AI in wealth management: what an advice firm can hand over, and what it can’t

AI in wealth management works best on the paperwork around advice, not on the advice itself. It can turn meeting notes into a fact-find, prepare a first draft of a suitability report, write client letters, check files and assemble onboarding packs, while a qualified adviser makes the recommendation and signs it off. The same split applies to financial advisers, planners and mortgage brokers. This guide covers where it helps, where it must not decide on its own, and what the FCA expects, in plain terms.

Where AI helps in wealth management

The useful tasks share a pattern: a lot of reading and writing, a known template, and a person who already checks the result. That makes them safe places to start.

• Fact-finds. Where the client has agreed to the meeting being recorded, AI fills the fact-find from the transcript and marks every gap, along with anything that conflicts with what is already on file. The adviser confirms each field rather than typing it.
• Suitability report drafting. AI assembles a first draft from the fact-find, the adviser’s notes on the recommendation and your approved wording. It sets out the client’s circumstances and objectives in plain English and lays out the reasons the adviser gave. It should never invent reasons of its own.
• Client communications. Review invitations, meeting summaries, replies to routine questions and plain-English explanations of a product, drafted in your firm’s tone for a person to approve.
• File checking. Before a file reaches compliance, AI checks it against your own checklist: is the fact-find complete, does the report match it, are the disclosures and signatures there. It flags problems; a person decides what to do about them.
• Onboarding paperwork. Reading identity documents, proof of address and transfer forms, pulling out the details and listing what is missing so someone can chase the client. The guide to intelligent document processing explains how that reading works and where it goes wrong.

Where it must not decide on its own

Some steps carry the weight of the advice. AI can prepare material for them, but a person makes the call:

• The recommendation itself: what the client should do, and why it suits them.
• Attitude to risk and capacity for loss. A tool can score a questionnaire; an adviser judges whether the answers make sense for this client.
• Vulnerability. AI can flag words that suggest it, but deciding how to support the client is a human job.
• Anything sent to a client without a person reading it first.
• Moving money, changing a client’s instructions or updating bank details.
• The outcome of a complaint.

What the FCA expects, in plain terms

The FCA’s approach to AI is that it does not plan to introduce extra regulations for AI. It relies on the rules that already apply, and describes its approach as principles-based and focused on outcomes. In practice, AI doesn’t change your obligations. It changes how you show you meet them. This is practical guidance, not legal advice: check your own position with your compliance adviser.

• The Consumer Duty. In force since 31 July 2023 for products open to sale, it requires firms to act to deliver good outcomes for retail customers: act in good faith, avoid causing foreseeable harm, and enable customers to pursue their financial objectives. A letter drafted by AI still has to meet the consumer understanding outcome, which means clear, accurate and suited to the person reading it.
• Suitability. The FCA’s suitability rules (COBS 9, and COBS 9A for MiFID and optional exemption business) make the firm responsible for advice being suitable for the client, and MCOB 4.7A sets the equivalent test for mortgage advice. None of that changes because a model drafted part of the file.
• Accountability. The FCA says its rules on senior managers’ accountability, under the Senior Managers and Certification Regime, are relevant to the safe use of AI. Someone senior should own each AI tool, know what it does and be able to explain it.
• Records. If your regulator, an auditor or the Financial Ombudsman asks why a recommendation was made, the file needs to show the adviser’s reasoning, not just a well-written report.
• Client data. UK GDPR still applies inside an AI tool. Use business accounts with proper data terms and put in only what the task needs. The guide on using ChatGPT with customer data covers the practical rules.

Mortgage brokers have a useful signal too. In its December 2025 feedback on the mortgage rule review, the FCA reported a strong consensus among respondents that AI should improve the advice process but not replace advisers. It added that existing frameworks such as the Consumer Duty and the Senior Managers and Certification Regime can already manage many of the risks of new technology.

Why a person signs off the advice

Once the drafts get good, it is tempting to treat sign-off as a formality. It isn’t, for four reasons.

• The duty sits with the firm. The suitability rules make your firm responsible for the advice. A model cannot carry that responsibility, and a vendor’s terms will not take it from you.
• AI is fluent, not careful. It can produce a confident paragraph that misstates a client’s income or repeats a detail from an old template. Polished writing makes errors harder to spot, not easier.
• The adviser knows what isn’t written down. A client’s hesitation, a family situation, a remark at the end of a meeting. That context shapes good advice and rarely reaches the file in full.
• Sign-off is where the system improves. Every correction an adviser makes shows where the tool goes wrong. Logged and reviewed, those corrections tell you what to fix, and when it is safe to rely on the draft a little more.

I learnt this inside a regulated firm. At Alter Domus, a global fund administrator, I founded the AI Governance Committee that let our AI work run under the EU AI Act. The lesson that carries over to advice firms: clear controls are what let a tool reach production, not what stop it. The rest of that track record is on the AI for financial services page.

A sensible first project

1. Pick one document. Meeting notes into a draft fact-find is a good start: frequent, low risk and easy to check.
2. Use an approved, business-grade tool. No personal accounts, and data terms you have read. Write the rules down; the free AI acceptable use policy template is a starting point.
3. Define the check. Write down exactly what the adviser or paraplanner confirms before a draft moves on.
4. Keep the source next to the output. Each field in the draft should point back to the part of the transcript or document it came from, so checking is quick.
5. Run it alongside the old way. For the first few weeks, compare AI drafts with the ones your team would have written.
6. Measure. Time per file and the number of corrections, before and after, using your own numbers.

Who this suits, and who it doesn’t

It suits advice firms with a consistent process, standard templates and a compliance function that wants better evidence, not less of it. It also suits firms where paraplanners and administrators spend most of their week drafting and chasing. If you need the controls agreed and written down before anything is built, that is the work of an AI governance consultant.

It doesn’t suit a firm that wants AI to give advice to clients with no adviser involved, or one hoping a tool will replace its compliance checks. And if files and templates vary from adviser to adviser, tidy that first. Automating an inconsistent process gives you inconsistent drafts, faster.

If you run an advice firm and want to know which of these tasks would pay back first, book a free 30-minute AI consultation. We can walk through how a client file moves today and which step to start with.