Junior roles change first and most. The uncomfortable part is that the tasks disappearing are the ones that produced senior lawyers.
Start a conversation with the AI Adoption Concierge, already scoped to the talent pyramid. Pick a starting point, or describe your situation directly.
Law firm leverage rests on a pyramid: many juniors doing volume work under fewer seniors, with the volume work doubling as the training. Document review taught you what matters in a contract. Research memos taught you the shape of an area. First drafts taught you how a document is built. AI compresses exactly those tasks — the consensus across the commentary is that junior roles change first and most dramatically, shifting from production toward review, validation and oversight. The headcount question is genuinely unsettled and people disagree in good faith about it. The training question is not really in dispute: if juniors stop doing the work that made them senior, something has to replace it, and very few firms have said what.
The point is not that these tasks were valuable output. It is that they were how judgement got built.
Taught what matters in a document and what a problem looks like. Heavily automated.
Taught the map of an area, including the wrong turns. Compressed hard.
Taught how a document is constructed and why each clause exists. Largely automated.
Taught pattern recognition across many deals. Compressed.
The new junior task. Teaches evaluation — but only if they know enough to evaluate.
Always taught the most and was always rationed. Now the main remaining channel.
How firms are responding.
The firm's partners in fifteen years are its juniors now. That is the whole argument.
The plan where AI produces and juniors verify assumes juniors know enough to verify — and the knowledge came from doing the work AI now does. Firms that have not addressed this have a circular plan.
Genuinely unsettled, and the two views are both coherent. One holds that AI automates most of what first- and second-years do, so firms need fewer of them and hire later in the career. The other holds that cheaper legal work expands demand, and that supervising AI output at scale still takes people — different work, similar headcount. Some commentators expect an influx of technologists and data specialists into firms alongside lawyers. What is much less disputed is that the composition of junior work changes sharply, which affects training regardless of how the headcount question resolves.
Deliberately, because it no longer happens as a by-product of the work. The approach several firms have taken is to have juniors do a task both ways early in their development — construct the research or the draft themselves, then compare against the tool — so they experience directly where the tool was wrong and why. It is expensive in time and it is the only reliable way to produce someone who can evaluate output rather than merely accept it. Firms that skip it get juniors who are fast and cannot tell when the answer is wrong.
It is real training and it is not sufficient on its own, and the distinction matters. Evaluating a synthesised answer is a genuine skill, and it presumes a body of knowledge the reviewer already has. A junior who has never constructed the analysis themselves has no basis for judging whether the synthesis is sound — they can check that citations exist, not whether the reasoning holds. Review works as training on top of foundational experience, and fails as a substitute for it.
Start with what juniors are actually spending time on, because most firms have not looked. If production work has already fallen substantially, the training that came with it has fallen too and nobody has noticed, because the output still looks fine. Practical near-term moves: get juniors into client meetings and matter strategy earlier than tradition allows, build the dual-track comparison into the first year, and have seniors review what juniors approved rather than only what they wrote — approval errors and drafting errors look different.
Describe how your juniors spend their time. The Institute will help you find the training that quietly disappeared.