AI should make people better thinkers, not just faster producers

What the paper develops

Your team now submits AI-assisted work faster. One recommendation reads well and reaches the review early. Then a leader asks which assumptions carry the decision and what evidence could overturn it. No one can answer without opening the chat again. The tool did not just help write the analysis. It supplied the first view, the challenge, and the wording of the final decision.

That is the risk behind an answer-first rollout. Adoption can rise while independent judgment becomes harder to see. The cost appears when the model is wrong, the case is unusual, or the tool is not available. A person who only approved a finished answer may not be ready to rebuild the reasoning.

The same model can support a different habit

Two randomized trials show why the interaction matters. Bastani and colleagues studied nearly 1,000 high-school students. Students with a standard AI tool later scored 17 percent below a no-AI group on an exam without the tool. A guarded version of the same model led students through hints and largely avoided that loss. Kestin and colleagues tested a custom AI tutor in introductory physics at Harvard. The AI group had more than twice the median learning gain of a class using strong active-learning methods.

The trials used different students, subjects, and designs. Neither tested experienced workers. They do not prove that a workplace will get the same result. They do show that access to a capable model does not decide whether people learn. The way people use it can keep reasoning active or let them skip it.

A review step needs skill and ownership

The usual phrase, “human in the loop,” is too weak on its own. It does not say whether the person can spot a poor assumption, check the evidence, or reject the answer. It also does not say who makes the final decision. A late signature is not the same as review.

Workplace evidence makes that distinction useful. In a field experiment with 758 consultants, AI users did better on tasks inside the model's tested range. On one task outside that range, they were 19 percentage points less likely to reach the correct answer. People whose results fell tended to accept the AI output and question it less. Other studies show that a wrong AI explanation can sound as convincing as a right one.

Make people form a view before they ask

For high-judgment work, start with a short pre-mortem. Before opening the tool, the worker states the answer they expect, names the assumptions that must be true, and says what evidence would change their mind. AI then challenges that first view. It can find a weak assumption, raise a counterargument, or point to missing evidence. The worker checks the challenge and owns the conclusion.

This pattern does not make the model more reliable. It makes the person's reasoning visible. It also gives a reviewer something concrete to inspect: the first view, the challenge, the evidence, and the final decision.

Support the habit in three places

Training should teach people to test assumptions and evidence, not just write prompts. IT and operations should give them a safe, usable tool and require stronger checks when the cost of error is high. Managers should ask what changed the person's view and who makes the decision. If any one of those pieces is missing, deadline pressure will push the work back toward ask, receive, and paste.

Start with one quarter of real work

Choose one function where judgment already matters, such as strategy, risk, client advice, or product decisions. Name one executive owner. For one quarter, require the three-step pre-mortem on selected work. Compare speed, rework, error, and review quality with the old method.

Then decide whether to expand, change, or stop. The goal is to keep people able to form, test, and own a view when the answer matters, without adding needless delay. AI should help people think better, not make their thinking disappear behind a polished response.