Aipedia: Auditors
Auditors can use generative AI to summarise board minutes for risk assessment, but the FRC warns against treating a summary as a complete record. Minutes are where related-party deals, disputes and funding worries first appear.
Part of: AI in Auditors
In short
Summarising board and committee minutes is one of the two worked examples in the FRC's March 2026 generative and agentic AI guidance. A generative AI tool reads a year of minutes and produces a short summary for the audit team. The risk is that the team reads the summary instead of the minutes and misses a related-party transaction, a litigation threat or a funding concern that never made it into the summary. The FRC calls this misuse of output. If a missed signal leads to an audit failure, a negligence claim will ask whether reading the source would have caught it.
Reading board, audit committee and subsidiary minutes is a basic risk assessment step. For a group with many entities, that can mean hundreds of pages. A generative AI tool can produce a summary of decisions, risks and actions in minutes, and can be asked targeted questions such as whether any director declared an interest.
The FRC uses this as a worked example in its March 2026 guidance, and lists minute summaries among the uses it sees in practice. Its factsheet says auditors should document on file that a GenAI-enabled tool was used.
The FRC's three risks apply directly. Deficient output: a summary can omit or distort, or simply hallucinate a decision. Misuse of output: treating a summary as a complete account of the minutes, or using a general summarising tool to answer a specific related-party question it was not designed for. Non-compliant methodology: a risk assessment that rests on the summary alone may not meet the standards.
The mitigation the FRC stresses here is human-in-the-loop review, where professional scepticism and an understanding of automation bias are important. ISA (UK) 220 keeps responsibility with the engagement partner, and ISA (UK) 230 paragraph 8 requires the file to show how the tool's output informed the team's judgements.
Illustrative scenario: an audit senior uploads a year of subsidiary minutes to an approved tool and asks for a summary of key decisions. The summary covers capital spending and trading but omits a short item in which a director disclosed an interest in a supplier and the board approved a large prepayment to it. Nobody reads the full minutes, because the summary looked thorough and the deadline was close. The related-party disclosure is missing from the accounts, the supplier later fails, and the prepayment is lost. Shareholders replace the board, and the company pursues the auditor.
Similar gaps appear with going concern: a summary that reports 'refinancing discussed' without the board's concern that terms might not be agreed. See AI in going concern assessments.
A claim that an audit missed a related-party transaction or a going concern signal would usually be dealt with under the firm's PI policy in the normal way, subject to its terms and excess. The fact that a summary tool was used is unlikely to change whether cover responds, but it may make the claim harder to defend if the minutes clearly recorded the point and the file shows only the summary was read. Whether the auditor owed a duty for the particular loss is a separate question; see Caparo for auditors and audit firm PI cover.
Related: AI contract review and AI in risk assessment and planning.
Apex arranges PI for audit firms and can talk through how your policy and your AI review procedures fit together if a missed-disclosure claim arises. See the AI in audit hub.
As a starting point, yes, but not as a substitute for the minutes. The FRC treats reading a summary as complete as a misuse risk. Teams should read the source, or at least check targeted questions against it, before concluding there is nothing relevant.
That a GenAI-enabled tool was used, its version, which minutes were provided, how completeness was checked, and how the team reviewed the summary against the source. The aim is that a reviewer can see what the team actually read.
The FRC uses minute summaries as a worked example of how generative AI can be used with appropriate mitigation, not as a blanket approval. Its point is that the four mitigation categories, especially human review, must give appropriate confidence in the output for that use.
It can be. The FRC notes the cyber risk when audited entity data is put into cloud-based AI applications but treats it as outside its guidance. Firms must protect confidential information under the ethics Code, so use only firm-approved tools.
Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.
Start a PI quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.