AI for financial advisers in the UK: a plain guide for 2026

Every adviser has now been told that AI will change their business. Far fewer have been told, in plain terms, what it is actually good at, where it goes wrong, and what the regulators expect from a firm that uses it. That gap makes it hard to judge which tools are worth trying and which claims to ignore.

This guide to AI for financial advisers in the UK sets out what today's AI tools do well and badly in advice work, the FCA and data protection position (including the ICO, which became the Information Commission on 30 September 2026), the questions to ask about client data, and a sensible way to start.

What AI does well in advice work

The AI models behind tools such as Claude and ChatGPT are very good at working with large amounts of text. In a planning firm, that maps onto a familiar list of jobs:

• Reading long documents quickly. A client folder of fact finds, valuations, letters and review notes can be read in minutes rather than hours.
• Extracting figures. Pulling values, contributions and charges out of platform valuation PDFs and putting them side by side.
• Writing a first draft. Turning a meeting transcript into a discovery summary or a review narrative, in a set structure and house style.
• Comparing against a checklist. Checking whether a file contains what it should and noting what is missing.
• Measuring a conversation. Counting talk share and open and closed questions in a transcript.

The common thread is that each of these is mostly reading and assembling, with a decision at the end that a person makes. That is where AI saves time without taking over responsibility.

Where AI goes wrong

The same models have weaknesses that matter a great deal in regulated work:

• Confident errors. A model can state something fluently that is not in the source document, or get a figure wrong while sounding certain.
• Incomplete reading. A scanned page, an unusual table or a file that will not open can be skipped without anyone noticing.
• No context beyond what it is given. It knows nothing about the client except what is in the documents and transcript, and it cannot read the room.
• No accountability. It cannot take responsibility for advice. The firm and the adviser do.

Good tools are designed around these weaknesses. When you look at any AI product, ask whether it shows where each figure or finding came from, and whether it tells you when it could not read something. In our own apps, for example, Strata traces every figure back to the report and page it came from and reports any file it could not read, and Scrutineer flags any review that could only read part of a folder and quotes the passage behind each finding. Whatever tool you use, that kind of traceability is what lets a person check the work quickly.

The FCA position on AI

There are no AI-specific FCA rules for advice firms. The FCA's AI Update explains that it relies on its existing framework, including the Consumer Duty, the Senior Managers and Certification Regime and SYSC. Its blog of 8 June 2026 repeats that the FCA will "rely on existing frameworks, including the Consumer Duty, the Senior Managers and Certification Regime (SM&CR), and our expectations on governance and controls."

In practice, that means the questions you would ask of any other process apply to AI too:

• Outcomes. Under the Consumer Duty, PRIN 2A.2.8R says: "A firm must avoid causing foreseeable harm to retail customers." An AI error that reaches a client unchecked is a foreseeable harm.
• Accountability. Someone senior should own the firm's use of AI, decide where it is allowed, and be able to explain the controls.
• Records. SYSC 9.1.1R requires "orderly records" sufficient to enable the FCA to monitor compliance. Keep the approved version of any AI-drafted document, and know who approved it.

The ICO position on AI and client data

Client data in advice work is personal data, often including health and financial details. The ICO's guidance on AI and data protection is direct about what that means. It says: "In the vast majority of cases, the use of AI will involve a type of processing likely to result in a high risk… and will therefore trigger the legal requirement for you to undertake a DPIA."

It is equally clear that the decision on lawful basis is yours: "it is your responsibility to decide which lawful basis applies". Note that the ICO marks this guidance as under review following the Data (Use and Access) Act, so check the current version when you write your assessment.

Questions to ask about where client data goes

• Which company processes the data, and in which account?
• Does the vendor hold a copy of client documents on its own servers?
• What do the AI provider's terms say about using your data to train models, and about retention?
• Can you switch off or delete anything stored, and what happens when you leave?

Our own approach is that the apps run on the firm's computers and the AI work goes through the firm's own Claude or ChatGPT account, chosen in setup. IFA Tools does not hold the firm's client data, though the AI provider the firm chose does process it, under the firm's own account and terms. We compare the main models in where client data goes with AI tools.

How to start safely

A cautious start is quicker in the end than a rushed one. A simple sequence:

1. Pick one job. Something frequent, mostly mechanical, and always checked by a person before it reaches a client.
2. Complete a DPIA. Describe the data, the providers involved, the risks and the controls. Record your lawful basis.
3. Decide the review step. Who checks AI output, against what, and how that is recorded.
4. Measure before and after. Time per item, number of corrections and the reviewer's view of quality.
5. Tell your team the rules. Which tools are approved, which data can go into them, and what must never be pasted into a personal account.

Our rollout plan for introducing AI in an advice firm goes through each step in more detail.

Frequently asked questions

Does the FCA allow financial advisers to use AI?

There are no AI-specific FCA rules. The FCA says it relies on existing frameworks, including the Consumer Duty, SM&CR and its expectations on governance and controls. Firms remain responsible for outcomes, whatever tools they use.

Do I need a DPIA before using AI with client data?

The ICO says that in the vast majority of cases AI processing is likely to be high risk and will trigger the legal requirement for a DPIA. For client advice data, plan on completing one.

Can AI write suitability reports?

AI can draft from a transcript, fact find and figures, but the adviser must check and approve every report. The suitability judgement and the reasons for it remain the firm's responsibility.

Scrutineer reads a whole client folder, scores it against FCA rules and your firm's own standard, and quotes the evidence behind every finding. Read more about Scrutineer, or book a twenty-minute demo with Adam Booth on Microsoft Teams, run live on a demo firm with nothing for you to prepare.