The AI competence test every junior solicitor should pass


Guest post by Dr Gleb Tsipursky

Tsipursky: Firms need to develop juniors’ skills despite AI

A junior solicitor can produce a polished legal argument with an AI tool before lunch. The more revealing question comes afterward: can that solicitor explain which authority supports the argument, how it applies, and where the analysis might fail?

A firm that cannot answer those questions is taking an avoidable risk, even when its AI policy looks impressive.

The Solicitors Regulation Authority has given law firms a timely reason to examine that gap. Its consultation, opened on 8 October and running until 3 December 2026, proposes updating the Statement of Solicitor Competence to address technology and AI in research, drafting, professional judgment, and supervision.

The SRA favours embedding AI within existing competences rather than creating a standalone technology competence. These are proposals, with final versions expected in April 2027; they are not yet new binding rules.

That approach makes sense. Competence has always involved reaching a defensible conclusion from sources, facts and professional duties. AI changes the speed and appearance of the work. It can conceal whether a lawyer has practised the judgment that the polished output seems to demonstrate.

Consider the case Legal Futures reported on 26 May. A junior solicitor at Pinsent Masons used AI to help draft misleading letters to the High Court. The judge examined a conversation in which the AI supplied wrong or misleading legal material, while the lawyers reviewing the letters failed to catch the problem. The firm self-reported to the SRA.

The case matters because it exposes two weaknesses at once: the user’s verification and the supervisors’ oversight.

In August, the SRA warned firms about inaccurate AI-generated court material and confidentiality risks. Its notice reiterates that solicitors and firms remain responsible for competent service and effective supervision.

A written instruction to check AI output may help establish expectations. A firm still needs evidence that its people can do the checking.

I would begin with a practical competence exercise that takes roughly half an hour and uses a simulated matter.

Give a junior solicitor a short AI-assisted memorandum containing a sound point of law, a plausible but unsupported proposition, and a genuine ambiguity. Ask for a corrected memo and an explanation of the changes. Score three abilities separately:

First, test source verification. Can the solicitor locate the primary legislation or judgment, identify the relevant passage, check whether it remains authoritative, and show exactly which proposition it supports?

The task should require an independent route to the source. A second AI-generated summary of the first is no substitute for checking the authority itself.

Second, test legal reasoning. Ask the solicitor to explain why the authority applies to these facts, which distinction could change the outcome, and what an opposing lawyer might argue.

This step matters even if every citation is genuine. Accurate quotations can still be assembled into a misleading legal conclusion.

Third, test escalation. Present one fact that makes the apparent answer uncertain. Does the junior identify the uncertainty, state what additional information is needed, and seek supervision before giving a client or court an unsupported assurance?

A competent lawyer knows the boundary of an answer. Confident language should never be mistaken for confidence justified by evidence.

The exercise should be repeated under two conditions: once with the firm’s approved AI tools and once using the underlying sources without generated analysis.

That comparison offers a view of what the solicitor can do with assistance and what expertise remains available when the tool fails. It is a training diagnostic, not a proposal for an extra regulatory examination.

The supervisor needs a parallel test. Give a senior lawyer the junior’s apparently polished memorandum and a short account of how it was prepared. Can the supervisor identify where to sample-check, which risks justify a deeper review, and what feedback would build the junior’s judgment?

The Pinsent Masons episode illustrates why asking only whether a supervisor approved the document is insufficient.

Law firms should also record what these exercises reveal. A short review sheet can capture the unsupported claim discovered, the authority checked, the decision to escalate, and the coaching provided.

Retesting after further practice shows whether competence is improving. The purpose is to improve professional learning and case quality, not to turn every file into an elaborate audit trail.

The exercise must respect confidentiality and proportionality. Start with fictionalised or properly sanitised matters. Use the firm’s authorised systems and ordinary supervision structures.

Give juniors a clear explanation that surfacing uncertainty is an expected professional behavior, not evidence of weakness. Otherwise, people will learn to hide doubts while producing fluent answers.

AI can spare trainees hours of routine searching and drafting. That is valuable. Yet the early years of practice have also been where junior solicitors learn to notice the odd fact, distrust the too-neat answer, and understand a supervisor’s challenge.

If firms remove some repetitive work, they need to preserve deliberate opportunities to develop those capabilities.

The SRA’s consultation creates an opening for firms to treat AI competence as something observable. The sensible test asks whether a junior can verify, reason and escalate, and whether a supervisor can genuinely supervise those judgments.

The legal profession should welcome faster tools while insisting that the person signing the work can still defend it.

Gleb Tsipursky PhD is a behavioural scientist, chief executive of Disaster Avoidance Experts, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026)




Leave a Comment

By clicking Submit you consent to Legal Futures storing your personal data and confirm you have read our Privacy Policy and section 5 of our Terms & Conditions which deals with user-generated content. All comments will be moderated before posting.

Required fields are marked *
Email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Loading animation