A landmark 2025 KPMG study involving 48,000 respondents across 47 countries exposes a critical tension in artificial intelligence adoption. While 73% of users report personal efficiency gains from AI, a staggering 56% admit to making errors by relying on unverified AI outputs.
The industry term for these fabricated facts is "hallucination." This phenomenon is no longer theoretical. In May 2026, EY was forced to retract a cybersecurity report after discovering that 16 of its 27 citations were entirely fabricated by AI. Similarly, Deloitte agreed to refund portions of a government contract following the discovery of AI-generated falsehoods in its deliverables.
Conducted with the University of Melbourne, the data confirms that this risk profile is consistent globally, from Tokyo to Toronto. Professional services firms are now heavily investing in AI governance frameworks. The strategy involves balancing productivity gains with rigorous human oversight to intercept hallucinations before they reach clients.
For investors, these findings serve as a stark reality check. The high benefit rate drives continued adoption, but the significant error rate introduces material risk for any business model treating AI output as inherently reliable without verification.