AI for Financial Advisers UK: Suitability Letters and Meeting Notes

Within FCA rules, an adviser can use AI to draft suitability letter wording and turn meetings into file notes, provided the adviser decides the recommendation, checks every fact against the fact-find, signs personally and keeps evidence of the review. The FCA says it plans no extra AI regulation and relies on existing rules such as the Consumer Duty.

The temptation is to ask the tool to write the whole letter. The better use is narrower and safer: it organises facts you have already established, writes them clearly for the client and records what was said in the meeting. The recommendation, the judgement and the signature stay with you, and the client file should show that they did: a dated input pack, a checked draft and your initials.

Follow me on Instagram@sagnikteaches

What the FCA has said, and what it leaves to you

The FCA's page on its approach to AI states that it does not plan to introduce extra regulations for AI, and relies on existing frameworks: the Consumer Duty, which expects products and services that meet target customers' needs and give fair value, communication that meets customers' information needs and support that meets their needs, and the Senior Managers and Certification Regime, which places accountability on named individuals. Read plainly, that means a suitability letter drafted with AI is judged exactly as one drafted by hand. If it is unclear, unfair or wrong, the fact that a tool wrote the first draft does not help.

Connect on LinkedInSagnik Bhattacharya

The FCA's page on assessing suitability reminds firms that they must obtain the necessary information from the client to make a suitable recommendation: financial situation, objectives, attitude to risk, capacity for loss and knowledge and experience. It also covers insistent clients, where you give suitable advice, explain the risks of the alternative and make clear the client's action goes against your advice. Those are the pieces a draft letter must reflect accurately. Check the FCA's current pages and your compliance adviser's guidance for anything specific to your permissions; what follows is a workflow, not a regulatory opinion.

Subscribe on YouTube@sagnikteaches

The suitability report rule sets the minimum a letter must contain, and it makes a good test for any AI draft. In the version of the FCA Handbook I could read (COBS 9.4.7R, in an archived copy from December 2023, because the live Handbook would not load as text), the report must at least specify the client's demands and needs on the basis of the information obtained, explain why the firm concluded that the recommended transaction is suitable having regard to that information, and explain any possible disadvantages of the transaction for the client. A draft whose disadvantages section is thinner than its reasons section is not finished. COBS 9.4 is one of several suitability chapters in the Handbook, so confirm with your compliance adviser which rules govern your products and the timing of the report.

The drafting workflow in five steps

Step 1: decide the recommendation first, outside the tool

Complete the fact-find, the risk profile, the research and the cashflow work in your normal systems. Write the recommendation yourself in a short structured form: what is recommended, why it meets each objective, what alternatives were considered, charges, risks and any limits. This is the input the tool works from. If you let the tool decide, you cannot evidence that you did.

Step 2: prepare a clean input pack

Put the facts in one block with labelled fields and no more personal detail than the letter needs. Use the client's first name or a code rather than a full address and date of birth. A good pack looks like this (illustrative):

CLIENT: [first name], age 52, employed, one adult dependant.
OBJECTIVES: (1) retire at 60 with 28,000 a year in today's money;
            (2) keep 40,000 accessible as an emergency fund.
ATTITUDE TO RISK: medium (questionnaire score 5 of 10, discussed, client agrees).
CAPACITY FOR LOSS: moderate; a 20% fall would delay retirement by about 18 months.
RECOMMENDATION: consolidate two workplace pensions (values 112,400 and 38,700) into
  a single plan with a medium-risk model portfolio; keep the emergency fund in cash.
REASONS: lower combined charge (0.62% against 0.95% and 1.10%), single statement,
  investment range matches risk profile. LOSSES: loses guaranteed annuity rate on
  the 38,700 plan.
ALTERNATIVES CONSIDERED: leave as is; consolidate only the larger pot.
CHARGES: initial advice fee 2,400; ongoing 0.50% a year.
NEXT STEPS: client signs application by [date].

Notice the line on the lost guaranteed rate. The pack states the downside because the letter must, and because a model left alone tends to write the good news first and the risks vaguely.

