- A KPMG report on Agentic AI has been found to contain numerous inaccuracies believed to be AI-generated.
- The report, titled Redefining Excellence in the Age of Agentic AI, was published in October 2025.
- The research group GPTZero discovered several non-existent examples of AI use within businesses.
- The Financial Times verified these errors with the organizations mentioned in the report.
- The report claimed that UBS uses AI Agents for investment advice, risk management, and compliance on a platform co-developed with Microsoft.
- UBS stated that this information is “not true.”
- KPMG also claimed that Swiss Federal Railways (SBB) deployed AI Agents to assist with journey planning, booking, and travel optimization.
- SBB confirmed this information is inaccurate.
- Another section alleged that Transport for London uses AI Agents to forecast congestion, personalize notifications, and coordinate multi-modal transport.
- Transport for London stated that this description is misleading.
- The report further claimed that NHS Greater Manchester uses AI Agents to predict readmissions, triage patients, and automate referrals.
- An NHS representative said the content does not reflect the original documents KPMG appeared to reference.
- Following the reports, KPMG removed the document from several websites for investigation. The company stated it always requires employees to perform human oversight and independent source verification when using AI.
- This is the latest in a series of AI-related incidents in the professional services industry.
- Previously, EY had to withdraw a report because it contained fake citations and references generated by AI.
- GPTZero warned that misleading reports from major consultancies could create a “secondary hallucination” effect, as false information continues to be cited by the press and other businesses.
📌 The KPMG incident creates a notable paradox: one of the world’s leading consulting firms for AI implementation was found using information seemingly fabricated by AI in a report about AI itself. The incident shows that even organizations specializing in AI governance consulting are struggling to verify AI-generated content. This serves as a continued warning that human oversight and independent source verification remain mandatory when using AI in research, consulting, and professional publishing.
