KPMG: AI agents in a trial sometimes asked better questions than its own auditors
KPMG used more than 100 AI agents to audit the accounts of a FTSE 250 company as a trial alongside a conventional audit. The official opinion rested on the human audit, and the claim that the agents asked better questions is KPMG’s own.
Cath Burnet, KPMG’s UK head of audit, made the claim to the Financial Times, and the Irish Times published the FT’s story. In several cases the AI was able to go ‘so much deeper’ than the regular audit team and, in some situations, asked ‘better questions’, Burnet said.
In the trial, the agents gathered, analysed and documented material and drafted conclusions. They collaborated with one another autonomously and proposed next steps. The FT reported that AI had not previously been used at every stage of a major UK audit.
A team built the agents in four months, and a KPMG audit partner signed off the experimental audit. KPMG did not, however, rely on it when giving its opinion on the company’s financial statements. The claim that the agents went deeper in some cases and sometimes asked better questions is KPMG’s own, not an independent comparison. The FT story does not say on what basis the questions were compared.
In one part of the audit, a human had to intervene when an agent failed to match a set of regulations to the company’s documents. Staff set the agents’ goals, gave them information about the company and earlier audits, checked their work from time to time and challenged their conclusions.
The UK audit regulator, the Financial Reporting Council (FRC), published guidance on 16 November, accountant.nl reported. The guidance says AI can pose a ‘clear threat’ to an auditor’s integrity and independence if the agents’ findings are used to advise the client. The FRC attributes the risk to the agents’ ability to review far more transactions than a sample-based audit and to find, for example, cost savings that could be used to sell consulting services, The Decoder wrote. ‘Whatever technology a firm uses, the auditor remains accountable,’ said the FRC’s Mark Babington. Babington said auditors must be able to explain their use of AI, challenge its results and answer for the evidence.
KPMG sees ‘clear time-saving opportunities’ but did not say whether audit fees would fall. The firm plans to extend the trial to up to 20 larger audits and hopes eventually to offer AI-based audits to its clients. Burnet said the technology is not yet ready for thousands of audits, and few companies have data clean enough for the agents.