How should a law firm supervise junior staff who use AI?

Supervising a junior who uses AI is not a new obligation. It is your existing supervision duty applied to a tool that produces plausible text quickly. The qualified person remains accountable for the work, the output is reviewed before it is used, and the junior understands where the tool helps and where it must not be trusted. Nothing about AI changes who signs off.

Start with the boundaries, not the tool

A junior needs to know what they may do with AI before they touch it. That is a matter of instruction, not software.

Be explicit about:

• Which tasks AI may assist with: summarising papers, first drafts, locating a passage, checking a date calculation.
• Which tasks it may not decide: whether to advise a course of action, what to tell a client, whether a point is arguable.
• What must always be checked by a person before anything is sent, filed or served.
• What may never be entered into a general tool, and which tool the firm has approved for matter papers.

Write these down. A junior who has a short, clear set of rules will use AI better than one left to guess.

Review the output, not the process

You do not need to understand how the tool works. You need to check what it produced, in the same way you would check a trainee's note or a paralegal's draft.

When you review AI-assisted work, ask:

1. Does every factual claim trace back to a source you can see? A summary of the papers should point at the pages it came from.
2. Does every point of law name an official source you can open and read?
3. Are there gaps the junior has flagged, or gaps they have quietly filled?
4. Would you put your name to this if you had written it yourself?

The last question is the real test. If the answer is no, the work is not ready, whoever or whatever produced the draft.

The risks specific to junior use

AI raises particular problems for less experienced staff, and supervision should be alert to them.

A junior may not yet have the knowledge to spot a confident answer that is wrong. Experience is what tells you when something reads plausibly but does not sit right. A junior who trusts a fluent summary because it is fluent is the situation to guard against.

There is also a training risk. If a junior leans on AI for the reasoning as well as the drafting, they may not build the judgement the firm needs them to have in three years. Use AI to remove drudgery, not to remove the thinking a junior is meant to be learning.

Set the expectation early: the tool prepares, the person decides, and the person must be able to explain the decision without pointing at the tool.

Keep a record of who checked what

When work is AI-assisted, it helps to know who reviewed it and against what. This need not be heavy.

• The junior notes that AI was used and for which part of the task.
• The reviewer confirms the citations and calculations were checked.
• Anything the tool flagged as missing or unreadable is recorded and resolved.

This protects the junior as much as the firm. If a client or a regulator later asks how a piece of work was produced, you can answer plainly. It also lets you see, over time, where AI is genuinely helping and where it is creating rework.

Match the tool to the task

Supervision is easier when the tool itself shows its working. A tool that produces a smooth answer with no sources leaves the reviewer to verify everything from scratch, which defeats the point. A tool that names the page or the section behind each point lets the reviewer check quickly and the junior learn from what they see.

When you choose what juniors may use, favour tools that make review possible in minutes rather than hours. The measure of a good AI tool for a supervised setting is not how impressive the output looks, but how easy it is to confirm.

Where Alesis fits

Alesis prepares drafts for a qualified person to review and sign off, and it does not file, serve or email anything, so the accountable person stays in the chain. It answers questions about a matter from the matter's own papers and names the page each answer came from, and when it reads official sources it names each one so a reviewer can open it in one click. When it cannot find support for a point it says what is missing instead of guessing, which makes a junior's work easier to check. Inside a firm people see only the matters they are on, and seniority alone grants no view.