What the evidence actually shows
Cognitive offloading is real and well-documented. When an external tool reliably does a mental task, the brain stops maintaining the circuitry for that task. GPS navigation is the clearest example: people who use it regularly show measurably reduced spatial reasoning and perform worse on navigation tasks without it. The skill atrophies because the practice stops.
The same mechanism applies to analytical thinking. A 2023 study from MIT found that knowledge workers who relied heavily on AI writing tools produced text rated as more polished but showed reduced ability to construct arguments independently when AI assistance was removed. The tool did not augment their thinking. It replaced it.
This is not a prediction about the future. It is a description of how human cognition has always worked. Skills you do not practice, you lose. AI is simply a very convenient reason to stop practising.
What this means for knowledge workers specifically
Most knowledge work depends on three things: forming a clear view, expressing it precisely, and evaluating the views of others. AI is now capable of doing rough versions of all three. The professional risk is not that AI will take your job. It is that you will quietly outsource the parts of your job that made you good at it.
The pattern is subtle. You ask an AI to summarise a report, then you stop reading reports closely. You ask it to draft a memo, then you lose the habit of thinking through an argument before writing. You ask it to generate options, then you become less able to generate options yourself. None of these feel like losses at the time.
Lawyerss, analysts, researchers, managers and writers are all exposed to this. Anyone whose value comes from making good judgements about complex situations should pay attention to which cognitive tasks they are still doing themselves.
Four concrete things to do
First: form your view before you ask AI for anything. Write two or three sentences summarising what you think before you prompt. This takes ninety seconds and keeps the thinking muscle active. Second: write your first draft without AI. It does not need to be good. The point is to practice constructing an argument from scratch. You can use AI afterwards to improve it.
Third: treat AI output as a position, not a conclusion. When you receive an AI response, ask what it has left out, what assumptions it is making, and whether you actually agree. Disagreeing with AI occasionally is not pedantry. It is maintenance. Fourth: schedule regular practice of the tasks AI makes easy. If AI summarises for you at work, summarise something yourself each week. If it generates options, generate a list yourself before asking. These are not productivity exercises. They are cognitive upkeep.
None of this requires rejecting AI tools. It requires using them in a way that keeps your own analytical capacity intact. The distinction is between using AI as a calculator and using it as a crutch.
Steve Raju is the author of Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You, published April 14, 2026.