Discussion

AI in financial advice: existing rules already put competence on the compliance agenda

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#13748

Financial firms cannot hand responsibility for AI-assisted work to a supplier or wait for a bespoke rulebook, argues Gareth Fatchett, a partner at FS Legal. His practical point is that existing duties around oversight, competence and customer outcomes already follow the work into AI use.

Watch Desk analysis

What happened

In an opinion article for Money Marketing, Fatchett describes a paraplanner using AI to summarise a client meeting. The summary misses a reference to deteriorating health, but still reads well enough to inform an adviser’s recommendation. The example shows why fluency is not the same as a complete account of a client’s circumstances.

Fatchett argues that existing regulatory obligations apply to AI-assisted processes. He points in particular to firms’ duties around effective management and control, suitable delegation and oversight, and keeping staff competent for the work they do. He also stresses that responsibility depends on the activity: an AI tool does not make every administrator an adviser, and SMF16 or SMF17 responsibility is not automatic for every AI application.

Why it matters

The article gives firms concrete questions to ask before an AI tool becomes part of routine work. Which tools and uses are approved? What information may staff enter? Which decisions need human approval? A firm should also check whether client information could be retained or used for model training, and whether removing names is enough to protect confidentiality.

Fatchett’s advice is to make human review substantive: check outputs against the fact-find, product information and applicable requirements, rather than asking the same model whether it is sure. He also urges task-specific training, records of competence and authorised uses, and monitoring for errors, overrides and complaints. Those are his recommendations for putting existing responsibilities into practice, not a claim that the FCA has issued a universal AI checklist or annual training timetable.

Our read

The useful shift here is from “Which AI tools have we bought?” to “Can we show who is competent to use each tool, for which task, and how the result is checked?” A polished summary that omits a health concern is not a minor formatting quibble when it feeds into advice.

Fatchett’s strongest point is that a supplier demonstration cannot do a firm’s oversight for it. Start with an inventory of AI features already embedded in ordinary software, then define approved uses, information boundaries and meaningful review. The paperwork matters; so does making sure the person checking the output can spot what it missed.

What to watch

  • Whether firms set clear approval, data and human-review boundaries for each AI use.
  • Whether training and competence records reflect the actual work staff perform.
  • Whether monitoring catches omissions, unsupported claims, client complaints and failures early.

Discussion spark: Should financial firms have to demonstrate staff competence for each AI-assisted task before using it with clients, or are existing oversight duties enough?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.