ChatGPT Prompts for Financial Advisers: Follow-Ups and Review Prep

Financial advisers can use ChatGPT for admin and communication: follow-up email drafts, review agendas, plain-English explanations of processes and meeting-prep checklists, built from placeholders rather than client data. Never paste identifying details or ask it for recommendations, and add a human compliance check before anything is sent.

Most prompt lists for advisers skip the thing that matters most to a regulated adviser, which is what you must not put in. Reverse the order and the prompts get safer: confidentiality rules first, prompts written to work within them, and a closing line on each draft that sends it through your normal review.

Follow me on Instagram@sagnikteaches

The rules the prompts are built on

Before any prompt, agree these with whoever owns compliance in your firm. If you are a sole practitioner, you are that person, so write them down anyway.

Connect on LinkedInSagnik Bhattacharya

1. Admin and communication only. Drafting, structuring, summarising your own text, checking tone. Not recommendations, suitability reasoning, product selection or tax and legal conclusions.
2. No client data in the prompt. No names, contact details, account or policy numbers, balances, holdings, dates of birth, health or family details. Use placeholders: [Client], [Spouse], [Portfolio value], [Review date].
3. A person reads everything. The adviser who sends it owns every word. AI-drafted text goes through the same review and record-keeping as any other client communication.
4. Check the firm's policy and the tool's terms. Some firms ban consumer AI tools; some allow an approved business plan. Do not assume that because a colleague does it, it is permitted.
5. Keep your own record. Save the approved final text in your normal client file system, not in the chat tool.

If you are unsure how far compliance generally goes, what compliance allows with ChatGPT covers the bigger picture, and AI note takers for advisers covers the separate question of meeting records.

Subscribe on YouTube@sagnikteaches

Where to keep the standing rules in ChatGPT

ChatGPT's custom instructions apply to all your chats. OpenAI's help page says they are available on all plans, with up to 1,500 characters on Free and Go and 5,000 on Plus, Pro, Business, Enterprise and Education. The same page says information from your use of custom instructions can be used to improve model performance unless you opt out, and that plug-in developers may receive relevant information from them. So the instructions are the place for your style and your rules, not for anything about a client. A short set might read:

I am a financial adviser. I use you for administration and communication only: drafting emails, agendas, checklists and plain-English explanations. Never give recommendations, suitability views or tax or legal conclusions. I use placeholders such as [Client], [Amount], [Date] instead of real details. If I paste anything that looks like a real name, account number or balance, stop and tell me to remove it. Write in plain English, short sentences, no jargon, no promises about performance, no guarantees. End every draft with a line "FOR ADVISER AND COMPLIANCE REVIEW BEFORE SENDING".

That last line is a small habit with a real effect: the draft says on its face that it is not ready to send. Remove the line when you paste the approved text into your email system.

Follow-up prompts

Prompt 1: the post-meeting follow-up

Draft a follow-up email to a client after a review meeting. Use placeholders for all personal details. Include: a thank-you in one line, a bullet list of the actions we agreed (I will list them: [action 1], [action 2], [action 3]), who is responsible for each and by when ([date]), what documents the client needs to send me, and when the next contact will be. Tone: warm, plain English, no jargon. Do not mention performance, returns or recommendations. Under 180 words.

Illustrative output: "Dear [Client], thank you for your time on [date]. Here is what we agreed. Actions for me: [action 1] by [date]; [action 2] by [date]. Actions for you: send [document] by [date]. I will be in touch on [date] to confirm. If anything here does not match your recollection, please tell me and I will update it." What you fix: put in your real details, check that the actions match your meeting record rather than your memory, and delete any sentence that implies advice you did not give. Then the usual review.

Prompt 2: the chaser for missing information

Write a polite, firm email asking a client for documents that are still outstanding. Placeholders: [Client], [document names], [original deadline]. Explain in one sentence why I need them (so that I can complete [task]). Offer two easy ways to send them securely through our usual portal. Make clear what happens if they do not arrive by [new deadline] without sounding like a threat. Three short paragraphs.

Sample output is competent. What you fix: the "secure channel" sentence must match what your firm actually uses. The model will not know, and may suggest email attachments. Replace it with your approved method.