Step 3: ask for a draft with strict rules

Draft a suitability letter in plain English for a client reading level of
about age 14. Use ONLY the facts in the pack below. Rules:
- Use my headings: Your objectives, Our recommendation, Why it suits you,
  What you give up, Alternatives we considered, Costs, Risks, Next steps.
- Quote every number exactly as written. Do not round or add numbers.
- Do not add products, funds, tax statements or benefits not in the pack.
- State the disadvantages as clearly as the advantages.
- If a field is missing, write [MISSING: field name] and do not fill it in.
- No marketing language. No guarantees.
[pack]

Step 4: check the draft line by line

Read the draft against the pack, not from memory. Strike anything not in the pack. A realistic fault: the draft says "the new plan has lower charges and no exit penalties" when the pack said nothing about exit penalties. The tool filled the gap with a typical claim. Because the rule said to write [MISSING] and it did not, tighten the prompt or shorten the pack. Another common fault is smoothing a risk: "may fall in value" for "could lose more than 20% in a bad year", the figure you discussed. Put the specific wording back.

Step 5: sign off and keep the evidence

Save four things in the client file: the input pack, the final letter, a tick-list showing the checks done and your initials with the date. You do not need to keep every intermediate draft unless your compliance adviser says so. The point is to show a person owned the content. The wider governance questions are in our AI governance checklist for advice firms, and the question of whether such a draft passes review is explored in can AI draft a suitability report that passes compliance.

Meeting notes: from recording to file note

Meeting capture is the other half of the job and usually the larger time saving. The method has three layers: capture, summarise, verify.

Choosing a note-taker

You have three kinds of option. A general note-taker such as Otter (Pro from $8.33 a user a month billed annually, $16.99 monthly), Fireflies (Pro $10 a seat annual, $18 monthly) or Fathom (Team $15 a seat annual, with a two-user minimum) joins calls and writes summaries. Adviser-specific tools and CRM add-ons produce file-note formats and push notes to the client record; compare them in the cost of a built-in note-taker against a standalone one and which features are native in an adviser CRM. A third option is the transcript feature in your meeting software. Prices above are the vendors' USD list prices, checked in November 2026 (Fathom Team is $19 on monthly billing, and Otter was showing a temporary discount on Pro at the time); check the current page.

Whatever you choose, ask the vendor five things in writing: is business content used to train models; where is data stored and for how long; can you delete recordings and transcripts; how does the tool handle the client's special category data such as health details; and does it produce an audit trail showing who edited a note. A tool that cannot answer is not for client meetings.

Tell the client and keep the notice

Add a line to the meeting invitation and open the meeting by saying that you use a note-taking tool, what it records, how it is stored and that they may decline. If a client declines, take notes by hand. Record that the notice was given in the file note. This also serves the UK GDPR requirement to be transparent about how personal data is used. The ICO's guidance on AI and data protection covers transparency, accuracy and accountability, and it is direct about impact assessments: it says that in the vast majority of cases using AI will involve processing likely to result in a high risk to individuals, which triggers the legal requirement for a data protection impact assessment, and that if you conclude a particular use is not high risk you still need to document how you reached that view. The ICO also notes that the guidance is under review following the Data (Use and Access) Act, so read the current version. Your compliance lead or data protection adviser should set the firm's position.

The file-note prompt

From the transcript below, write a file note under these headings:
Attendees and date | Client's circumstances changes | Objectives discussed |
Risk and capacity discussion (quote the client's own words where given) |
Advice given | Client decisions | Actions, owners and dates | Anything unclear.
Rules: use only what is in the transcript; mark any figure that was not
stated clearly as [CHECK]; do not infer feelings or intentions; do not
add recommendations that were not made in the meeting.

Illustrative output (extract): "Client's circumstances: moved to part-time hours from 1 March; household income now about 3,100 a month net [CHECK: client said 'around three thousand one hundred'; confirm]. Risk: client said 'I can live with a dip as long as I am not forced to sell.' Decision: client to think about consolidation and reply by 15 April. Action: adviser to send charge comparison by 8 April." The [CHECK] flag earns its place, because transcription services often mishear numbers.

Verify before filing

Skim the draft against the recording for the passages that matter: numbers, risk statements, decisions and anything the client said they did not want. Correct and file. If you do this within a day, memory helps. If you leave it a week, the shortcut becomes a liability.

Getting Consumer Duty evidence from the same material

The Consumer Duty asks firms to show good outcomes for retail customers, including that communications are understandable and support meets needs. You can build evidence from material you already produce. Three simple habits help.

