In the second installment of the Thomson Reuters Legal team's series on designing judgment in AI-mediated collaboration, we examine how AI changes the way junior lawyers learn to reason.

Evidence shows that AI is not replacing lawyers in 2026; in fact, hiring is up. What is changing, however, is that AI is altering how legal professionals learn to exercise judgment together.

Legal judgment has long been built through the cycle of trying, failing, being corrected, and trying again with junior lawyers taking the first pass and senior attorneys engaging with their reasoning. This process was inefficient by design, but it did foster legal reasoning well. Senior lawyers reviewed not only the juniors’ work product, but also how junior lawyers framed the issue, what they noticed, and what they missed. The gap between the junior’s approach and the senior’s approach became the site of learning.

Today, however, when AI tools now generate the first draft of a memo, a contract, or an issues list, juniors no longer begin with a blank page. And that may feel like an efficiency gain, but it quietly alters the conditions under which judgment is formed.

How AI reorders the sequence

The critical feature of the apprenticeship model was sequencing. Cognitive effort came first and review followed, which ensured that at least one round of original thinking occurred. As AI becomes embedded in intake, triage, and early-stage guidance, however, routine questions are handled automatically, and the matters that escalate arrive better scoped and preprocessed. Indeed, efficiency improves, but junior lawyers are less involved in the earliest stages of problem framing, which is where uncertainty is highest and judgment has traditionally been formed.

Indeed, when a system provides the first answer or directs the next step, it can unintentionally become the first mentor. Increasingly, the first hard thinking in a matter may occur inside the workflow before a junior lawyer is engaged, quietly removing opportunities for the junior to wrestle with uncertainty and frame the problem independently.

When AI is used as a competent starting point in a work product, juniors are repositioned as editors rather than originators. Work still looks collaborative because juniors still touch the document, seniors still review, and outputs may even improve. Beneath the surface, however, the order has flipped. Seniors engage with an AI's output rather than a junior's reasoning.

This feedback shifts toward polishing results instead of examining thought processes. Because the junior did not generate the original framing, there are fewer chances to discuss alternative approaches, uncertainties, or blind spots.

This is the delegation failure mode we described previously, now operating across interactions. Over time, confidence in independent judgment does not develop through refinement alone. The ability to face a blank page and take responsibility for what follows cannot be learned by editing something that was never fully developed by the junior lawyer.

Redefining roles, not resisting tools

To make sure that juniors are learning judgement in AI-mediated workflows, roles must be redefined. Junior lawyers must be expected to interrogate AI output, which means identifying assumptions, testing sources, surfacing what might be missing, and considering alternative framings. Treated as a genuine expectation, this interrogation becomes a new site of cognitive effort.

Seniors, in turn, must move beyond merely redlining AI-assisted work products and engage directly with junior lawyers to make their reasoning visible. When a senior lawyer asks a junior to Walk me through how you got here, it helps recreate the mentorship that the work product itself may no longer reveal.

Workflow design can reinforce these roles. Escalation points, for example, offer a natural opportunity. When a system cannot resolve an issue, escalation creates a structured moment in which a junior forms an initial view before senior review occurs. AI output then becomes a shared object of analysis rather than an unspoken starting assumption. In this workflow, efficiency remains while the cognitive engagement deepens.

What leadership must own

Law firm leaders must learn that apprenticeship in the AI era must be an organizational design choice because AI adoption is as much a talent-strategy decision as it is a technology decision.

Leaders should ask not only how much time AI saves, but rather what their people are doing with this time. Freed-up senior capacity can be reinvested in mentorship and judgment development. Further, metrics focused solely on speed and volume can mislead, because work product may look excellent even as reasoning thins out underneath.

It is leadership's responsibility to ensure that collaboration norms and workflow design continue to reward thinking, explanation, teaching, and the development of independent judgment.

Indeed, the most dangerous scenario is success that masks long-term erosion. If juniors advance after years of editing excellent outputs without learning to frame problems independently, the gap in their skills may surface only when they are called on to exercise judgment under real pressure.

Critically, AI does not eliminate the need for apprenticeship and junior development. Quite the opposite — it makes it easier to assume apprenticeship is still happening when it is not.

Those legal organizations that preserve judgment formation among their juniors will better be able to recognize this risk early and redesign collaboration intentionally before the cost appears in the next generation of legal leaders.

In our next installment, we will turn from people to practice, examining how AI-enabled collaboration in legal research and contracting can create the appearance of shared understanding without the interpretive work behind it.‍