Skip to content
All articles
Management

Who Does AI Strengthen More: the Novice or the Experienced Professional?

Who Does AI Strengthen More: the Novice or the Experienced Professional?

Two 2026 experiments with lawyers show that AI quickly lifts the quality of a novice's work, while independent judgment grows among the experienced. A short training on the tool delivers more than access alone.

Early studies of generative AI produced a convenient conclusion: the tool pulls weaker performers up to the level of stronger ones. The 2026 studies show a more complex picture.

How it started

In a 2023 experiment with 758 BCG consultants, participants with below-average results gained 43% in quality, and participants with above-average results gained 17%. In a customer support study of 5,172 employees, average productivity rose by 15%, noticeably more for newcomers. Both measurements were made on chat assistants from 2021-2023, and both measured the quality of work done together with AI.

One question remained: what happens to the person. Do they become stronger as a professional, or is it only the pairing of person and tool that gets stronger?

Three months with AI, then a test without it

That question was answered by an experiment from David Autor's group, published in 2026. It involved 133 practicing patent lawyers from 11 US firms. Some of the participants worked with an AI assistant for three months, the others without it.

With AI, the quality of work rose: by 0.34 standard deviations on day ten and by 0.38 on day ninety, more for junior lawyers. This matches earlier studies.

The lawyers were then tested on a task without AI: they had to redline someone else's text. The result split.

GroupGain in independent skill without AI
Everyone who worked with AI0.32 standard deviations
Senior lawyers0.45
Junior lawyersno average gain

The advantage sat entirely with the seniors. Among juniors the results diverged: fewer middling pieces of work, more weak ones and more strong ones.

The explanation runs as follows. An experienced lawyer checks the AI's answer against their own understanding and refines it. A beginner more often accepts the ready answer, and the understanding does not form.

Access without training does harm

The second experiment was run by Benjamin Chen and Hong Bao. 164 law students took an issue-spotting exam in three groups: without AI, with AI, and with AI after a short training.

  • •The group with AI and no training wrote shorter answers, misstated cases more often and scored slightly below the group without AI. Most of these differences are not statistically significant.
  • •After training, 41% of students used AI against 26% of the untrained ones. The trained students scored 0.27 points higher and stated the applicable rules more accurately.
  • •Training works mainly through the willingness to use the tool.

A separate observation by the authors changes the usual conclusion. Without training, strong students opt out of AI, while weak ones adopt it and perform worse. The leveling effect appears only after instruction.

What follows from this

Both studies were done on lawyers, the second one on students. The figures should be carried over to other professions with caution. The direction is consistent with labor market data: employment of experienced workers grows faster in occupations that rely heavily on tacit knowledge.

For a company this means three practical steps.

  1. 1Teach working with AI deliberately. Handing out access falls short: in the experiment, a short briefing changed both the share of users and the quality.
  2. 2Build the development of juniors around the skill of verification. Some tasks without AI, and a review of finished answers together with a senior colleague.
  3. 3Assign review to the experienced. They are the ones whose independent judgment grows from working with AI.

How this fits into the overall picture of a team is in our overview "AI and Narrow Specialization".

Related service

Express Financial Diagnostics

An independent look at your finances in 5 days: 3-5 findings and a plan for what to fix first

Learn more

Start with a 30-minute consultation

Free and with no obligations: we will review your task, propose a working format, and quote the cost. If your team can solve it on its own, we will say so directly.

We respond within 1 business day