KPMG quietly removed its agentic AI flagship report after fabricated case studies were exposed. UBS, the NHS, Swiss Federal Railways and Transport for London all denied the AI use cases described in the document. The scandal lands right in the middle of the consulting industry’s AI sales push.
Key Takeaways
- The KPMG report “Redefining excellence in the age of agentic AI” has been withdrawn
- Four named organizations publicly denied the AI use cases attributed to them
- GPTZero detected the fabrications, the Financial Times verified them
The flagship report pulled in silence
KPMG has quietly pulled “Redefining excellence in the age of agentic AI“, a flagship report published in October 2025. The document was used as a sales pitch to convince enterprise clients to adopt agentic AI systems. It leaned heavily on case studies meant to prove the maturity of the field.
Several of these case studies do not exist. The report described AI deployments at four clearly identifiable organizations. All four publicly disputed the use cases or called them misleading.
The list is heavy. UBS, the UK National Health Service, Swiss Federal Railways and Transport for London. These are not unknown startups. They are listed banks, national public institutions, and infrastructure operators that move millions of users every day.
KPMG confirmed the withdrawal through a spokesperson. According to the statement, the firm expects all staff to follow internal guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources. An internal investigation is underway.
When AI traps the consultants selling AI
The mechanics of the fabrications are now documented. Research group GPTZero spotted the inaccuracies first and attributed them to AI hallucinations. The Financial Times then verified the chain, organization by organization, before publishing.
The KPMG episode is not isolated. The previous month, EY had to retract a report on loyalty programs for strictly comparable reasons. Footnotes made up from scratch, citations pointing to sources that never covered the topic, data presented as coming from surveys that never existed.
Two Big Four firms tripping on the same pattern within weeks raises a structural question. Consulting houses billing for AI expertise are themselves using AI to draft their flagship publications without enough oversight. The output ends up published under the firm’s official logo, dressed as research.
The damage is amplified by how consulting reports travel. These documents get cited by other reports, quoted in trade press, referenced in corporate strategy decks. A hallucination slipped into a KPMG PDF becomes a secondary source in ten other documents within months.
The whole ecosystem is exposed to this pollution risk. The enterprise AI market relies heavily on adoption numbers published by the Big Four, the same way it relied on the gap between public sentiment and the figures pushed by firms in a recent Anthropic survey on AI and jobs. The credibility of that whole information layer just took a serious hit.
Also on Horizon:
- Meta AI Gateway: Bosworth cuts off internal tokenmaxxing
- OpenAI probe: state attorneys general turn up the heat
- Manus deal: Meta unwinds $2B under Beijing’s order
What this changes for buyers and for the consulting market
In the short term, procurement teams and strategy departments that cited the KPMG report internally will have to rework their materials. Business cases built on these fabricated studies are contaminated. Any investment decision validated at committee with this document as a reference is now exposed.
KPMG will have to publish a corrective document or a detailed erratum, or risk seeing wronged clients escalate. The four named organizations can also consider legal follow-up, especially the listed bank and the publicly funded ones. Unauthorized use of an institution’s name in a commercial report stays a serious legal matter.
In the medium term, Big Four firms will need to rebuild the production line for their publications. A workflow where AI drafts, a junior skims, and a partner signs no longer holds. Internal risk teams will impose systematic double checks on every figure, every quote, every case study. That will slow the publishing cadence.
The commercial cost is real. KPMG, like EY before, sells agentic AI as a productivity rupture for its clients. When the firm itself stumbles on the quality of its own AI-assisted output, the sales pitch loses force. Competitors not hit by this kind of scandal, including AI-native consultancies, will capitalize on the gap.
For the rest of the market, the episode is a caution signal on every publication produced between 2024 and 2026. The benchmarks, the adoption numbers, the agentic case studies that have circulated for eighteen months must now be read with sharper scrutiny. Source traceability is back as a real commercial selection criterion.
Follow the story on Horizon.
