“Don't use AI. Think for yourself.”
Fair enough—but who's checking the facts? Who decides what the work requires? And who signs off before an answer goes to a customer?
If none of that is clear, banning a chatbot won't magically create good judgment. Copying an AI answer without checking it is risky, too. Using your brain and refusing a tool aren't the same thing.
Faster writing doesn't mean better decisions
A 2023 experiment involved 453 professionals completing work-style writing tasks. Participants given access to ChatGPT finished, on average, 40% faster, and their work received quality scores 18% higher.
Those are meaningful results for the tasks tested. They don't mean every workplace will get the same gains. And they don't show that the people using AI became better at making important decisions.
Writing quickly and exercising judgment are two different skills.
The risk of switching your judgment off is real
In a 2025 survey of 319 knowledge workers, people who reported greater confidence in generative AI also tended to report less effort spent on critical thinking while using it.
That is a correlation in self-reported experience, not proof that AI caused anyone's thinking skills to decline.
Still, it highlights an everyday trap: an answer can be neatly written and entirely wrong. A polished paragraph isn't a checked fact, no matter how reassuring the formatting looks.
Decide who drafts, who verifies and who owns the call
There is a practical division of labor.
Let AI suggest a structure, draft plain-language text or surface questions you've missed. Have a person check sources, confirm company rules, decide what matters and take responsibility for the result.
“The AI said so” is not a job title. Nor is it an explanation to the person affected by the decision.
And don't feed confidential documents or personal information into an AI service without confirming that your workplace allows it and that the service is appropriate for the data.
Teaching judgment takes more than a ban
For training, someone could write a first draft unaided and compare it with an AI version. For routine documents, a reviewed AI draft may be the more sensible path. The right approach depends on the work and the risks.
There is no universal winner called “AI banned” or “AI everywhere.” If you want people to think critically, ask what they verified—not just which tools they used.
References (2)
- Noy & Zhang (2023), Experimental evidence on the productivity effects of generative artificial intelligence, Science. https://doi.org/10.1126/science.adh2586
- Lee et al. (2025), The Impact of Generative AI on Critical Thinking, CHI 2025, Microsoft Research. https://www.microsoft.com/en-us/research/publication/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers/
