Give Away the Work, Not the Judgment: AI and Lawyers
7 Sept 2026
Everyone in the room was fine with their lawyer using AI on their case.
Everyone in the room was fine with their lawyer using AI on their case.
Roxana Sharifi and Anaïs Franck asked us directly at the IMD Alumni Club Switzerland evening in Zurich last week. One of the options she offered was: I would rather they did not use it. Nobody chose it. She said that two years ago in a legal audience the same show of hands would have gone very differently.
Then we got to the part I care most about, and the one I work on with every team. We are told to give away the work to AI, but not the judgment.
But if AI does the work, how does anyone build the judgment in the first place?
In a law firm it used to be built by grinding through 150 contracts as a junior. Slowly. Painfully.
→ One firm is considering mandatory no-AI hours for its juniors
→ A US law school has banned AI in the first year
→ But can we really turn back the clock, or do we have to start training entirely differently
It is one of the biggest challenges we have to solve, and the speed is only picking up.
PS: thank you Rafael Martín de Agar IMD Alumni Club Zürich for the evening, and Roxana Sharifi of CMS Switzerland and Anaïs Franck of Thomson Reuters for facilitating a real conversation on AI Impact on Legal🙏 It was lovely having you Dany Kostadinova and Kishwar Chishty!
Would your leaders be ready for this shift?
Want to think through what AI means for your own leadership? My Private AI Advisory is here.
How judgment gets built when AI does the drafting
This is not really a legal-industry problem. Every profession that built expertise through repetition, consulting, finance, coaching itself, is now facing the same gap. The grind used to be the training. Remove the grind and you have not removed the need to train judgment, you have just removed the method that used to build it by accident.
In the executives I advise on AI adoption, the pattern repeats: senior people trust their own judgment because they spent years doing the work badly before doing it well. Junior people coming up now skip straight to AI-assisted output, and it looks competent immediately, which is exactly what makes the gap hard to see until it matters.
What I ask leaders to do instead of banning the tool or ignoring the problem:
→ Name the specific judgment calls a junior needs to practise, not just the output they need to produce
→ Build in a deliberate slow version, even one case a month done the old way, so the muscle does not disappear
→ Review the reasoning, not only the result, so you can tell whether the judgment is actually forming
This is the same question I help clients think through in one-to-one advisory: what to hand to AI, and what you deliberately keep doing yourself so the underlying skill stays alive, which is close to what senior leaders actually need from AI right now.
If you want a structured way to work through your own version of this, I have a free AI Coach Tünde Light with self-coaching prompts that help you decide what to hand over and what to keep practising yourself. Leave your email here and I will send it over: https://www.thechangerepublic.com/free-resources