Prompt 3: tone variants for the same message

When the same message goes to clients of different temperaments, ask for variants: "Rewrite the email in three tones: concise and businesslike, warm and reassuring, and formal. Keep every fact and placeholder identical." Pick one and edit. The identical-facts instruction matters because tone changes can quietly change meaning.

Review prep prompts

Review prep is where advisers lose evenings. Most of that work involves client data you must not paste, so the prompts here work on structure and wording, with the data handled in your own systems.

Prompt 4: a review agenda from a template

Create a 60-minute annual review agenda for a client meeting. Sections: welcome and changes since last time, review of goals, review of the plan's progress (I will bring the figures from our systems), upcoming life events, risks and protection check, actions and next steps. Add a suggested time for each and two open questions I could ask in each section. Do not include any figures or recommendations.

A usable result gives a timed agenda: 5 minutes welcome, 10 minutes changes since last time, 15 minutes goals and progress, 10 minutes upcoming events, 10 minutes risks and protection, 10 minutes actions. Questions might include "Has anything changed in your work or family that we should factor in?" What you fix: adapt the sections to your firm's review process, and align the questions with the fact-find fields you are required to update. How advisers use AI to prepare annual client reviews goes further into that workflow.

Prompt 5: a pre-meeting question list for the client

Write a short email (under 150 words) asking a client to bring or send ahead of their annual review: any changes in employment, family, health or property; any big expenses planned in the next two years; and anything they would like to discuss that I have not listed. Give them an easy way to reply. Use placeholders for names and dates.

This one is safe because the client supplies the data, not you. Sending it a week earlier usually means a better review meeting and fewer follow-up emails.

Prompt 6: plain-English explanations of your own process

Explain in plain English, in under 150 words, what happens in each stage of our onboarding: fact-find, risk profile questionnaire, planning, implementation, and first review. Use "you" and "we". Avoid jargon; if a term is needed, define it in a few words. Do not describe specific products or give any performance expectations.

Your onboarding guide benefits from this. The explanation is about process, not advice, so it passes the "admin and communication" test. What you fix: make sure the stages and timings match what your firm really does, and have compliance confirm the wording matches your documents.

A compliance review line and a short checklist

Add this checklist to the bottom of your drafting prompt or print it by your desk. It takes under two minutes per message.

• Does the text contain any recommendation, forecast, promise or guarantee? If so, remove it or route it through your advice process.
• Do all facts, dates and actions match the meeting record or client file?
• Are placeholders all replaced, and are there no stray bracketed words left?
• Is the tone suitable for this client and for this situation (a bereavement or a market fall needs a different register from a routine reminder)?
• Does it comply with your firm's rules on required wording, disclosures and records?
• Has it been approved by the right person, and is the approved version filed?

A suitable review line for your policy document might say: "Communications drafted with AI tools use placeholders only, are checked by the sending adviser against the client file, follow the same approval route as other client communications, and are stored in the client record." Have your compliance lead or adviser adapt it.

A six-adviser firm's review season, in minutes saved

Illustration: a firm of six advisers each holds about 120 client relationships. Review season means about 20 review meetings per adviser per month. Each review involves a pre-meeting email, an agenda, and a follow-up email, so three pieces of writing per review.

The firm's compliance lead approves five prompts (follow-up, chaser, agenda, pre-meeting email, process explanation) and the custom-instruction set above, and each adviser keeps the prompts in a shared document. Time per piece falls from roughly 12 minutes to roughly 6, including the review checks. Across 20 reviews and three pieces, that is 60 pieces; 6 minutes saved on each is six hours per adviser per month, or 36 hours across the firm. These numbers are illustrative: your own timings will differ, and the saving is only real if the review step is not skipped.

The firm does three other things that matter more than the arithmetic. It keeps a short log of where AI-drafted text was used, so it can answer questions later. It spot-checks five messages a month for stray details and unapproved claims. And it reviews the policy every quarter, because tools and rules change. For a wider rollout plan, the AI readiness checklist for professional firms is a good companion.

A placeholder convention the whole firm can share

Placeholders only work if everyone uses them the same way, because a half-replaced "[Client]" in a sent email is an embarrassing error, and a real name in a prompt is a breach. Agree one convention and put it in the shared prompt document.

Before sending any draft, search it for the "[" character. Zero results means every placeholder was replaced. This two-second check catches the most common slip.

A follow-up email, before and after review

Here is the kind of change a careful review makes. Illustrative text, with placeholders.

