KPMG UK tested more than 100 AI agents in a full parallel audit of the financial statements of a FTSE 250 company, whose name was not disclosed. The experiment, internally codenamed Project Exponential, was revealed on Wednesday (18/11) by the Financial Times and is, according to the firm, the first time AI agents have followed every stage of a real audit, working alongside the conventional team.
What the agents did
According to Cath Burnet, KPMG’s UK head of audit, the AI agents went further than the human team in several cases: they investigated issues in greater depth and asked “better questions” than the auditors. The agents collected, analysed and documented data on their own, drafted preliminary conclusions and coordinated work among themselves, dividing responsibilities throughout the investigation. KPMG says the system records every interaction between the agents and the documents, with full visibility and traceability.
The limits of the experiment
The trial also exposed a clear limit: a human auditor had to step in when an agent failed to correctly match a set of regulations to the company’s documents. KPMG made clear that the experiment did not replace the official audit — the formal opinion came from the human team, and the agents’ findings were not used to approve the company’s accounts.
What comes next
The firm plans to expand the project to as many as 20 larger audits, but acknowledges that the technology is not yet ready for thousands of engagements — and that the quality of companies’ data remains an obstacle. This week, the UK’s Financial Reporting Council (FRC) published guidance applying existing independence rules to the use of AI in audits.
As an editorial reading, the case goes beyond a productivity test: it places AI agents at the centre of the analytical work of one of the professions most dependent on human judgement. If the experience repeats at scale, the traditional audit model — large teams of junior professionals reviewing documents under partner supervision — could change structurally, with humans focused on exceptions, complex judgements and final responsibility for the opinion.
Image: KPMG logo, via Wikimedia Commons (public domain; original source: kpmg.com).
Sources: European Business Magazine; The Decoder; Business Standard; ClayoCraft. Original reporting by the Financial Times, published on 18 November 2026.
Transparency: this article was produced with the assistance of artificial intelligence, based on public sources on the internet listed above. We recommend consulting the original sources to confirm the facts.
By Muse