ChamberOne

Law school curricula and AI: what graduates should be able to do

By the ChamberOne team 4 minute read

A graduate who has written excellent essays can still struggle in the first month of practice. Add AI tools to the mix and the gap widens. Some students use them heavily and without guidance. Others avoid them out of fear of misconduct rules. Employers meanwhile assume that new recruits can use them sensibly, or can at least be trusted to ask. This article sets out what a graduate should be able to do, where teaching and practice diverge, and how educators and employers can talk to each other about it without pretending that anyone has settled answers.

Key points

  • Teach capabilities that outlast any tool: verifying, protecting confidentiality, instructing clearly and disclosing use.
  • Employers should ask for judgement and humility rather than expertise in a particular product.
  • Assess AI-aware work through annotated prompts, error hunts, oral follow-ups and supervised practice.
  • Protect the slow work, such as reading cases and drafting unaided, that builds judgement.
  • Hold a regular conversation between schools and local firms about what graduates need.

The gap between the seminar room and the file

Academic work asks a student to analyse a closed set of materials and defend an argument. Practice asks someone to act on incomplete facts, for a named client, to a deadline, with a supervisor reading the result. The skills overlap but the habits differ.

AI tools sit awkwardly across that gap. In an essay setting, using a tool might be forbidden or tightly limited. In an office, not using one might be odd. A student who has learned only to hide AI use has learned the wrong lesson for the workplace, where the expectation is closer to disclosure, checking and accountability.

Academic work
  • Analyse a closed set of materials
  • Defend an argument
  • AI use may be forbidden or tightly limited
Practice
  • Act on incomplete facts for a named client
  • Work to a deadline with a supervisor reading the result
  • AI use expected to come with disclosure, checking and accountability
Academic work and practice, and where AI sits in each

Capabilities worth teaching

Rather than teaching tools, which change, schools can teach capabilities that will outlast any product. A reasonable list for a graduate entering practice:

  • Explain in plain words what a language model does and why it can produce confident errors.
  • Verify any authority an AI tool mentions by finding the original in a trusted source.
  • Judge what client information may be entered into a tool, and what must never be.
  • Write a clear instruction for a task and refine it when the first result is poor.
  • Compare an AI draft against the source documents and mark omissions and distortions.
  • Say, without embarrassment, when and how a tool was used on a piece of work.
  • Recognise tasks where a tool does not help, such as assessing a witness or advising on risk.

What employers can reasonably ask

Employers sometimes expect graduates to arrive already skilled with AI. That expectation is unrealistic, partly because firms use different tools under different policies. A more useful expectation is judgement and humility: a graduate who knows to check, to ask and to keep confidential material out of unapproved tools.

In interviews, scenario questions work better than questions about favourite tools. Suppose a supervisor asks for a short note on a point of law and the candidate has an AI draft in front of them. What would they do before sending it? A good answer mentions finding the sources, reading them, checking the jurisdiction and telling the supervisor how the draft was produced.

Do
  • Ask for judgement and humility
  • Use scenario questions in interviews
  • Look for a graduate who checks, asks and keeps confidential material out of unapproved tools
Don't
  • Expect graduates to arrive already skilled with AI
  • Interview about favourite tools
Employer expectations of new graduates

Assessing AI-aware competence

Assessment is the hardest part for schools. Closed-book exams test memory and reasoning but not workflow. Take-home essays now invite the question of tool use. Several approaches are being tried, none perfect.

  • Open-tool exercises where the student must submit the prompts, the raw output and an annotated verification note.
  • Error-hunt tasks: give students a flawed AI-written memo and ask them to find and correct the problems.
  • Oral follow-ups, where a student defends the reasoning in work they have submitted.
  • Supervised clinical work in which tool use is logged and discussed with a supervisor.
  • Short, timed tasks without tools to confirm that core skills have not been outsourced.

Protecting the basics

There is a real risk that students skip the slow work that builds judgement: reading a case in full, drafting from a blank page, struggling with a statutory scheme. Tools can make that struggle disappear, and the struggle is how people learn. Educators can respond by being explicit about which tasks must be done unaided and why.

Employers share that responsibility. If a firm gives juniors only the tasks a tool does well, the juniors never develop the judgement needed to supervise the tool later. Rotating juniors through tasks where AI is used and tasks where it is not helps keep both skills alive.

A short case for teaching verification first

If a school can add only one thing, verification is the best candidate. A student who can find the original source behind any statement, and who feels uncomfortable relying on text they have not checked, is safer with a language model than a student with excellent prompts and no instinct to doubt. It is also the habit supervisors notice first.

Verification can be practised in an ordinary seminar. Hand out an AI-written summary of a case the class has read. Ask students to mark every statement as supported, unsupported or wrong, and to cite the paragraph that proves it. The exercise takes under an hour and teaches close reading as well as scepticism.

Working together

Schools and employers rarely sit in the same room. A modest start is a short advisory conversation each year between teaching staff and a few local firms about what graduates struggle with. Another is placements or clinic projects that include a review task: the student receives an AI-assisted draft and must correct it under supervision.

Professional bodies and regulators may publish expectations for new entrants. Check the guidance from your regulator or bar before designing assessments or induction programmes that depend on it, and revisit the design when that guidance changes.

A first-month plan for new recruits

Suppose a mid-sized firm receives its graduate intake. In the first week it explains which tools are approved, what may be entered into them and who to ask. In the second week each graduate completes a supervised exercise: verify an AI-generated research note against sources. By the end of the month they have a short written reflection on what the tool did well and badly. The firm keeps these notes to refine induction for the next cohort. None of this is expensive, and it makes expectations visible from the start.

This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.

Read next

Law school curricula and AI: graduate skills | ChamberOne