First AI draft: "Dear [Client], it was great to see you. I'm pleased to say your plan is on track and I'd recommend increasing your monthly contribution to [Amount] to make the most of the opportunity. Markets look strong, so now is a great time to act."

After the adviser's review: "Dear [Client], thank you for meeting on [Date]. As discussed, I will send you the updated plan summary by [Date]. If you would like to talk about your monthly contribution, I have suggested we set up a call on [Date] so that we can cover it properly. Please send [Document] before then."

The first version contains three problems. "On track" is a statement about performance that may not be supportable. "I'd recommend increasing" is a recommendation in an email drafted by a tool, outside any suitability process. "Markets look strong, so now is a great time to act" is a market view with a nudge to act. The reviewed version keeps the courtesy and the next steps and moves the contribution discussion into a proper conversation. This is why the instruction set says "no promises about performance, no guarantees" and why the review checklist asks about recommendations first.

Difficult moments: a different register

Some messages should not be produced from a template at all. A bereavement, a significant market fall, a complaint, news of a fee change. For these, ChatGPT can still help you think about structure, as long as the words end up yours.

I need to write to a client after a complaint about [issue]. Do not draft the full email. Suggest: (1) a structure of five short parts, (2) three phrases that acknowledge the problem without admitting liability for anything I have not confirmed, (3) three things I should check with compliance before replying, (4) a checklist for the tone: what to avoid.

The output is guidance for you, not text for the client. Typical points: acknowledge the concern in the first sentence, state what you will do and by when, avoid defensive language, do not speculate about causes, keep it short. The "check with compliance" list will usually include your complaints procedure and any required wording and time limits. Use it as a prompt for your own process. Your firm's complaint-handling rules govern the reply, and a model does not know them.

For a market-fall note to clients, a similar approach works: ask for a structure and tone guidance ("calm, factual, no predictions, remind clients how to reach you"), then write it yourself. Avoid asking the tool to explain what the market will do, because it has no reliable way to know and neither do you.

Internal admin that carries no client risk

Some of the most useful uses have nothing to do with clients. These sit safely inside the "admin only" rule because no client information is involved.

• Process documents. "Turn these bullet points about how we onboard a new client into a one-page checklist for a new team member." The bullets describe your process, not a person.
• Job descriptions and interview questions for a paraplanner or client-service assistant.
• Meeting agendas for team meetings, training sessions and compliance refreshers.
• Template wording for your own newsletter, with the content and numbers supplied by you and reviewed by compliance.
• Spreadsheet help. Questions about formulas for your own tracking sheets, using fake sample rows. See ChatGPT and Excel for how that works.

Starting with these builds the habit and the policy without touching client material at all. After a month of use you have a clearer view of where the savings are and which prompts are worth refining.

Onboarding colleagues and keeping the prompts current

A prompt set is a small piece of firm knowledge, and it needs an owner. Name one person as the keeper of the document. New staff get a half-hour walk-through: the rules, the placeholder convention, the approved tools, the review checklist, and two practice runs with fictional data. Their first three AI-assisted messages are checked by a more experienced colleague. That is the same way you would introduce any new communication standard.

Review the set quarterly. Things to check: whether the tool's terms or settings have changed (OpenAI's help pages describe custom-instruction limits and data-use options that may change), whether your regulator or firm has issued new guidance, whether any prompt keeps producing text that reviewers rewrite, and whether the approved tools list is still right. Date every change. When someone leaves, check that firm material is not left in their personal accounts.

Five slips advisers make with AI-drafted messages

• Data creeping into prompts. A balance, a first name and a family detail in a "quick" prompt. Spot it with a monthly search of your recent chats for numbers and names. Fix with the placeholder habit and the "warn me" line in the instructions.
• Advice slipping into drafts. A follow-up that says "you should consider moving to..." is a recommendation. Spot it with checklist question one.
• Overconfident reassurance. Phrases like "your plan is on track" without basis. Remove them unless you have verified the position from your own records.
• Template sameness. Clients notice when every email sounds identical. Edit each draft so it refers to something specific from the meeting.
• Unapproved tools. An adviser using a personal account for client work. Make the approved tools list explicit and short.

