Home / News

AI agents out-questioned KPMG auditors in a trial on a FTSE 250 company's accounts

KPMG's audit trial shows that AI agents can now review financial documents and raise questions a busy team would miss, as long as a human checks every conclusion. For a small business, the practical move is to give an agent one review task, such as matching invoices against contracts, and keep an accountant as the final sign-off.

According to European Business Magazine, KPMG UK ran more than 100 AI agents on the accounts of an unnamed FTSE 250 company, alongside its regular audit team. The agents collected and documented data, drafted conclusions and suggested next steps. In several cases they went deeper than the human auditors did. The system took about four months to build, and its findings were not used for the official audit.

What a small business can take from this

You do not need 100 agents. The pattern that worked is narrow: each agent checks documents against rules and flags what looks off. I use the same approach in a legal archive audit assistant I built. Claude reads hundreds of files and surfaces contradictions, and a lawyer decides what matters. The weak spot is the input data. KPMG calls messy data a major obstacle, and one agent failed to match regulatory requirements to the documentation until an auditor stepped in. If your invoices live in scanned PDFs and three different Excel files, clean that up first.

Why it matters: agents are most useful in document review as a tireless question-asker, not as the one who signs off. Start with a single process where a missed error is cheap to catch and easy to verify.

Source: Hacker News