AI & professional risk

AI Use and Your PI Insurance: What to Document

In short: If your firm uses AI tools in client work, the way to protect your professional indemnity position is to document it: which tools you use, for what, who reviewed the output, and why you decided to rely on it. Because PI policies are claims-made and claims surface years later, contemporaneous records are what stand between you and an argument you cannot evidence. Apex is an independent, FCA-authorised UK broker (FRN 724952) advising professional firms across the UK on exactly these questions.

This page is for any professional firm — architects, surveyors, accountants, consultants, lawyers, engineers — that has started using AI tools in fee-earning work and wants its insurance position to hold up. It is about records and process, not about whether you should use AI at all.

First, the regulatory picture — stated carefully

One professional body has, as far as we have verified, made AI governance mandatory for its regulated firms: RICS, whose professional standard on the responsible use of artificial intelligence takes effect on 9 March 2026 and applies to RICS members and regulated firms. If you are a surveying practice, that standard is not optional and we have written about it separately — see the RICS AI standard explained.

For other professions, we are not aware of an equivalent mandatory AI standard at the time of writing. Everything that follows is therefore framed as good practice and emerging expectation, not as any body’s requirement — but the direction of travel is clear, and the firms that document now will be the ones comfortable later, whether the question comes from a regulator, an insurer or a claimant’s solicitor.

Why claims-made cover makes records vital

Professional indemnity policies are written on a claims-made basis: the policy that responds is the one in force when the claim is made, which is routinely years after the work was done. That time lag is the whole problem. If a piece of advice produced in part with an AI tool is challenged in several years’ time, the questions will be: what tool was used, what was it used for, who checked the output, and on what basis did a qualified person decide to rely on it?

If the answers exist only in memory, you are defending a negligence allegation with recollection against hindsight. If they exist in a contemporaneous record, your defence starts from evidence. In professional negligence disputes the firm that can show its process — what was done, by whom, and why — is in a categorically stronger position than the firm that can only assert it. Nothing about that is new; AI simply adds a layer of process that most firms have not yet built the habit of recording.

What to record, per engagement

You do not need an enterprise governance programme to start. You need a consistent, dated record — in the matter file, where your other working papers live — covering four things:

1. Which tools, and which versions. Name the AI tools used on the engagement, including tools embedded inside software you already use. Note the version or date where you can: tools change quickly, and “the model behaved differently then” is an argument you can only make with a date.

2. What each tool was used for. Be specific about the task: first-draft text, summarising documents, checking calculations, generating options, transcription. The record should let a stranger distinguish between AI as a drafting aid and AI as the source of a professional judgement — because a claimant will try to blur exactly that line.

3. Who reviewed the output, and how. Name the qualified person who checked the AI-assisted work and what the review consisted of — checked against source documents, recalculated, verified citations, site knowledge applied. Human review is the control that underwriters and courts will look for; an unnamed “the team checked it” is barely a record at all.

4. The decision to rely, and why. The most valuable line in the file is the one that says a named professional considered the output, satisfied themselves of it for stated reasons, and adopted it as their own work. That line converts “the machine said so” into “a professional exercised judgement” — which is the position every PI defence wants to start from.

Alongside the per-engagement record, keep a short firm-level note: which tools are approved, which uses are prohibited, and how client confidentiality and data protection are handled when material is put into a tool. Date it and update it when your practice changes.

Client disclosure as risk management

Telling clients how you use AI — in your engagement letter or terms — is not just transparency for its own sake; it is risk management. A client who was told, at the outset, that the firm uses AI tools under professional review has a much weaker version of the “we never agreed to this” argument than a client who discovers it after a dispute has started. Disclosure also forces the internal discipline of knowing what you actually do with these tools, which is precisely the record described above. Keep the wording honest and specific enough to be meaningful; a vague boilerplate line that does not match reality can be worse than silence.

What underwriters are likely to ask

Proposal forms in most professions do not yet interrogate AI use in depth, but underwriter interest is growing and questions are beginning to appear. Expect, over the coming renewal cycles, to be asked some version of: which AI tools the firm uses in client work; what they are used for; whether output is reviewed by a qualified person before it goes to a client; whether the firm has a written policy on AI use; and how confidential client information is protected when tools are used. Firms that have kept the records described here can answer in a sentence each. Firms that have not will face the uncomfortable choice between a vague answer and an audit of their own practice under renewal deadline pressure.

We will not speculate on how AI use will affect premiums. What we can say from experience is that underwriters price uncertainty, and a firm that can evidence a controlled, reviewed, documented approach presents a clearer risk than one that cannot.

Frequently asked questions

Do I have to tell my PI insurer I use AI?

Answer every question on the proposal form honestly and completely — if it asks about AI, tell them. Under the duty of fair presentation, a commercial policyholder must disclose what a prudent insurer would want to know, so if AI is material to how your firm delivers its work, the safe course is to present it plainly rather than leave it unsaid. Documented, reviewed use is not something to hide; it is something to present well.

Is any professional body’s AI standard actually mandatory?

As far as we have verified, RICS is the professional body with a mandatory AI standard, effective 9 March 2026 for its members and regulated firms. For other professions, published material is currently guidance and good practice rather than binding requirement — but check your own body’s current position, because this area is moving quickly.

Does using AI invalidate my PI cover?

Using AI tools does not, of itself, void a PI policy, and we are not aware of standard UK PI wordings that exclude AI-assisted work as such. The realistic risks are different: a claim arising from unreviewed AI output that you cannot defend for lack of records, or a non-disclosure argument if the proposal asked and the answer was incomplete. Both risks are addressed by the same thing — documented, reviewed use, honestly presented. Check your own wording, and ask us if anything in it is unclear.

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority (FRN 724952). This article is general information, not advice on your specific circumstances.