By Jamie Ng Global Clients & Markets Partner at Ashurst Perkins Coie
Much of the conversation about artificial intelligence and the legal profession in the age of AI has focused on what this technology might replace.
Which tasks will disappear? How quickly will firms adopt it? What happens to the billable hour as a result? These are all very relevant questions, but I wonder whether we are overlooking a more fundamental one: what happens when the work we have traditionally used to train junior lawyers becomes the very work AI performs well?
For decades, legal apprenticeship has followed a remarkably consistent pattern – junior lawyers learnt by researching authorities, summarising documents, preparing first drafts and working through problems under the guidance of more experienced practitioners.
Those tasks were not only productive but they were also developmental.
Research forced you to distinguish a useful authority from a superficially relevant one; and drafting required you to decide how an argument should be structured. Having work returned covered in comments taught you that the answer you initially thought was obvious often looked quite different through the eyes of someone who had dealt with the problem many times before.
None of this was especially efficient, but efficiency was never its only purpose. Those hours contained thousands of small learning moments through which professional judgement was gradually developed.
Generative AI can now compress many of those hours spent at the junior desk into minutes and research that once occupied an afternoon can be assembled almost instantly. A first draft can appear before a junior lawyer has worked through the structure for themselves, while documents that once needed to be read and compared carefully can be summarised in seconds.
From a productivity perspective, this is extraordinary but from a professional development perspective, it means firms need to think much more deliberately about what was being learnt inside the work that technology is starting to remove at pace.
When production and Development Separate
Historically, law firms did not need to design every element of the apprenticeship because development was embedded naturally into production. A client needed the research; so a junior researched. A document needed drafting; so a junior drafted it.
A senior lawyer needed to review the work before it went out; so feedback followed.
AI begins to separate those two things.
The work may still be produced – and often produced exceptionally well – without requiring a junior lawyer to travel through the same developmental process. This does not mean junior lawyers become less important but it does mean the capabilities we need to develop in them become a lot clearer. The legal skill increasingly lies beyond simply finding the answer.
It is now in recognising what the model has assumed, understanding what it may have missed, appreciating how the client’s commercial circumstances alter the analysis and deciding whether an apparently convincing answer (on the surface) deserves to be relied upon. This is judgement and judgement cannot be developed simply by giving people access to better tools.
Why Proximity may Become more Important
One of the assumptions surrounding AI is that it will naturally accelerate independent and remote working. If technology can perform more of the work that once required teams of people to sit together, it seems reasonable then to assume that physical proximity matters less.
I suspect, however, there is a credible argument for the exact opposite.
If junior lawyers spend less time developing judgement through the mechanics of producing work, then they may need more opportunities to observe judgement being exercised by people with greater experience.
Watching a partner ask why a client wants to pursue a particular course of action, navigate an uncomfortable conversation or explain why the technically safest answer is commercially unrealistic teaches something very different from reading the finished advice afterwards.
These moments reveal the part of our professional expertise that rarely appears in a final document. They show how lawyers listen, frame uncertainty and how they recognise that the legal question presented by a client is not always the real problem that needs solving.
Technology makes it easier than ever for us to work across borders and collaborate without being in the same room and that is an enormous advantage; but it does not eliminate what we learn from spending time around people who approach clients, problems and decisions differently from us.
If anything, as more routine production becomes automated, that type of exposure may become more valuable and the office becomes less about producing documents and more about transferring judgement.
Learning when to Disagree
There is another risk here that deserves more attention. Junior lawyers using AI is not the risk – they should and the technology will become an ordinary part of legal practice. The greater risk is that they become very good at obtaining answers without developing the instinct to challenge them.
Generative AI is particularly powerful because its output is often logical, well structured and persuasive. It can produce a competent first draft or a credible interpretation extremely quickly for anyone using it. Frequently, the answer will be good. Sometimes it will be wrong. Perhaps more difficult is the occasions when it is almost right: technically defensible yet incomplete because a commercial nuance, unusual fact or competing interpretation changes what should happen next for the client.
Traditional legal training created a useful counterweight to this and junior lawyers learnt to compare authorities, defend their reasoning and have their conclusions challenged by people with more experience.
The value was not simply reaching the correct answer but learning how to interrogate an answer that initially appeared correct.
If AI shortens that process, firms need to recreate the challenge deliberately and junior lawyers need senior practitioners asking why they believe something, which assumptions sit underneath the analysis, what the other side might argue and what would need to change before their conclusion changed.
As AI takes on more of the first-pass production, partners may increasingly need to see development as part of their role rather than an incidental consequence of reviewing junior work. Reviewing improves the work; but coaching improves the professional and the future partner’s role must shift from supervising output towards strengthening judgement in the people and systems around them.
The legal profession has spent enormous energy asking how AI can make lawyers more productive, but the next challenge is making sure that increased productivity does not quietly remove the experiences through which expertise was built.
The apprenticeship is not disappearing but the firms that produce the strongest lawyers over the next decade are likely to be those that stop assuming it will happen automatically and start designing it with considerably more intent.
Jamie Ng is Global Clients & Markets Partner at Ashurst Perkins Coie and was named by the Financial Times as one of the Top 20 Legal Innovators of the Past 20 Years in 2025.
