The number — 66% of AI users act on its answers without checking them, and 56% have already made a mistake because of it

Two-thirds of people who use AI act on its answers without checking their accuracy, and more than half admit they have already made a work mistake because of it. The danger isn’t that AI is sometimes wrong — it’s that a wrong answer gets used as-is, confidently, inside a quote, a plan or a decision. The smoother and faster the tool feels, the more the review step quietly disappears — and that is where the value leaks out. For a small or mid-sized company, the fix costs almost nothing: on every AI use that leads to a decision, add a named, time-stamped human verification step before acting, and track the error rate it catches in the first months.

Source: KPMG & University of Melbourne, “Trust, Attitudes and Use of AI: A Global Study 2025,” April 2025 (48,000+ respondents, 47 countries).