• Understanding checks. Add a line to the file note template: "How did we check the client understood the recommendation?" and record what you asked and what they said.
• Vulnerability flags. Include a field for any signs of vulnerability noticed or disclosed (bereavement, health, low confidence with money), and what adjustment you made. Do not ask the model to infer vulnerability from the transcript; capture what was said and what you saw.
• Readability. Ask the tool to rewrite a paragraph of the letter in simpler language, compare the two, and keep the version that is accurate and clear. Ask for a short summary box at the top that states the recommendation, cost and main risk in four lines.

Keep the evidence where you can retrieve it by client. A folder of untagged notes proves little in a review.

A worked example: one pension consolidation case

Illustration. An adviser in a three-person practice sees a client on a Tuesday. The note-taker, with notice given, produces a transcript. By 4 p.m. the adviser has the draft file note, corrects a figure marked [CHECK], and files it, about fifteen minutes of work against forty minutes writing from memory and scribbled notes. On Thursday the adviser completes the research and builds the input pack above. The first draft of the letter takes ten minutes to generate. Checking each line against the pack takes thirty-five minutes and finds three issues: an unsupported "no exit penalties" claim, a softened risk sentence and a missing ongoing-charge figure. After fixes the adviser signs, saves the pack and the tick-list, and sends it. The firm estimates the process saves about an hour and a quarter per case across notes and letter. Treat those minutes as an illustration; time your own cases for a month before you plan around them.

Four ways an AI-drafted letter or note goes wrong

The confident invention

Tools write fluent text, and fluent text feels checked. The cure is the pack and the rule that missing fields become [MISSING]. A letter with an invented sentence about tax treatment is an error the adviser owns.

Copying a previous client's detail

If you reuse a conversation across clients, details can bleed. Start a new conversation for every client, never paste two clients into the same chat, and turn off any memory feature that carries information between conversations unless your compliance lead has approved it.

The note-taker that joins every call

Auto-join features can record meetings you did not intend, such as an internal discussion about a complaint. Switch auto-join off and invite the tool only to client meetings where notice has been given.

Treating the summary as the record

A summary is shorter than the meeting and will omit things. Keep the transcript or recording according to your retention policy so the original is available, and note the reason if you delete it.

Before and after: fixing a risk paragraph

The clearest way to see what checking means is to compare drafts. Here is an illustrative first draft of a risk paragraph, and the adviser's corrected version.

First draft from the tool: "Your money will be invested in a diversified portfolio designed to grow steadily over time. While values can go down as well as up, the portfolio is managed by experts and is well placed to deliver the returns you need to retire at 60."

Problems: "designed to grow steadily" implies a smooth path; "managed by experts" is marketing; "well placed to deliver the returns you need" is close to a promise; nothing mentions the 20% fall discussed in the meeting or the 18-month delay it would cause.

Corrected by the adviser: "The value of your investments can fall as well as rise, and you may get back less than you put in. In a bad year your portfolio could fall by around 20%. Based on our calculations, that would delay your retirement date by about 18 months. We have recommended a medium-risk portfolio because you told us you could accept a fall of that size, as long as you were not forced to sell. Nothing in this recommendation is guaranteed."

The corrected version is longer, more specific and less attractive, which is what a suitability letter should be. Keep a library of before-and-after pairs like this one. They make the best training material for new advisers and the best test cases when you try a different tool: feed the same pack to a new model and see whether it makes the same mistakes.

A second case: an insistent client

Illustration. A client wants to transfer a defined benefit pension into a drawdown arrangement against your advice. Pension transfer advice carries further rules of its own, so treat this as an example of the letter, not of the transfer process. The FCA's suitability page says that for insistent clients you give suitable advice, explain the risks of the alternative course and make clear that the client's action goes against your advice. A letter for that situation is delicate, so the pack must carry three extra fields: "ADVICE GIVEN: remain in the scheme", "CLIENT'S DECISION: proceed against advice, reasons in client's words" and "RISKS EXPLAINED: list with dates of the meetings". Instruct the tool: "Do not soften the advice given. Do not describe the client's choice as sensible or reasonable. Do not use the words 'we support'." A model's instinct to be agreeable is precisely wrong here. The adviser should write this letter's core paragraphs by hand and use the tool only to tidy structure and readability, and the compliance lead may want to see every such case.

Readability: rewrite, compare, keep the accurate one

Plain English is part of acting in clients' interests, and AI is good at it, with one risk: simplifying can change the meaning. A sound method is to ask for a rewrite of one paragraph and then ask the same tool to list what changed in meaning:

Rewrite the paragraph below at a reading age of 12. Then list every place
where the rewrite says something different from, or less than, the original.
Do not remove any risk, cost or condition.

