Professionals still rely too heavily on AI-generated answers, global study finds
2–3 minutes
A global study conducted by KPMG in partnership with the University of Melbourne surveyed over 48,000 people across 47 countries. Around 66% of professionals who use artificial intelligence at work admitted to trusting the tool’s answers without verifying their accuracy. Furthermore, 56% reported making professional mistakes due to the use of this technology.
Why verification takes a backseat
Generative AI tools produce text based on statistical patterns rather than fact-checking, which can lead to hallucinations. The fluency of the responses reduces the urge to review, and productivity pressures also play a role. Although 47% of respondents received AI training, many still use the technology without formal guidance on how to verify results.
Risks in practice
In daily work, errors manifest in tangible ways. An analyst might receive a contract summary that omits a critical clause, while a coordinator could present outdated market data in a meeting. At higher levels, an executive might approve an investment based on unverified projections. Additionally, 48% of participants admitted to inputting company data into public AI tools, exposing confidential information.
How to change the habit
Define what needs verification
Not all tasks require the same level of scrutiny. A draft email poses less risk than a number used in a financial presentation. Data, dates, values, and quotes should always be verified.
Create simple review routines
Ask the tool to provide its sources and review each one. Compare figures with the original source, such as official reports or internal systems. For critical decisions, have a second person review the material before it’s sent. A Freshworks survey found that 69% of professionals would trust AI more if human review were mandatory.
Formalize rules and training
Companies that establish usage policies, designate approved tools, and provide training reduce the likelihood of errors and data leaks. For those seeking advanced education, long-term programs like an MBA in Artificial Intelligence focused on business are available.
What to watch for going forward
The data shows that the adoption of artificial intelligence has outpaced the development of verification practices. Checking sources, numbers, and assumptions, defining review levels by task type, and training teams are measures that reduce risks without compromising productivity.
For professionals, the key takeaway is to treat AI-generated responses as a first draft that requires validation before becoming a report, presentation, or decision.
Saint Paul offers a Pre-MBA in Artificial Intelligence with four online practical classes for R$ 37.
