Generative AI tools — large language models, drafting assistants, expert systems — have moved from novelty to daily workflow in many professional services firms. Solicitors draft first-cut contracts through AI assistants. Accountants run analytical models. Surveyors use AI-generated valuation comparables. Financial advisers explore AI research aids. The productivity gains are real. The professional liability position is unchanged: the professional remains liable for the advice given, regardless of the tools used to produce it.
A client engages a professional under a contract of retainer. Under the Supply of Goods and Services Act 1982 s.13 the professional must exercise reasonable skill and care. Where advice is given, the common law standard from Bolam v Friern Hospital Management Committee [1957] 1 WLR 583 applies — the professional is judged against the standard of a reasonably competent practitioner in the field. That duty is owed by the professional to the client. It cannot be delegated to a software tool.
A solicitor who accepts AI-generated drafting without review, an accountant who signs off AI-produced calculations without checking, or an adviser who repeats AI-generated research without verification, remains on the hook if the advice turns out to be wrong. The tool is not a defendant. The professional is.
The Bolam test does not ask whether the AI was reasonable. It asks whether the professional's use of the AI was reasonable — a question of process. Reasonable use, in the current state of the market, generally involves treating AI output as a draft to be checked rather than a finished work-product. It includes an awareness of the tool's known limitations, verification of any material factual claim, and a documented review before the output is passed to a client.
Large language models are known to generate plausible-sounding content that is factually wrong — fabricated case citations, invented statutory provisions, subtly incorrect clause wording. This is not an edge case; it is a documented feature of the technology. A professional who relies on AI output without independent verification carries the risk of that hallucination reaching the client. The ICO's guidance on AI and automated decision-making reinforces that professionals accountable for a decision must be able to explain and justify it.
The SRA has published guidance on the use of AI in legal services, focusing on competence, confidentiality and client communication. The ICAEW has published guidance for accountants on AI use in professional work, emphasising the professional's responsibility for the output. The FCA has run thematic reviews on AI in insurance and financial services, and has been clear that firms remain responsible for outcomes of any AI-supported process. AI is a tool the professional uses; it is not a substitute for professional judgement.
A PI claim arising from AI-assisted advice is treated as any other negligence claim. The trigger is the professional's breach of duty — usually a failure to check, verify, or exercise judgement over AI output — and the resulting loss to the client. Cover responds subject to the usual policy terms, aggregation rules and any profession-specific minimum terms. Under the Insurance Act 2015 the firm's duty of fair presentation applies at inception and renewal, so material use of AI in delivering advice may need disclosure where the proposal form asks.
A solicitor uses an AI drafting tool in 2025 to prepare a commercial contract. The AI-produced draft includes a cited case authority that does not exist, and an indemnification clause with a subtly incorrect scope that the solicitor does not spot on review. The client relies on the contract, a dispute arises, and the loss is quantified at £200,000. The solicitor's PI insurer treats this as a standard Bolam analysis — did the solicitor exercise reasonable skill and care? The failure is not the AI's; it is the solicitor's failure to check. Cover under the SRA Minimum Terms and Conditions responds, subject to aggregation.
A written policy on AI use, kept current. Verification protocols — no AI-produced citation, calculation or clause reaches a client without an identified human check. Documentation of the check, so the audit trail exists if a claim later arises. Client-facing disclosure where AI has been used materially in delivering advice, in line with the firm's communication and transparency obligations. Training for anyone using the tools, so the limitations are understood.
Some PI insurers have begun asking about AI use in proposal forms — the tools used, the volume of AI-assisted work, and the firm's internal controls. Others have not yet reached that point. Firms should expect the questions to become standard at renewal in the near term. Answering them fully is part of the Insurance Act duty; understating AI use to smooth a renewal creates a materiality problem down the line.
Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Firm reference number 724952. This entry is general information, not advice on any particular policy.