If the list comes back with "removed the reference to the guaranteed annuity rate", the simpler version is wrong and you keep the original wording, perhaps with a short explanation added beneath it. A reading-level target is a guide; the client's understanding is the test, which is why the understanding-check line in the file note matters.

Data protection and retention in practice

Client files contain personal and often sensitive data, so the practical rules are about minimising and keeping records. Use the smallest input that works: first name and age bands rather than full identifiers where the letter does not need them. Check where the tool stores data and for how long; choose a business plan where the vendor states it does not train on your content. Decide your retention for recordings and transcripts alongside your file retention policy, and write it down. Agree who can delete a recording at a client's request, and how you handle a subject access request that covers transcripts and AI-generated notes: those are records about the client, so your system must be able to find them. Record these decisions in the impact assessment you completed when the tool was introduced, so the reasoning sits in one place; your data protection adviser should confirm the details for your firm.

A pilot plan for a small firm

1. Week 1: agree the rules with your compliance lead; pick one note-taker and one drafting tool; write the notice wording and the input pack template.
2. Weeks 2 to 3: use the note-taker with clients who agree, on ten meetings. Compare each file note against your own notes and count corrections.
3. Weeks 4 to 5: draft five letters from packs and compare with letters written by hand for similar cases. Record time saved and faults found.
4. Week 6: compliance lead reviews all fifteen files. Decide to continue, change the prompts, or stop. Record the decision.

A small pilot with written outcomes is the best evidence that you introduced the tool carefully, and it gives you real numbers instead of a vendor's claims. For the next step, see how the same notes feed annual client reviews.

A review checklist to adopt

The sample audit closes the loop: if the compliance lead finds repeated faults of one kind, change the prompt or the pack rather than telling advisers to be more careful. The broader decision of how much to trust a general chat tool is in can financial advisers use ChatGPT, which covers what compliance allows.

If you want to build this with your own templates and a tool your compliance lead has approved, that is what my AI implementation consultation covers, and you can start this week with the input pack, because it improves your letters even without AI.

Adviser questions about AI drafting and the FCA

Does the FCA have specific AI rules for advisers?

On its AI approach page the FCA says it does not plan to introduce extra regulations for AI and relies on existing frameworks, including the Consumer Duty and the Senior Managers and Certification Regime. Your existing duties apply to anything an AI tool helps produce.

Can I let the AI write the recommendation itself?

No. The recommendation comes from your analysis, your research and your assessment of suitability. AI can draft wording around a recommendation you have already made and recorded. If it proposes a product or a switch, treat that as an error to remove.

Do I need the client's consent to record a meeting?

Tell clients you use a note-taker, say what it records, where the data goes and how long it is kept, and give them the choice to decline. Record the notice in your file. Take your compliance adviser's view on the exact wording and on whether the audio itself is kept.

Who is responsible if the AI draft contains a mistake?

The adviser who signs the letter and the firm. A tool is not a defence. This is why every figure, fact and recommendation in a draft is checked against the fact-find and the research before it goes out.

Further reads

• Can AI Write File Notes From Adviser Meetings? Draft, Check, Sign Off — A draft, check and sign-off routine for file notes.
• AI Financial Planning Software for Advisers: What to Test First — What to test before trusting planning software with AI.
• AI Agents for Financial Advisers: What They Can Do Without Advising — Where agents help with admin and where they must stop.
• AI for Wealth Managers: Drafting Proposals With a Compliance Review — Drafting proposals with a compliance review built in.
• AI Note Takers for Financial Advisers: Compliant Meeting Records — Compliant meeting records and how to choose a note-taker.
• AI Receptionist for Financial Advisers: Cost, Scripts and Limits — Cost, scripts and limits for an AI receptionist.
• Claude for Financial Advisers: Plugin, Connectors and Plan Needed — What Claude offers advisers and which plan it needs.
• Jump AI With Wealthbox: Syncing Notes, Tasks and Contacts — Connect Jump to Wealthbox, set the automations, then prove notes, tasks, contacts and workflows each land in the right place.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: FCA pages on its approach to AI (updated 2 October 2026) and on assessing suitability; FCA Handbook COBS 9.4 (archived December 2023 copy); ICO guidance on AI and data protection and its accountability chapter (archived September and October 2026 copies); Otter, Fireflies and Fathom pricing pages, checked November 2026.