In an ironic turn of events that underscores the growing pains of the generative artificial intelligence era, professional services giant KPMG has been forced to retract a high-profile report titled "Redefining excellence in the age of agentic AI." The document, intended to serve as a roadmap for businesses navigating the integration of autonomous AI systems, has instead become a cautionary tale regarding the pitfalls of automated content generation and the erosion of editorial due diligence.
The retraction follows a wave of pushback from several high-profile organizations—including UBS, the UK’s National Health Service (NHS), Swiss Federal Railways, and Transport for London—all of which categorically denied the report’s characterizations of their specific AI adoption strategies. The controversy suggests that in its pursuit of technological authority, KPMG may have fallen victim to the very technology it sought to champion, employing AI-driven research methods that resulted in a series of sophisticated, yet entirely fabricated, claims.
The Anatomy of an Error: A Chronology
The timeline of this incident highlights a rapid descent from corporate thought-leadership to public embarrassment.
October 2025: The Launch
KPMG published "Redefining excellence in the age of agentic AI" in October 2025. Positioned as a flagship publication, the report aimed to provide C-suite executives with a landscape analysis of how global institutions were implementing "agentic" AI—AI systems capable of executing complex tasks with minimal human intervention. The report was marketed as an authoritative guide for digital transformation.
Late 2025: Detection by GPTZero
The inconsistencies were not flagged by traditional media watchdogs, but by GPTZero, a research group specializing in AI detection and linguistic analysis. Upon reviewing the report, analysts at GPTZero identified recurring patterns of "hallucinations"—a phenomenon where AI models confidently assert false information as fact. When these findings were cross-referenced with the public records of the companies mentioned, the discrepancies became impossible to ignore.
November 2025: Public Rebuttals and Retraction
Following inquiries from the Financial Times and other outlets, the organizations cited in the report began to issue formal denials. UBS and the NHS, among others, confirmed that the specific AI use cases attributed to them by KPMG were either pure fabrications or grossly distorted versions of their actual operational realities. Faced with mounting evidence of inaccuracy, KPMG pulled the report from its digital platforms, signaling an internal crisis of quality control.
A Catalogue of Inaccuracies: The Impact on Subject Organizations
The report’s claims were not merely minor errors; they were structural fabrications that misrepresented the technological posture of some of the world’s most significant institutions.
- UBS: The Swiss banking giant, which has been vocal about its controlled and conservative approach to AI, found itself described in the report as having deployed autonomous AI agents for tasks it had not yet begun to automate.
- The National Health Service (NHS): The report attributed specific diagnostic AI integration strategies to the NHS that the health service confirmed were non-existent, creating potential confusion regarding patient data and operational safety protocols.
- Swiss Federal Railways and Transport for London: Both transport authorities were cited as early adopters of agentic AI for complex scheduling and maintenance, claims both organizations characterized as misleading or entirely untrue.
The common denominator in these errors is the "hallucinatory" nature of the output. In the context of generative AI, a hallucination occurs when the model predicts the next likely word in a sentence based on statistical probability rather than factual truth. When tasked with synthesizing "research" on AI usage, the model likely aggregated existing industry trends and projected them onto specific, high-profile companies to fill perceived knowledge gaps in its prompt-generated response.
Official Responses and Internal Accountability
KPMG’s response to the controversy has been one of damage control, focusing on the failure of internal processes rather than the failure of the technology itself. A spokesperson for the firm stated: "We expect all our people to follow our guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources."
The firm has since launched an internal investigation to determine how such a report could pass through editorial and legal review processes without the inaccuracies being caught. This investigation likely aims to identify whether the errors were the result of a "black box" AI workflow—where a staff member relied too heavily on AI-generated research—or a breakdown in the firm’s standard fact-checking procedures.
Critics, however, argue that the issue is systemic. By leaning on AI to generate research about AI, the firm created a circular dependency where the machine was effectively evaluating its own hypothetical utility.
The Broader Context: A Trend of Institutional Failure
This incident is not an isolated event. It represents a worrying trend within the professional services sector, where the pressure to be perceived as an "AI-first" firm is clashing with the reality of current AI limitations.
Just one month prior, accounting and consulting powerhouse EY (Ernst & Young) was forced to withdraw its own report on loyalty rewards programs. In that instance, investigators found that the report contained "fake footnotes"—citations to non-existent academic papers and industry studies. The fact that two of the world’s "Big Four" accounting firms have been caught publishing AI-generated misinformation within weeks of each other suggests a widespread vulnerability in the industry’s knowledge-management pipelines.
Implications for the Future of Professional Services
The retraction of these reports has profound implications for the professional services industry, which trades primarily on the currency of "trust" and "accuracy."
1. The Erosion of "Expert" Status
For decades, firms like KPMG and EY have acted as the gold standard for industry intelligence. When these firms publish a report, it is often cited by policy makers, journalists, and corporate boards to justify massive capital investments. If the source material is revealed to be AI-generated fiction, it threatens the very foundation of the consulting business model.
2. The Legal and Reputational Risk
The potential for legal liability is significant. If a client were to rely on a KPMG report to make a multi-million dollar investment decision regarding AI, only to find the underlying data was a hallucination, the firm could face substantial litigation. Furthermore, the reputational damage—being exposed as an entity that cannot verify its own research—is difficult to quantify but potentially catastrophic.
3. The Revaluation of Human Oversight
The industry is now facing a mandatory pivot back toward "Human-in-the-Loop" (HITL) workflows. The consensus among technologists is that while AI is an excellent tool for summarizing and organizing existing data, it is currently ill-equipped to perform primary research or synthesize novel industry insights without rigorous human validation. The "KPMG incident" will likely lead to the implementation of mandatory, auditable "fact-check trails" for all AI-assisted publications within major firms.
4. A New Standard for AI Transparency
There is a growing call for firms to disclose when AI has been used in the generation of research reports. Just as financial statements are audited for accuracy, there is now a movement toward "AI audits," where the methodology of content generation—including the prompts used and the human-led verification steps taken—is made transparent to the public.
Conclusion: The "Hallucination" Trap
The paradox of the current AI boom is that those who stand to gain the most from its adoption are often the ones least equipped to handle its failures. KPMG’s attempt to project expertise in the field of agentic AI ended in a demonstration of the technology’s most dangerous flaw: the ability to sound profoundly authoritative while being entirely incorrect.
As organizations scramble to integrate AI into every facet of their operations, the lesson from this incident is clear: technology is a powerful accelerant, but it is a poor substitute for rigorous, human-led verification. Until professional services firms can reconcile the speed of AI generation with the slow, deliberate pace of institutional truth-seeking, they remain at risk of trading their hard-earned credibility for the convenience of an algorithm.
In the age of agentic AI, "excellence" is not defined by how quickly one can produce a report, but by the integrity of the data that informs it. For KPMG and its peers, the path forward requires a return to the foundational principles of auditing: verify everything, trust nothing, and remember that when it comes to the truth, there is no substitute for the human eye.
Laily UPN
Suro Senen