The 'Hallucination' Scandal: KPMG Retracts Major Agentic AI Report After Fact-Check Backlash

Global professional services giant KPMG has been forced to pull a high-profile report after featured companies exposed its case studies as entirely fabricated.

For decades, the "Big Four" have been the gold standard for corporate research, providing the data that boards and governments rely on to set their strategies. But this week, that reputation suffered a jarring blow. KPMG was forced to scrub its website of a flagship document titled "Redefining excellence in the age of agentic AI" after the very companies it profiled publicly disowned the findings.

The report, which promised to outline how major organisations were successfully deploying advanced "AI agents" to manage complex tasks, turned out to be a collage of fiction. From the Swiss bank UBS to the UK’s National Health Service (NHS) and Transport for London (TfL), the institutions named in the text were quick to push back. They didn’t just disagree with the findings; they stated the claims were factually incorrect and, in some cases, completely untethered from their actual operations.

When the Machines Write the Truth

How did a top-tier firm get it so wrong? The consensus among tech researchers, including the team at GPTZero who first flagged the errors, is that KPMG relied on the very technology it was supposed to be analysing. It appears the report was drafted using generative models that "hallucinated"—confidently inventing success stories where none existed.

The specific inaccuracies were glaring. The report claimed UBS had integrated AI agents into its investment advisory systems using a custom Microsoft platform. A UBS spokesperson dismissed this as entirely inaccurate. Similarly, the NHS Greater Manchester clarified that the claims about them using AI to automate patient data and predict readmissions simply did not align with their actual workflows. TfL similarly labelled the claims regarding their use of AI for congestion coordination as misleading.

The Bigger Picture: Why It Matters

This incident is not an isolated glitch; it is a symptom of an industry racing to appear "AI-ready" at the cost of rigour. When professional services firms—whose primary product is trust and verification—start outsourcing their research to the same black-box models they are selling to clients, the circle of accountability breaks.

We are witnessing a "poisoning the well" effect. When global consultancies fail to audit their own publications for basic factual accuracy, they undermine the credibility of the entire digital transformation narrative. Coming so soon after a similar retraction from rival firm EY, which also fell foul of AI-generated errors in its research, this suggests a systemic failure in quality control. For businesses, the lesson is stark: in the rush to adopt agentic AI, even the giants are forgetting to switch on their critical thinking.