PI insurance for firms using AI in client work
Who this affects
Almost every professional firm, whether it has an “AI policy” or not. Accountants using AI to draft analysis and correspondence; surveyors using it in report production; consultants using it for research and first drafts; architects and engineers using AI-assisted design tools. The common feature is that AI output feeds into advice or work product a client relies on — which puts it squarely inside the territory professional indemnity insurance exists for.
It also affects firms that believe they do not use AI. In practice staff often adopt tools individually before any policy exists. From an insurance point of view, what matters is what actually happens in the firm, because that is what the duty of fair presentation covers — not what the official position says.
What underwriters now ask
Underwriters increasingly want to know whether firms use AI in client work, and the questions are becoming more specific over time. Expect some or all of the following at renewal: whether the firm uses generative AI in producing client deliverables; which tools, and whether they are enterprise versions or public free tools; whether client data is entered into them; whether output is reviewed by a qualified person before it reaches a client; whether there is a written AI-use policy and staff training; and whether the firm’s engagement terms say anything about AI use.
None of these questions is designed to catch you out. They are the underwriter’s way of separating firms that have thought about the risk from firms that have not — and pricing accordingly.
Turning AI governance into a positive risk story
The instinct at renewal is to say as little as possible. With AI it is the wrong instinct. A firm that can present a short, concrete governance picture reads as a better risk, not a worse one. The elements underwriters respond to are simple: a list of approved tools and what they may be used for; a rule that no AI output reaches a client without review by someone qualified to catch its errors; restrictions on entering confidential client data into public tools; training so staff know both; and someone senior who owns the policy.
None of that requires a technology budget. It requires the same thing PI underwriters have always rewarded: evidence that the firm supervises the work that goes out under its name. AI does not change that principle — it just gives you a new chapter of it to write down.
What to check in your own wording
Most PI wordings were written before generative AI, so the starting point is finding out how yours would treat it. The things worth checking: whether any AI or “technology” exclusion has appeared at renewal — wordings change year to year and new exclusions are easy to miss; how “professional services” or “the business” is defined, and whether AI-assisted work sits comfortably inside it; whether there are conditions about supervision or quality control that an ungoverned AI workflow might breach; and how the policy treats claims arising from breach of confidentiality, if client data has gone into external tools.
Where a wording is silent, that ambiguity itself is worth a conversation with your broker — some insurers will confirm their position, and some now offer express AI cover. What you should not do is assume silence means safety.
How Apex approaches it
We treat AI the way we treat every material change in a client’s risk: describe it accurately, present it well, and make sure the wording matches the reality. That means helping you put the governance story on paper, checking the current wording for gaps and new exclusions, and putting the right questions to insurers before renewal rather than after a claim. Firms that do this once find the next renewal straightforward — the story is written, and it usually reads well.
Frequently asked questions
Will using AI increase our PI premium?
Not automatically. Underwriters price the overall risk, and a firm using AI with clear governance and human review can present as well as — or better than — a firm with no stated position. What tends to cost money is the absence of a story: undisclosed use, no review process, no policy.
Do we need to tell our insurer which AI tools we use?
If asked, answer accurately — proposal-form answers form part of your presentation of the risk. Even unasked, material AI use in client work is the kind of fact the duty of fair presentation is designed to capture. A short, factual description of tools, uses and controls is usually all that is needed.
Our staff use AI informally without approval. Does that matter for insurance?
Yes, because your disclosure has to reflect what actually happens in the firm, not just official policy. Unofficial use is also where the riskiest behaviour tends to sit — public tools, client data, no review. An internal AI-use policy closes the gap and gives you something accurate to disclose.
Does our PI policy cover work produced with AI?
Most wordings neither confirm nor exclude it — they were written before generative AI. Cover then depends on general definitions and conditions, which is workable but ambiguous. The fix is to have the wording reviewed, raise the question with the insurer where it matters, and close any gap before a claim does it for you.
Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority (FRN 724952). This page is general information, not advice on a specific policy.