Your team is using AI. They've stopped thinking.
The real AI problem isn't hallucinations. It's your team's judgement.
The pattern is everywhere now. A marketer gets a brief, opens ChatGPT, pastes it in, and sends the output to their manager with light edits. An ops analyst asks Claude to summarise a supplier report and presents the summary in a meeting without reading the original. A writer submits a first draft that’s technically clean, completely bland, and doesn’t contain a single opinion they actually hold.
None of them thinks they’re doing anything wrong. They got the job done. They hit the deadline. The output looks fine.
That’s the problem.
The research confirms what you’re already seeing
In early 2025, Microsoft Research published findings from a study of 319 knowledge workers. Participants shared 936 real-world AI use cases and reflected on how it changed their thinking and mental effort.
The headline finding was uncomfortable: knowledge workers who trusted AI more tended to believe it reduced the mental effort required for critical thinking tasks. Confidence in AI reliability influenced not just how often people used it, but how they perceived their own cognitive abilities.
The better they thought the AI was, the less they thought for themselves. And they didn’t notice they were doing it.
A separate study, published the same year, surveyed 666 people across age groups and found a significant negative correlation between frequent AI use and critical thinking ability, driven by cognitive offloading. That’s the technical term for what your team is doing when they paste a brief into a chatbot and ship the output. They’re moving the thinking outside their own heads.
Do it long enough, and the thinking muscle weakens. Researcher Lisanne Bainbridge identified this irony in the context of factory automation decades ago: by mechanising routine tasks and leaving exception-handling to the human, you deprive that person of the routine practice that builds judgement. The muscle atrophies from disuse. She was writing about assembly lines. It maps perfectly onto a marketing team in 2025.
What it actually looks like
The early signs are easy to miss. Work gets faster but blander. Outputs are technically correct but weirdly hollow. People stop pushing back in reviews because they didn’t form a view in the first place.
Ask someone why they structured a report a certain way, and they can’t tell you. Ask what they’d do differently, and they reach for the AI to answer that too.
The dependency builds fast. Six months of using AI to draft every piece of analysis doesn’t just make someone slower at analysis. It makes them worse at knowing when the analysis is wrong. That’s the real problem. Not the outputs being bad today. The judgement being absent tomorrow.
Why most managers are making it worse
Most teams have one of two policies on AI use: “go ahead, use whatever helps”, or a vague “make sure you check it.” Neither of those is a thinking policy. They’re output policies. They say nothing about the cognitive process that should happen before, during, and after the AI gets involved.
When managers introduced AI to save time, they accidentally removed the practice reps that built skill. Every time someone lets AI draft the first version of something, they skip the part of the job that builds their ability to judge whether the draft is any good.
Telling people to “think critically” doesn’t fix this. That’s not an instruction. It’s a wish. You fix it with deliberate practice. Specifically, with one exercise that takes 45 minutes, costs nothing, and works.
What follows is the full exercise, the facilitation guide, and the debrief questions that generate the most useful discussion.