FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: Auditors
Auditors can use AI forecasting and stress-testing to challenge management's going concern case, but they need to understand and explain what the model shows. After a collapse, the going concern conclusion is often the first thing a claimant examines.
Part of: AI in Auditors
In short
AI tools can build independent cash flow forecasts, run many downside scenarios and compare management's assumptions with sector data. Used well, they sharpen the auditor's challenge. Used badly, a model nobody can explain gives false comfort. The FRC says the explainability needed depends on the use, and the going concern conclusion is a high-stakes use. If a company fails soon after an unmodified opinion, claims usually allege the auditor should have warned. Whether the auditor owed a duty for the loss claimed depends on Caparo and scope of duty principles, and PI usually responds subject to its terms.
Going concern work asks whether management's forecasts are reasonable and what happens if things go worse than planned. AI tools help in three ways: building an independent forecast from historic patterns, running large numbers of stress scenarios on revenue, margins and working capital, and comparing assumptions with sector benchmarks. Generative AI may also be used to summarise lender correspondence or minutes for signs of funding pressure.
The attraction is breadth. The risk is that a polished chart with many scenarios looks like rigorous challenge even when the inputs are management's own figures, or when the team cannot say why the model concluded headroom was adequate.
The FRC's June 2025 guidance defines explainability as "a measure of the extent to which the behaviour or decisions of a tool can be understood" and says the level required depends on the use. A going concern conclusion that a partner must defend needs a high level of it. The file should record the tool and model version, the data used and its source, why the tool was appropriate, and how its outputs informed judgements (ISA (UK) 230 paragraph 8).
The March 2026 guidance adds that faulty reasoning is a form of deficient output, and that human-in-the-loop review needs professional scepticism. ISA (UK) 220 keeps the conclusion with the engagement partner.
Illustrative scenario: an audit team uses a forecasting tool to stress-test a retailer's cash flows. The model is fed management's budget and runs downside cases on sales, all showing headroom. It does not model the expiry of a key lending facility shortly after the year end, because that sat in a loan agreement rather than the data. The opinion is unmodified with no material uncertainty disclosure. The retailer enters administration within the year, and the administrators allege the audit should have identified the funding risk.
Other failure points: stress tests that only flex one variable at a time, sector benchmarks that do not fit the client, and teams unable to explain the model's output when the partner asks.
Going concern claims tend to be large because they follow insolvency. An audit firm's PI policy would usually respond to the claim, subject to its terms, limit and excess, whether or not AI was involved. The key legal questions are who can sue and for what. Auditors' duty in negligence is generally owed to the company, not to every investor or lender, following Caparo, and the recoverable loss is shaped by scope of duty. A section 534 liability limitation agreement can limit liability to the company if properly made and approved. See audit firm PI cover.
Related: minute summaries, risk assessment and planning and auditing client AI systems.
Apex arranges PI for audit firms and can review whether your limit reflects the going concern exposure on your larger audits. See the AI in audit hub.
Yes, and it can improve challenge. The FRC expects the tool to be certified and the file to show how its outputs informed judgements. The team must be able to explain the outputs well enough for the engagement partner to rely on the conclusion.
The FRC says the level depends on the use. Because the going concern conclusion is a key judgement that may be challenged after a failure, auditors should be able to explain what drives the model's results and what it does not consider.
Usually the company, often acting through administrators or liquidators. Third parties such as lenders or investors generally need to show the auditor assumed a duty to them, following Caparo. This is fact-specific and applies whether or not AI was used.
Usually it would, subject to the policy terms, limit and excess, as with any audit negligence claim. The quality of the file record, including how the model was used and challenged, is likely to affect how well the claim can be defended.
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.
