The Irony of the Decade

KPMG set out to define excellence in the age of agentic AI. Instead, the firm ended up providing a masterclass in the technology’s most persistent failure. Last week, the professional services giant quietly pulled its October 2025 report, “Redefining excellence in the age of agentic AI,” after multiple organizations named in the document publicly refuted its findings.

The report, which was intended to serve as a roadmap for corporate AI adoption, appears to have been authored with the very tools it sought to analyze. According to research group GPTZero, the document was riddled with AI hallucinations—errors that were not just minor typos, but complete fabrications regarding how major institutions deploy technology.

When the Data Doesn't Exist

The inaccuracies were not subtle. The report made specific, detailed claims about the AI strategies of UBS, the UK’s National Health Service, Swiss Federal Railways, and Transport for London. When contacted by the Financial Times, representatives from these organizations confirmed the claims were either entirely untrue or fundamentally misleading.

It is a striking failure for a firm that advises the world’s largest companies on digital transformation. The incident suggests that the report’s authors bypassed the most basic layer of human verification, trusting an AI model to synthesize data that it had effectively invented.

A Growing Pattern in Consulting

KPMG is not alone in this stumble. Just last month, EY withdrew its own report on loyalty rewards programs after it was discovered to contain fake footnotes and hallucinated data. These incidents highlight a dangerous trend: the pressure to produce thought leadership on AI is leading firms to use the very tools they are still learning to govern.

For professional services firms, the stakes are higher than just a retracted PDF. Their business model is built on the assumption of rigorous, verifiable expertise. When that expertise is outsourced to a black-box model without human oversight, the firm’s primary product—trust—begins to erode.

Key Takeaways

  • The Retraction: KPMG removed the report from its digital channels following reports of fabricated case studies involving major clients like the NHS and UBS.
  • The Root Cause: Analysis by GPTZero indicates the inaccuracies were the result of AI hallucinations, suggesting the report was generated with insufficient human verification.
  • Industry Trend: This is the second major consulting firm in two months to retract a report due to AI-generated errors, signaling a systemic issue with AI-assisted research workflows.

What This Means for Corporate AI Governance

KPMG’s spokesperson stated the firm is conducting an internal investigation and reiterated that employees must follow guidelines requiring human oversight. However, the damage is already done. The incident serves as a stark reminder that "agentic AI" is currently a tool for drafting, not for fact-finding.

For the consultants and analysts currently drafting their Q1 2026 outlooks, the lesson is clear: if the AI provides a specific data point about a client’s internal operations, it is not a fact until it is verified by a human source. The firm’s next major report release will be the true test of whether these internal guidelines are being enforced or if the pressure to publish remains the priority over accuracy.