Descartes wrote, “I think, therefore I am.”
If he were writing today, he might need to add a warning:
AI thinks, therefore I choose.
That is starting to describe how a lot of people use AI.
Ask for options. Get options. Pick one.
Ask for a recommendation. Get a recommendation. Accept it.
Ask what to say. Copy, edit, send.
It feels efficient because it is efficient. But there is a problem hiding inside that efficiency.
Selecting an answer is not the same as thinking.
AI is very good at removing friction. That is one of its biggest strengths. It can summarize, organize, compare, draft, search, structure, and explain faster than most of us can.
But not all friction is waste.
Sometimes the useful part is the struggle.
Working through competing ideas. Realizing your first assumption was weak. Defending a position. Changing your mind. Noticing that something sounds convincing but does not actually hold up.
That is where judgment gets built.
The danger is not that AI will think.
The danger is that we will stop thinking because AI can.
This is where I think Learning and Development needs to pay much more attention.
Right now, a lot of organizations are focused on helping employees use AI better. Better prompts. Better tools. Faster workflows. More productivity.
All useful.
But there is another question underneath all of that:
Are people becoming better thinkers with AI, or just faster users of it?
Those are not the same thing.
An employee can ask AI for five solutions to a problem, choose one, and look productive.
Another employee can start with a point of view, explain the assumptions behind it, ask AI to challenge the logic, look for missing evidence, test the idea from different stakeholder perspectives, and then revise the decision.
Both used AI.
Only one really used it to strengthen judgment.
That distinction is going to matter more.
As AI makes information, synthesis, drafting, and recommendation cheap, the ability to judge the quality of an answer becomes more valuable.
Can the employee tell when the recommendation is too generic?
Can they spot the assumption that should not be there?
Can they explain why the answer does not fit the context?
Can they reject something that sounds polished but is still wrong?
Can they defend the final decision without saying, “That is what AI suggested”?
Those are not prompt-writing skills.
They are thinking skills.
So maybe the goal should not be to make AI do more of the thinking.
Maybe the goal should be to use AI to create better resistance.
Instead of asking:
“What should I do?”
Try:
“Here is what I think I should do. What am I missing?”
Instead of:
“Which option is best?”
Try:
“What would have to be true for my preferred option to be wrong?”
Instead of:
“Give me the answer.”
Try:
“Challenge my reasoning before you give me yours.”
That is a very different way to use AI.
It keeps the human in the work.
And that may be one of the most important shifts L&D has to make over the next few years.
We should absolutely use AI to remove low-value effort.
But we should be much more careful about removing the effort that builds judgment.
Don’t outsource your thinking. Outsource the friction around thinking.
Or even more simply:
Stop using AI to think for you. Use it to make you think harder.
