picture, so every plan starts a step behind.
In session one of our financial planning series, we'll show how you can use Hazel to:
* See every detail organized to the individual client and household, filed where it belongs
* Trace any fact back to its source: Ask Hazel to find the meeting it came from, the exact words that were said, and the date
* Keep your judgment in the loop, so uncertain details wait for your review
Hazel surfaces the deeper nuance — the details buried in your emails, notes, files and meetings — and organizes it with your CRM to create a single client profile that automatically keeps itself up to date.
April 30, 2026
What's new in Hazel? Modes, modes, modes.
Modes are a set of instructions you can provide Hazel to define the type of work you would like her to complete.
In this session, the Hazel team dives deep into what is possible with Hazel's Tax Planning agent.
February 24, 2026
Hazel AI Tax Planning reads 1040s, paystubs, statements, meeting notes, email, custodial, and CRM data to help advisors create personalized tax strategies and client-ready plans.
In this session, Jordan Haines from Drucker Wealth joins the Hazel team to talk through the modes he's created to help his team do more with Hazel.
November 18, 2025
In this session, you’ll see how Hazel connects to the Altruist custodial platform — a first for the industry — unlocking several powerful use cases, including the ability to:
* Pull information from Altruist on contacts, households, accounts, balances, and beneficiaries
* Compare meeting notes to household custodial data to confirm whether a portfolio is properly allocated, among other things
* Include a portfolio overview as part of meeting agendas and notes
* Use custodial information to draft email responses to client questions (coming soon)
You’ll also hear from Yohance Harrison, BFA, founder of Money Script Wealth, who shared how his firm uses Hazel to save over 75 hours per month by automating prep, follow-up, and communication.
This webinar also highlights Hazel’s commitment to secure, private AI, including our Zero Data Retention (ZDR) agreements with LLM providers — ensuring your data is never stored or used to train models.
Plus, get a sneak peek at Hazel’s roadmap for the rest of 2025 and into early 2026.
October 2, 2025
Transformative AI, Built Just for Wealth Managers.
In this session, Fernando guides us through how wealth managers can use AI to transform their firms and client relationships.
Rewatch for insights on:
* How to evaluate and choose the right AI platform for your firm
* Understanding Cybersecurity and AI
* How Hazel’s tools and tech are built different
You’ll also get a sneak peek into future Hazel plans, plus highlights from the live Q\&A featuring real advisor questions.