There is a certain poetic justice in this: KPMG, one of the world's largest consulting firms, had to pull a report on how companies use artificial intelligence - after it turned out that the report was partly written by artificial intelligence, and made things up while doing so.

The document, with the glamorous title "Redefining Excellence in the Era of Agentic AI", was published in October 2025. The research group GPTZero found a string of inaccuracies and told the "Financial Times" they stemmed from so-called hallucinations - when the model confidently invents facts that do not exist. In other words, the firm that charges millions for advice on how to use artificial intelligence wisely appears to have used it to write a report on artificial intelligence - without checking what it had written.

The embarrassment grows once you see who complained. UBS, the British National Health Service, the Swiss and London railways - all said the claims in the report about their use of artificial intelligence were either inaccurate or misleading. KPMG pulled the document from its pages while it conducts its own investigation and reminded everyone that "all employees are expected to provide human oversight and verify sources". A nice sentence, a shame it came after the fact.

And they are not alone. Last month EY, another of the big four, pulled a report on loyalty programmes that contained fake footnotes and similar fabrications. When the firms that sell expert authority itself start delivering texts the machine hallucinated, a simple question arises: what exactly is the client paying the consultant for? The technology that was supposed to boost productivity is for now mostly exposing one thing - who really reads their own work, and who just signs it.