United KingdomProfessional services2025–2026KPMG International
AI-assisted report on enterprise adoption of agentic AI
KPMG withdrew an agentic-AI report after named organisations disputed its claims
Problem
KPMG distributed research about enterprise AI adoption through consulting groups in several countries, with local marketing contacts. The publication supported its marketing of AI advisory services.
Approach
KPMG published the report in October 2025; reporting later attributed its inaccurate case studies to apparent AI hallucinations. The Financial Times checked findings raised by GPTZero and obtained responses from organisations named in the report: UBS rejected claims about its AI agents, while Swiss Federal Railways and Transport for London also disputed descriptions of their systems. KPMG removed the report from websites in June 2026 while investigating its publication. The firm said its responsible-AI guidelines required human oversight to validate content and verify independent sources.
Results
No published figures
Takeaway
The published case studies did not survive checks with the organisations they described. KPMG's stated requirement for human verification had not prevented the disputed claims from reaching publication.
Categories
- Size
- Large enterprise
- Industry
- Professional services
- Business function
- Marketing and salesInternal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Press reporting
Sources
Source summaries
The Financial Times checked disputed case studies with the organisations concerned and reported KPMG's withdrawal and response. It described the report's distribution by consulting groups as marketing material for potential clients.
Finextra reported that UBS rejected the report's description of its agentic-AI deployment. It also reported KPMG's removal of the publication pending investigation and its statement about verification requirements.
The article dated the report's publication to October 2025 and reported its subsequent removal. It attributed the identification of apparent AI hallucinations to GPTZero and described KPMG's investigation.
Full details are in the original sources.