← The Clair Blog

    ChatGPT for financial advisors: what to trust it with, and what to never

    Everything an advisor writes can be a regulated communication, which changes the whole AI question. Where it helps, where the regulator lives, and the client conversation that is coming either way.

    A note before the split, and here it carries real weight. This is general guidance, not compliance, legal or financial advice. Your regulator, your firm's compliance function and your professional obligations override every word below, and nothing here changes what counts as advice, a financial promotion or a suitability obligation in your jurisdiction. Tools change constantly too, so specifics are a snapshot.

    Most professions can start the AI conversation with the tasks. Advisors have to start somewhere else, because the defining fact of the job is that almost anything you write to a client can be a regulated communication, which makes a tool that generates confident text at speed rather more than a productivity question.

    It is a compliance surface. That is why the honest split for advisors does not run along what the tool is good at, it runs along what ever reaches a client, and what never does, and once that line is drawn a surprising amount of the week sits safely on the useful side of it.

    The short answer: behind the line, ChatGPT earns its keep the way it does in any words-heavy practice, on meeting preparation and follow-up summaries from your own notes, first drafts of newsletters and market commentary that then go through compliance exactly as they always did, dense fund documents translated into language a client can follow, and the admin volume. In front of the line sit the absolutes, which are no client-identifiable financial information on consumer versions, no generated content reaching a client without your review and your firm's process, and nothing that resembles a personalised recommendation ever coming from the tool, because suitability is a professional obligation and it is not transferable to software.

    Behind the line: the work it can carry

    The job What that looks like
    Meeting preparation and follow-up Your own notes turned into a review agenda beforehand and a summary letter draft afterwards, with client details stripped or handled inside a firm-approved tool
    Content first drafts The newsletter, the quarterly commentary, the explainer for the website, drafted in minutes and then entering your normal compliance review, which does not move
    Translation A fund factsheet, a policy document or a piece of regulation rendered into plain language, once for your orientation and once for the client-friendly version you will verify
    The admin layer Process documents, meeting checklists, the recurring correspondence that fills the gaps between reviews

    The pattern across the four is that the tool drafts and you, plus your existing compliance process, remain the gate. Nothing about the review chain changes, only how long the first version takes.

    In front of the line: where the regulator lives

    Three behaviours account for nearly all the ways this technology hurts advisors, and each maps to an obligation that predates it:

    • The paste. Client names, holdings, circumstances and account details entering a consumer tool that can learn from inputs and retain them, which is a confidentiality and data breach in a profession where that carries regulatory weight.
    • The invented number. These tools generate plausible figures rather than calculating, so they will produce performance claims, projections and comparisons from nothing, any of which lands somewhere between misleading and a breach if it reaches a client. Every figure in client-facing work traces to a real source, with no exceptions for numbers that look reasonable, since looking reasonable is what generated numbers do.
    • Drift into advice. The tool will happily produce what reads like a personalised recommendation if asked, and what makes a communication advice is its content and context rather than whatever produced it.

    The suitability obligation, the reasoning and the accountability are yours, which is less a limitation of the technology than the definition of the profession. There is more on the invention problem in how to tell when ChatGPT is making it up, and a broader view of the same line in ChatGPT for finance professionals.

    The clients who already use it

    The other half of this subject arrives from the opposite direction, because your clients have ChatGPT too, and some now turn up to reviews with its opinions about their portfolios, their fees or a product they have read about.

    It is tempting to hear that as the robo-advice threat returning in a new voice, though in practice it tends to run the other way. A generated portfolio take is generic by construction, since it knows nothing of the client's tax position, goals, history or the conversation you had in March. Walking a client calmly through where the machine's answer fits their actual situation, and where it cannot, is a live demonstration of exactly what the fee buys.

    The advisors who will feel threatened by this are the ones whose service was generic to begin with. The product an advisor sells is accountable, personalised judgement, and accountability is the one thing this technology structurally cannot hold. Clients asking better questions is not the threat, it is the sales pitch writing itself.

    The one-line version

    The tool drafts behind the line, everything client-facing passes through you and your compliance process unchanged, no client data touches a consumer version, no generated number survives unverified, and no recommendation ever originates in software.

    Hold that and the hours come back from exactly the places they should, the drafting and the translating, while the part of the job that justifies the profession stays untouched, because it was never automatable in the first place.

    Not sure where to start?

    The drafting hours back. The judgement untouched.

    The free finder tells you which AI to open for the task in front of you, in about 60 seconds, and the full "AI, sorted." reference lands in your inbox.

    Try the free finder →

    One question in, one answer out.


    Clair helps non-technical professionals know when to trust their AI, when to check it, and when to skip it.

    Common questions

    Can financial advisors use ChatGPT?

    Yes, behind a firm line. It is genuinely useful for meeting preparation and follow-up drafts from your own notes, first versions of newsletters and commentary that then go through normal compliance review, translating dense documents into client-friendly language, and admin. Client-identifiable data, unverified figures and anything resembling a personalised recommendation stay entirely off it.

    Is ChatGPT compliant for financial advice?

    No tool is compliant or non-compliant by itself. What matters is that regulated obligations do not move: client-facing content goes through your firm's review process regardless of what drafted it, suitability and recommendations remain your professional responsibility, and client data stays out of consumer versions. Your compliance function's guidance overrides everything.

    Can I put client information into ChatGPT?

    Not on consumer versions, which can use inputs for training and retain them, making client names, holdings and circumstances a confidentiality breach with regulatory weight. Strip identifying details into a genuine hypothetical, or work inside a firm-approved tool with a proper data agreement, and follow your firm's policy first.

    Why can't ChatGPT give financial advice?

    Because what makes a communication advice is its content and context, and the obligations attached, suitability, accountability, liability, belong to a person and a firm, not to software. The tool also invents plausible figures and projections, which in client-facing material sits somewhere between misleading and a breach.

    What do I do when clients quote ChatGPT at me?

    Treat it as an opening rather than a threat. A generated take knows nothing of the client's tax position, goals or history, so walking them through where it fits their actual situation, and where it cannot, demonstrates precisely what your fee buys. Verify any claim it made, correct it calmly, and let the comparison do the selling.

    The AI Starter Kit

    Set up ChatGPT and Claude properly. Once.

    The workflows, rules and habits that hand the boring recurring work to your AI and give you your evenings back.

    Get the Kit →