How do you stop your team losing the habit of thinking when they use AI?

For most teams the answer is to decide in advance where the thinking stays with a person, then build one small step into the work to make it happen. The simplest step is to have the person write their own answer in two lines before they ask the AI, then compare the two. It keeps their judgement in use, and it shows you quickly when the AI is wrong.

Why would AI make a team worse at thinking?

Because the cost is hidden by the speed. The reports get done faster. The emails go out sooner. Everything looks like progress, and nothing looks wrong until the day the AI is down, or confidently wrong, and nobody on the team has a view of their own to check it against.

Gartner put a number on the worry in October 2025. Its prediction was that through 2026, atrophy of critical-thinking skills from generative AI use would push 50% of global organisations to require "AI-free" skills assessments. In other words, employers expect to need a way of checking that candidates can still reason without the tool.

That is a forecast, not a measurement, and it is aimed at large organisations hiring at scale. Gartner's later list of predictions for 2027, published in September 2026, does not repeat it, so it is best read as one analyst's view of a direction rather than a settled trend. But the direction matches what we hear from small businesses, and the cost is easier to see in a small team, where a handful of people's judgement is most of the business.

What does it look like in a small business?

Picture the person who used to quote jobs. They knew which customers always change the scope, which supplier runs late in December, which price is too low to be worth the trouble. Now they paste the details into an assistant and send what comes back.

The quote goes out faster. It is usually fine. But the knowledge that made the old quotes good is no longer being used, and a person who is not using a skill tends to lose it. The first sign is rarely a disaster. It is a quote that looks reasonable, goes out, and loses money, with nobody able to say why.

What is the two-line habit?

Before you ask the AI for an answer, write your own in two lines. Then ask it. Then compare.

Where the two agree, you have a quick sanity check and you can move on. Where they disagree, you have found the interesting part. Sometimes the AI has seen something you missed, such as a clause, a figure or an option you had not considered. Sometimes you know something it cannot, like a customer's history or a supplier you would never drop. Either way you have used it to sharpen your judgement, not replace it.

It costs about thirty seconds. It also gives you something a plain review step does not: a record of what the person actually thought before the AI spoke.

Where should the thinking stay human?

You do not need to do this for everything. A good rule is to keep a person's own view in the loop for three kinds of decision.

Decisions about people and relationships, where the context lives in somebody's head and is rarely written down. Decisions that are hard to undo, such as a price, a contract term or a message to a client. And decisions where being wrong is cheap to the AI but expensive to you.

Everything else can lean on the tool more heavily, and should. The aim is not to use AI less. It is to be deliberate about where you do not want to hand over the pen.

Isn't a review step enough?

Only if somebody actually thinks during it. A review step nobody really thinks about is just a rubber stamp. People get used to approving what the AI gives them, the checking becomes a glance, and the control is still on the checklist while no longer doing anything.

That is why at Handiwork we build the comparison step into how a team uses AI from the start, rather than adding a sign-off at the end. Forming a view first is the part that cannot be skipped by accident.

What to do this week

Pick one decision your team made with AI help in the last few days. Ask whether anyone formed a view before the AI answered. If the honest answer is no, that is the place to start. Ask the person to write their two lines next time, and have a look at where they differ from the AI.

If you want to see where else your team leans on AI, our guide to starting with AI as a small business is a good place to begin, and inducting an AI tool the way you would a new hire covers the written standard that goes with it. Your team probably uses AI already, which makes the habit more urgent, not less.

Ready to find out where you stand?

If you want a straight read on where your team leans on AI and where it should not, our free AI Readiness Check takes about five minutes: AI Readiness Check.