In a startling turn of events that highlights the growing pains of our transition into an AI-driven world, KPMG has been forced to retract a major global report titled Redefining excellence in the age of agentic AI. The document, which was intended to serve as a high-level authority on how artificial intelligence is transforming industry and public services, was found to be riddled with factual inaccuracies. These errors were not merely simple typos or minor misinterpretations; they were sophisticated fabrications that appear to be the results of “AI hallucinations”—instances where artificial intelligence systems confidently present false information as absolute truth. By relying on automated tools to generate high-stakes content without rigorous human oversight, one of the world’s most reputable consulting firms found itself spreading misinformation about its own high-profile clients.
The errors were initially brought to light through the work of GPTZero, a research group dedicated to identifying AI-generated text. Their analysis flagged several passages as inherently suspicious, a finding that was subsequently verified by the Financial Times. The report had made bold assertions about how various major institutions, including UBS, Swiss Federal Railways, Transport for London (TfL), and NHS Greater Manchester, were deploying complex AI agents to manage their operations. Yet, when those specific entities were approached for comment, they categorically denied the claims. The fallout has been significant, not just for KPMG, but for the credibility of the “Big Four” consulting firms in an era where trust is the most valuable currency in business.
The breadth of these false claims is particularly concerning because they painted a picture of a technological utopia that simply does not exist yet. For instance, the report claimed that the wealth management giant UBS had integrated AI agents across its entire investment advisory and compliance sectors in collaboration with Microsoft. UBS felt compelled to publicly state that these claims were “factually incorrect,” effectively distancing themselves from a narrative that could have misled investors and competitors alike. Similarly, Swiss Federal Railways was described as using AI to orchestrate holistic, multi-modal travel journeys, a role they officially labeled as inaccurate. These companies rely on their reputations for precision and stability; being mischaracterized as “AI-first” innovators for technology they haven’t actually implemented is a significant breach of professional expectations.
Perhaps most unsettling was the misinformation regarding public services. The report asserted that NHS Greater Manchester was using AI agents to predict patient readmissions, handle triage, and automate medical referrals. In a sector where clinical accuracy is a matter of life and death, such a claim is not just an error—it is a dangerous distortion of the current reality. A spokesperson for the NHS body noted that the report’s claims did not align with actual operational practices, highlighting a massive disconnect between the AI’s generated prose and the complex, human-led workflows of the health service. When Transport for London also flagged the description of their congestion management systems as “misleading,” it became clear that the report was an embarrassing tapestry of fiction.
The criticism from figures like GPTZero CEO Edward Tian strikes at the heart of the modern information crisis: he warned that when institutions as influential as the Big Four produce and distribute error-filled, AI-generated reports, they “poison the well of information.” This is an existential concern for the business world. Consulting firms are paid a premium specifically because they act as filters, curating raw data into actionable, accurate, and reliable intelligence. If these firms are outsourcing their foundational research to hallucinating bots, they are delegating their core value proposition to software that mimics intelligence without possessing any understanding of truth. This case suggests that the pressure to be seen as “AI-native” might be causing these venerable firms to skip the crucial “human-in-the-loop” step that keeps reality grounded.
Ultimately, KPMG has pulled the report from its digital presence to conduct a full review of its production process. Their statement, affirming that they take “the accuracy and integrity of its published content seriously,” is a necessary step back from the brink, but coming on the heels of a similar scandal at EY—where another firm had to retract research over fabricated references and data—the industry has clearly hit a tipping point. These incidents serve as a cautionary tale for the entire corporate world: while AI is an incredibly powerful tool for summarization and drafting, it is an unreliable scribe for factual reporting. As we navigate the hype surrounding generative technology, it is clearer than ever that there is no substitute for human expertise, rigorous cross-referencing, and the accountability that only a living person can provide. Trust, once lost to an algorithm’s fabrication, is an incredibly difficult thing to repair.