The Slow Death of Figuring Things Out For Yourself

There was a time when staring at a blank Word document was just part of life.  

You opened the file, hated every sentence you wrote, rewrote it anyway, and somewhere between the third bad draft and the final one you learned something about the work and about yourself.  

Today, that same moment feels optional.  

If you are a student, you can paste the question into a chatbot and get a “good enough” draft in a few seconds. If you are early in your career, you can outsource the first attempt at an email, memo, or code snippet to an AI tool before you have ever really tried to do it alone.  

On the surface, this looks like great progress for sure.  

You save time, avoid frustration, and get something polished that might have taken you much longer on your own. But underneath that convenience, the habit of figuring things out for yourself is starting to look unnecessary, even inefficient.  

You can see this tension clearly in education.  

Pearson has reported that its own AI tutoring tools seem to improve grades, while generic chatbots used without structure do not reliably help students learn better (Reuters, 2026).

At the same time, researchers and professional bodies are starting to talk about a “judgment gap,” which is the risk that law, business, and other professional graduates become fluent at using AI while being less confident at forming their own conclusions without it (Thomson Reuters, 2026).

Work is not very different. Companies across sectors are already cutting roles while shifting investment toward AI, often starting with the junior tasks that used to teach people how the work actually functions (Reuters, 2026). The tasks such as first-pass research, basic drafting, simple data cleanup were boring, but they were also how you learned the difference between surface-level output and actual understanding.  

The old, painful process of figuring things out did something AI cannot easily do for you. It forced you to sit with confusion long enough to build an internal sense of what makes an answer good, not just what makes it look finished. When you always ask the machine first, you still get a result, but you do not build that inner map.  

This is not an argument for going back to the stone age of education or work. Used well, AI can be a scaffold for struggle rather than a replacement for it; it can be a way to get feedback, see alternatives, or check your reasoning, not just a way to skip the reasoning entirely.

The problem is that most systems, products, and incentives are pushing us toward removal of friction, not toward better-designed friction.  

So the question is not “Will AI make us stupid?”. We should ask, instead, what kind of people do we become if we rarely have to push through the part of the task where we do not yet know what we are doing?  

There is no simple policy lever for that.  

It is a series of small choices. Think of a teacher who allows AI but demands a reasoning log, a manager who cares how you got to the answer, not just that it looks clean, or the student or worker who chooses to try first and ask later, not the other way around.  

AI is very good at producing answers. The more we rely on it, the more we will have to deliberately protect the slower, less glamorous habit that used to build judgment, which is figuring things out the hard way, at least sometimes, when nobody is watching.