AI can produce the work, it can't produce the lawyer

Think back to the work you did as a junior lawyer - the afternoon spent researching a difficult point. The first draft that came back covered in comments. The precedent you initially thought was perfect until someone more experienced explained why it wasn't.

None of that was particularly efficient but efficiency wasn't the point. Those hours were where much of the learning happened and you wrestled with the problem, got things wrongand gradually developed judgement.

AI can now compress many of those hours into minutes. That's an extraordinary productivity gain, but it also raises a question we're talking about far less: what happens when the work we've traditionally used to train junior lawyers becomes the very work AI performs well?

Research that once occupied an afternoon can now be generated almost instantly. First drafts appear before a junior lawyer has had time to think through the structure themselves and summaries that once required careful reading and studied consideration of what is most important are produced automatically.

The question isn't whether AI replaces junior lawyers; but whether we're still relying on yesterday's apprenticeship paradigm to develop tomorrow's lawyers.

I don't think the apprenticeship is disappearing; I do think it's changing.

The office may become more important because of AI, not less

One of the assumptions surrounding AI is that it naturally accelerates remote and independent working. It follows that if technology can undertake much of the work that once required physical proximity, surely the office becomes less important?

That may sound logical, but I suspect the opposite may be true. Historically, a significant amount of learning occurred through 'doing the work' and drafting documents, researching precedents and preparing advice forced junior lawyers to wrestle with uncertainty before receiving feedback from someone more experienced. Those activities created thousands of small learning moments that accumulated into professional judgement.

As AI increasingly assists with those tasks, firms will need to think differently about where that judgement is developed and the answer may be even closer proximity to experienced lawyers rather than further away from them…

Watching a partner enquire as to a client's commercial objectives, navigate a difficult client conversation, negotiate a commercial compromise or explain why the technically correct answer may not be the right commercial answer teaches something AI cannot. These conversations expose how judgement is applied in practice.

It's something I'm thinking about personally as I prepare to relocate to the US. One of the things I'm looking forward to is spending more time alongside colleagues across Ashurst Perkins Coie and seeing first-hand how different people approach clients, problems and decisions. Technology makes it easier than ever to work across borders, but it doesn't replace what we learn from being around people who think differently from us. If anything, I suspect that kind of proximity is becoming more valuable.

If AI changes the work juniors do, then apprenticeship may depend more than ever and at an earlier stage on observation of personal interaction rather than repetition of tasks, and the office becomes less about producing documents and more about transferring judgement.

The biggest risk isn't junior lawyers using AI…it's never learning when to disagree with it

Much of the debate around AI assumes the danger is that junior lawyers become overly reliant on technology; but I've said before that I think the greater risk is that they become overly trusting of it. Generative AI produces answers that are logical, well structured and persuasive. Often those answers are very good; occasionally they're wrong; but more often, they're incomplete.

That matters because AI is also very good at producing the probable answer. As these tools become ubiquitous, everyone will increasingly have access to a competent first draft, a reasonable interpretation and something approaching the conventional view. In one sense, the 'average' answer becomes commoditised.

Traditional legal training developed a useful counterweight to that, whether by design or accident. We lawyers are almost universally professional sceptics and drawn towards finding the problem. Juniors learnt that instinct by researching competing authorities, comparing precedents, having their reasoning challenged by senior lawyers and debating why one interpretation should prevail over another. The value wasn't simply arriving at the right answer but rather learning how to interrogate one.

If AI removes some of that process, we need to create those experiences elsewhere and, again, I think people become central to the apprenticeship. A junior lawyer sitting alongside a senior practitioner and being asked 'Why do you believe that?', 'What are we missing?' or 'What would the other side say?' is learning something fundamentally different from prompting a model for another answer. Firms may need to make that kind of challenge, debate and contrarian thinking a much more deliberate part of development, and introduce it much earlier.

The aim shouldn't be to train juniors to distrust AI. It should be to develop lawyers who can use an extraordinarily capable tool without outsourcing their judgement to it. The competitive advantage may increasingly lie with lawyers who recognise when the most probable answer isn't the best one, and have developed the confidence and reasoning to explain why.

The best junior lawyers will spend less time writing and more time understanding clients

For most of the profession's history, client exposure was something lawyers earned. Junior lawyers first built technical capability through tasks like research, drafting and document review before gradually moving into client relationships.

The assumption was that commercial understanding followed technical competence but now, AI may reverse that sequence. If everyone instantly becomes the ultimate black letter lawyer and perfect draftsperson though use of generative AI, firms need another way of developing commercial judgement and that may come from exposing junior lawyers to clients much earlier in their careers.

UnderstandiT

The apprenticeship therefore shifts and more time needs to be spent on understanding the people legal advice is ultimately written for. Knowing clients deeply - including their businesses, industries, regulators, competitors and commercial realities - becomes part of legal training rather than something deferred until later in a career.

Perhaps we need to teach more human intelligence, not less

There's an idea from Stanford's Fei-Fei Li that I think is useful here. Li has spent much of her career arguing for human-centred AI and reminding us that, despite the name, there is nothing particularly 'artificial' about artificial intelligence: it is created by people and ultimately affects people. More recently, discussing AI and education, she has warned that the risk isn't simply students using AI, but what happens if reliance on it diminishes the motivation and agency involved in learning.

There is a parallel for professional apprenticeship. If AI can give a junior lawyer the answer, the purpose of training cannot simply be to get them to that answer faster. We need to think much more deliberately about what happens cognitively and professionally on the way there. How do they develop curiosity? How do they learn to read a room, understand someone's motivations, recognise ambiguity, challenge an assumption or make a judgement when there isn't an objectively correct answer?

That becomes particularly important in understanding clients. It's easy to say that professionals need to 'know their client'; it's much harder to do well. The skill isn't simply understanding the problem a client brings to you, but working out whether they've identified the right problem in the first place. What are they actually trying to achieve? What sits behind the question they're asking? Professionals can be very good at solving the problem presented to them without ever discovering that it wasn't quite the problem the client needed solved. That kind of curiosity and enquiry is something we need to develop much earlier in junior lawyers.

Those capabilities have historically accumulated alongside technical training. We may now need to teach them much more intentionally. That could mean earlier client exposure, more observation of senior lawyers, more discussion of why a decision was made and more opportunities for juniors to defend a view rather than simply produce a piece of work. In an age of increasingly capable artificial intelligence, developing human intelligence may become one of the most important functions of the apprenticeship.

The apprenticeship doesn't disappear. It gets redesigned.

The legal profession has adapted to every major technological shift it has faced and AI is unlikely to be any different. What feels different this time, however, is that the technology is changing not only how work gets done but also how capability is developed.

For generations, technical work and professional development were largely the same thing. AI is separating them and if firms continue relying on research, drafting and summarising as the primary training ground for junior lawyers, they may inadvertently remove many of the experiences that once built judgement.

That creates a challenge but also an opportunity.

The firms that produce the strongest lawyers over the next decade may not simply be those that adopt AI most effectively but those that redesign the apprenticeship most deliberately. They will create more opportunities for juniors to observe experienced practitioners, spend time with clients, challenge assumptions and develop commercial judgement earlier in their careers.

Clients have always valued judgement over information and AI simply brings that reality forward. The apprenticeship hasn't become less important - It just needs to prepare junior lawyers for a profession where producing legal work is instantaneous, but understanding people, exercising judgement and knowing when to challenge the obvious answer remain uniquely human capabilities.