Legal Mentorship Was Already Broken. AI Just Made It Obvious.

A strong mentorship culture is table stakes for any law firm that wants to grow. Recruiting materials promise hands on training and a clear path from junior lawyer to trusted adviser. However, mentorship has long been one of legal’s chronic weak spots.

The proliferation of artificial intelligence tools has made that weakness impossible to ignore— forcing the profession to reconsider not only how junior lawyers are trained, but why they matter. Teaching takes time, and time is what the billable hour monetizes.

When mentoring competes with what gets tracked, credited, and compensated, it loses. As David Maister observed in Managing The Professional Service Firm, organizations optimize for what they measure, and underinvest in activities that are harder to price, such as talent development. The result is a persistent gap between the mentorship firms promise and the training their business model actually produces.

Workflow filled the gap. Many firms effectively outsourced training to the work product. Juniors learned by “go-by”: file, save as, revise. The draft became both the deliverable and the curriculum. That was convenient for everyone; it produced artifacts that could be evaluated, and it kept training inside billable hours. Training was embedded in production, requiring little explicit investment.

The traditional apprenticeship model was widespread, but it was never designed to reliably produce judgment. Experience accumulated unevenly, feedback depended on who had time, and development often turned on access rather than design. Left to itself, the system becomes what Maister described as a kind of internal Darwinism in which associates compete for a busy partner’s attention and explanation. Survivors emerge; a scalable training system doesn’t.

That fragility matters now because generative artificial intelligence has removed the camouflage that made this system tolerable: the steady flow of drafting work that doubled as a proxy for training. As AI tools become standard for research and drafting, the “learn by drafting” apprenticeship breaks down. The industry is already moving in that direction, with major legal information platforms rolling out generative AI products designed to accelerate research and drafting.

The mistake is to assume that if drafting disappears, so do junior lawyers, and therefore training becomes irrelevant. While AI can accelerate production, it can’t absorb responsibility; just ask the plaintiffs’ lawyers in Mata v. Avianca, who were sanctioned for including made-up cases in their filings. Even the most tactical of legal work still requires human judgment, verification, and ownership, and those inputs don’t scale simply because drafting does.

Junior lawyers still matter—not because they need something to draft, but because a human must own pieces of the work and the judgment behind them. What is going away is the quiet bargain firms relied on for decades: that judgment would be trained automatically through drafting reps. If machines do more of the drafting, firms lose the ability to hide behind “learning by doing” as a substitute for teaching judgment. And for a law firm, judgment is the product.

The good news is that law school lays the foundation for this kind of training. ABA accreditation standards list legal analysis and reasoning as core learning outcomes, and every J.D. can attest that law school teaches thinking, not drafting.

The question is whether law firms will choose to explicitly build on this foundation by treating judgment as the focal point for early-career development. For firm leadership, this isn’t merely a cultural issue but a business one. Judgment failures surface as write-offs, lost clients, disciplinary risk, and reputational damage.

So, what will it take to train judgment in a post AI workflow?

Firms must make mentorship economically legible. If training is essential to quality, retention, and risk management, it must count. That means real credit for mentoring time, reflected in both performance expectations and compensation. People do what the system rewards.

Firms must invest deliberately in early leadership capability. If lawyers are expected to manage people and judgment, they need tools to do so. Coaching, delegation, and feedback are learned skills, not innate traits. Other industries formalize this; law largely improvises it. In a post ‑AI environment, that gap becomes untenable.

Firms must replace artifact based development with deliberate practice. When the draft no longer functions as the training ground, firms need structured repetitions: simulations, decision reviews, and postmortems that require associates to explain assumptions, tradeoffs, and verification steps. The expertise literature is clear that focused repetition with feedback, as opposed to merely time served, is how professionals level up.

The old system was never ideal. It was inefficient, inequitable, and often accidental. But it had one advantage: It forced repeated exposure to real problems. AI will reduce that exposure, and with it the ability to pretend that judgment develops automatically. Firms can either treat that as a threat—or as a reason to finally build mentorship that is real, measurable, and aligned with how legal value is actually created.

AI won’t eliminate the need for junior lawyers. It will eliminate the excuse that firms don’t need to teach them. In a profession where judgment is the product, the firms that thrive will be the ones that finally build training systems that reflect that reality.

An immaterial amount of this content was drafted by generative artificial intelligence.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law, Bloomberg Tax, and Bloomberg Government, or its owners.

Author Information

Jordan Teague is founding partner and chief strategy officer at Campbell Teague.

Write for Us: Author Guidelines