Is letting AI think for you weakening your own judgement?

You are halfway through typing the prompt when you catch yourself. You have asked the AI what you think about the acquisition, the hire, the restructure, before you had actually sat with it and formed a view of your own. The question went out before the thought did. It is a small moment, the kind that leaves no trace in any quarterly report, and it is the moment worth watching.

Here is the short answer, because you are busy and the rest of this earns its place underneath it: yes, letting AI think for you can weaken your own judgement, and the mechanism is well understood. It is called AI cognitive offloading, and while it frees capacity in the moment, it can atrophy the critical thinking a leader exists to provide. The good news sits in the same sentence. Judgement is a muscle, and muscles respond to load. Put the load back deliberately and the edge returns with it.

Why do I keep reaching for the AI before I have formed my own view?

Because it is faster, and because it works. That is exactly why it deserves your attention. Offloading a hard cognitive task to a tool that returns a fluent, confident answer is one of the most natural things a stretched mind can do. The trouble is what you stop doing while the tool does it.

In a 2025 study published at CHI (the Conference on Human Factors in Computing Systems, the main academic venue for human-computer research), researchers at Microsoft Research and Carnegie Mellon University surveyed 319 knowledge workers who shared 936 first-hand examples of using generative AI at work. The pattern was clear: the more confidence a worker placed in the AI, the less critical thinking they applied to the task. The more confidence they placed in their own expertise, the more they applied. They also found the work itself shifts shape: critical thinking moves away from gathering and generating information and toward verifying, integrating and stewarding what the machine produced. You become an editor of thoughts you did not have.

A separate 2025 study in the journal Societies, surveying 666 people across the United Kingdom, found a significant negative relationship between frequent AI tool use and critical thinking, with cognitive offloading sitting in the middle as the mechanism. Younger participants leaned on the tools more and scored lower. The direction of travel is consistent across very different methods, and it points at the same quiet erosion.

What is the actual cost when a leader stops thinking first?

The cost is discernment. The career you built was built on it: the read of the room, the number that feels off before you can say why, the strategy that pattern-matches against twenty years you cannot fully articulate. That is the asset. And it is precisely the asset that goes dormant when the first move is always to ask the model.

The clearest evidence so far comes from medicine, where the stakes make the erosion measurable. In a 2025 multicentre study published in The Lancet Gastroenterology and Hepatology, researchers tracked 19 experienced endoscopists across four centres in Poland before and after AI assistance became routine. When those clinicians went back to performing colonoscopies without the AI, their adenoma detection rate, the share of procedures in which they spotted a precancerous growth, fell from 28.4% to 22.4%. A six-point drop, in expert doctors, after only months of leaning on the tool. The authors call it deskilling, and it is the first real-world signal that routine AI use degrades the unaided human skill underneath, with patient outcomes attached.

An MIT Media Lab team found the same shape in a different domain. They put electrodes on people while they wrote essays across four sessions. The group using ChatGPT showed the lowest brain connectivity of the three groups studied, and when asked afterwards in the first session, 83% of the ChatGPT writers could not quote a single sentence from the essay they had just produced. The authors describe this as accumulating cognitive debt, and they have asked the public to avoid the lazy framings of brain rot or harm. I will honour that. The work is a preprint, not yet peer-reviewed, so hold it as a signal rather than a verdict. As a signal, it is a loud one: you can produce output and retain nothing, because the part of you that would have done the producing never woke up.

This is the substrate argument in miniature. The bottleneck is no longer the technology. The tools are extraordinary and getting more so. The constraint that decides whether all that capability turns into good decisions is the quality of the human doing the deciding, and that is the variable most leaders are quietly outsourcing.

Notice the colonoscopy line, because it reframes everything. These were not novices. They were doctors who had each performed thousands of procedures, and a few months of routine AI assistance was enough to dull the unaided eye by six percentage points. If that can happen to deep clinical expertise built over years, it can happen to the judgement you have built over a career. The skill that defines you is the one most exposed when the tool quietly takes the wheel.

How do I keep my edge while still using the tools every day?

You build deliberate friction back in. Not by using the AI less, but by changing the order of operations so your own mind goes first and the machine becomes the thing you test yourself against. The category of intervention here is simple to name and demanding to practise: think first, prompt second, and protect the conditions under which thinking happens at all.

That last part matters more than it sounds. Judgement runs on a body. A systematic review in Frontiers in Neuroscience found that higher resting heart rate variability, the natural beat-to-beat variation that signals a well-regulated nervous system, tracks with stronger executive function, the attention and self-control that good decisions depend on, after controlling for demographic and clinical factors. The neurovisceral integration model (the theory that the heart and the thinking brain regulate each other through a shared circuit) links that variability to how well your prefrontal cortex steadies itself. Put plainly, a depleted leader makes worse calls. The analysis of 1,112 Israeli parole rulings made the point famously: the chance of a favourable decision started near 65% and slid toward zero across a session, then reset after a break. Later work argues that effect is smaller than the headline, so hold it loosely, yet the underlying mechanism, that depletion degrades expert decisions, holds. Coherence, the head-heart alignment that keeps you steady under load, is a capability you can build, and it is what lets the rest of this stick.

1. Form the view before the prompt. Write your own answer in two or three sentences first, even rough. Then ask the AI. Now you are integrating, the high-value work the research describes, rather than receiving.
2. Use the model as a sparring partner. Ask it to argue against your position, to find the hole, to name what you have missed. You stay the one who decides; it stays the one who pressures.
3. Reserve the irreversible calls for your own mind. Offload the reversible and the routine freely. The decisions that define your tenure deserve your own wrestling first.
4. Protect the conditions. Guard the recovery, the breaks, the rested mornings. The parole and heart-rate-variability findings agree: discernment is physiological before it is intellectual.

So bring it back to that moment at the keyboard. The fix was never to type less. It was to think first, then type, and to treat the machine as the sparring partner that sharpens you. This is the shape of the third phase of work. Phase One, the Age of Effort: work hard, get a little more, linear growth. Phase Two, the Age of Scale: build once, sell to millions, exponential growth. Phase Three, the Age of Acceleration: output decoupled from human effort almost entirely, the phase AI unlocks. Phase One was muscle. Phase Two was machine. Phase Three is mind, and the leaders who win the next decade are the ones who keep theirs awake while everyone around them quietly hands it over.

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.