Article 5.3 · The Junior Lawyer Problem

T3PS Legal Dynamics · Series 5: The Human Factor · Fact-verified June 2026

Early in my career I spent the better part of a fortnight reading a complete run of correspondence on a dispute that, on paper, looked simple. Nobody told me why.

The partner just handed me the file and said “get to know it.” Somewhere around day nine I noticed that two letters, six months apart, contradicted each other on a date — and that contradiction turned out to be the case. I did not learn that from a textbook. I learned it from the tedium. The profession has always trained judgement that way, indirectly, and mostly without admitting it. We called it bundling, disclosure review, attendance notes, first drafts, research memos, chronologies, proofreading, indexing, due diligence, and sitting quietly in meetings trying to work out why the partner asked the question in that particular way.

Much of that work was tedious. Some of it was inefficient. A great deal of it was badly supervised. But it had one redeeming feature: it exposed junior lawyers to the raw material from which judgement is formed. AI is now moving directly into that work — and that creates a problem the profession has not properly confronted.

The question nobody quite asks

If AI performs the first pass on the documents, produces the first chronology, drafts the first research note, summarises the authorities, reviews the contract pack and prepares the issue list — what, exactly, is the junior lawyer learning? The comfortable answer is that juniors will be freed from drudgery and lifted into higher-value work. That is true, up to a point, and dangerously incomplete. You cannot lift someone into higher-value work if they never acquired the pattern recognition that the lower-value work used to build.

There is a reason an experienced litigator can spot the odd document in a disclosure set. A reason a good costs lawyer can sense, reading a file history, that the time narrative does not match the procedural reality. A reason a senior associate can read a witness statement and feel that the chronology has been quietly improved beyond what the witness could really recall. Those instincts are not downloaded. They are accumulated. AI can accelerate that accumulation if it is used well. It can also hollow it out.

The right diagnosis

The risk is not that junior lawyers become lazy. That is the wrong diagnosis, and it slanders a generation. The risk is that firms remove the work through which juniors historically learned, without replacing it with a deliberate training architecture. The result would be a cohort of lawyers who can review AI outputs fluently but have never developed the underlying judgement needed to know when those outputs are wrong. That is not an efficiency gain. It is deferred professional risk — the kind that does not show up on this year’s figures and shows up badly in about seven years’ time.

There is now hard evidence the concern is mainstream, not nostalgic. Wolters Kluwer’s 2026 Future Ready Lawyer Survey found that over half of respondents expect AI to reduce billable hours, and explicitly flags reduced opportunity for junior lawyers as routine tasks are automated. The profession is not just worrying about this at the coffee machine. It is showing up in the surveys.

Administrative versus formative

The answer is not to preserve bad work for sentimental reasons. Nobody needs trainees paginating bundles by hand at midnight because previous generations suffered. But firms do need to draw a line the billable hour always blurred — between work that is merely administrative and work that is genuinely formative.

Manual pagination is administrative. Reading a full run of correspondence to understand how a dispute actually developed is formative. Typing a dictation is administrative. Drafting a first advice note and then comparing it, line by line, against the partner’s version is formative. Clicking through a thousand irrelevant disclosure documents is inefficient — but understanding why three of those thousand change the case is the most formative thing a young litigator can do. Colin Chapman’s instruction to his engineers was "simplify, then add lightness": strip out everything that does not earn its place. Applied to junior work, that means AI should remove the waste — and it should not remove the apprenticeship. Those are different cuts, and a firm that cannot tell them apart will make the wrong one.

Training juniors to interrogate the machine

The firms that manage this well will redesign junior work around supervised AI interaction. Juniors should not simply be told to use AI. They should be trained to interrogate it. That means a standing set of questions on every AI-assisted task:

—   What did the AI miss?

—   What assumptions did it make?

—   Which documents did it treat as important — and why?

—   Which authorities did it rely on, and are the citations real?

—   Is the summary neutral, or subtly weighted?

—   Does the output reflect the client’s actual commercial objective?

—   What would you have concluded without the tool?

This is a different training model: more explicit, more structured, and more demanding of supervisors. It requires partners and senior associates to teach their reasoning, not merely mark up a draft. That is uncomfortable, because legal training has long relied on osmosis — the junior absorbing judgement by proximity. AI makes osmosis insufficient. The reasoning now has to be said out loud.

The client dimension, and where the billable hour distorts it

There is a client angle that has to be faced honestly. Clients will not want to pay junior lawyers to do work an AI tool performs in minutes. That is entirely reasonable. But clients do have a real interest in the profession continuing to produce competent senior lawyers — the ones they will need in ten years. The answer is not to hide training time inside bills. It is to be honest about what is chargeable, what is investment, and what is part of a properly supervised quality-control process.

This is another place the billable hour quietly distorts the conversation. Under an hourly model, junior training time becomes a recoverability problem — something to disguise or write off. Under a value-based or phase-based model, the firm has room to treat training as part of its own delivery cost. The client buys the outcome. The firm decides how to develop the people who will deliver that outcome sustainably. That is healthier for everyone, and it is one more reason the pricing question and the human-factor question are the same question wearing different hats.

The real question

AI will change the junior lawyer’s work, and it should — much of the traditional junior workload was a poor use of intelligent people. But if firms simply strip out the foundational tasks and replace them with output review, they will open a professional-development gap that becomes visible only years later, when today’s juniors are tomorrow’s supervising lawyers and discover they were never given the raw material that made their own supervisors good. The question is not whether AI should do junior work. The question is how juniors learn when it does.

Three questions to take back to your desk

—   Which tasks in your firm are genuinely administrative — and which are formative training disguised as low-value work?

—   If AI now produces the first draft, what structured exercise teaches the junior lawyer to produce, challenge and improve that draft?

—   Are clients being asked to fund inefficiency — or is the firm investing deliberately in the next generation of judgement?

How a firm trains judgement in an AI era is not an HR footnote — it is a strategy question, and it sits right next to how the firm prices and structures its work. If you are rethinking one, it is worth thinking about the other in the same breath. A Fika is a good place to start that conversation.