Anthropic released a very interesting article and white paper recently. They analyzed 1.5 million AI conversations and found AI tools can sometimes undermine human judgment through helpfulness, not deception.
When users rely too heavily on AI to navigate significant decisions they can end up with less accurate beliefs, shifted priorities, and actions they later regret — a pattern the researchers call disempowerment. Critically, users tend to rate these interactions positively in the moment; the regret only comes up later.
For businesses, the implications could be catastrophic. If employees (or worse, management) delegates judgment to AI systems then those systems are no longer just “thinking partners” – they are doing the thinking for you.
AI tools are wonderfully helpful if used properly. But business leaders need to be alert for the office equivalent of the “disempowerment” Anthropic highlighted. Is an AI making vendor decisions for you, deciding on personnel matters, or telling your accountant how to prepare the financials? Are AI-drafted messages and contracts being sent as-is?
None of this argues against using AI — I use them all the time. But it does highlight that organizations need policies for when human judgment is non-negotiable. Read the full Anthropic research to learn more.