Checking the prompts work

Run a short test before rolling a prompt out. Use fully fictional placeholders, generate the draft, and have a colleague read it as a client would. Then have the compliance lead read it as a reviewer would. Compare their comments. If both want the same fix, change the prompt, not just the draft. Record the version of the prompt and the date it was approved. When a model changes or your firm's wording changes, re-test.

It also helps to compare two or three real follow-up emails you wrote before using AI with the AI-assisted version of the same type. The better version is usually the one that mentions a specific detail from the meeting, which only you can supply. If the AI versions read as more generic than your own, the prompts need more of your own phrasing in them.

Where the line sits, and who decides

The prompts above sit comfortably on the admin and communication side. Where it gets harder is the grey area: summarising a long product document, drafting a client-facing explanation of a rule, creating a market commentary. Each of those can slip into advice or into unapproved statements, and whether a firm allows them is a compliance decision, not a prompting one. A practical approach is a three-tier list: allowed without special review (agendas, chasers, process explanations); allowed with compliance review (educational explanations, newsletters); not allowed (anything involving client data, recommendations or performance claims). Write the list, date it, and update it as your experience and the rules develop.

If you would like help setting up the tiers, the prompt set and a simple review step for your firm, that is what my AI implementation consultation covers. It does not replace your compliance lead, who should always have the final say on what your firm may do.

Questions advisers ask before using ChatGPT

Can a financial adviser use ChatGPT at all?

Often yes for admin and communication, if the firm's policy and regulator's rules allow it and no client data goes in. Check your compliance manual first. Many firms permit drafting from placeholders and forbid anything that identifies a client or gives advice.

What counts as client data I must not paste?

Names, contact details, account and policy numbers, balances, holdings, dates of birth, family circumstances, health details and anything else that could identify a person or their finances. When in doubt, replace it with a placeholder or leave it out.

Should compliance review AI-drafted client emails?

Treat them like any other client communication: the adviser is responsible, and your normal review and record-keeping rules apply. Add a line to your process saying AI-drafted text is checked by a person before sending, and keep whatever records your firm requires.

Can ChatGPT give a suitability recommendation?

No, and you should not ask it to. Keep it to drafting, structuring and checking wording. Recommendations, suitability reasoning and anything that amounts to advice stay with the qualified adviser and the firm's documented process.

Further reads

• How to Use ChatGPT as a Business Adviser Without Being Misled — Why ChatGPT should not be your adviser.
• AI Engagement Letters for Accountants: Faster Drafts, Clear Scope — Drafting engagement letters in a neighbouring profession.
• Turning Hours Saved by AI Into Advisory Work Clients Pay For — Turning time saved into advisory work.
• Claude vs ChatGPT for Financial Advisers: Meeting Prep and Records — Which assistant suits adviser meeting prep and records.
• ChatGPT for Insurance Agents: Ten Safe Uses and Three to Avoid — Safe ChatGPT uses in another regulated trade.
• What to Do When AI Gets Something Wrong With a Customer — Steps to take when an AI draft goes wrong.
• HubSpot for Financial Advisers: Pipeline and Where Breeze Stops — Prospect and annual-review pipelines in HubSpot, what email logging does and does not keep, and where Breeze AI needs limits in an advice firm.
• AI Lead Generation for Financial Advisers Without Cold Outreach — Referral thank-yous, seminar follow-ups and newsletter segments drafted with AI, plus a channel-by-channel compliance checklist and one quarter's numbers.
• How to Use AI in a Small Consultancy: Proposals, Research and Delivery — Stage-by-stage AI use for a three-person consultancy: proposals, research and delivery, with the confidentiality rules that decide which tools you can use.
• Ongoing Advice Review Evidence File: A Checklist for UK Adviser Firms — A grouped checklist for proving ongoing reviews happened: meeting record, review pack and delivery log, with a filled-in sample file.
• Can AI Write File Notes From Adviser Meetings? Draft, Check, Sign Off — Draft a structured file note from a meeting transcript, then make adviser sign-off a named step with consent wording, a checklist and a retention rule.
• How Can AI Help Financial Advisors Find and Keep Clients Compliantly? — A solo adviser's month with AI: research prospects, draft content and nudge reviews, with an approval checkpoint before anything leaves the building.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI Help Center pages on ChatGPT custom instructions and Projects. Prompts, sample outputs and figures are illustrative; check your own firm's policy and your regulator's rules.