How should a law firm train its fee earners to use AI?

Train on the work your firm actually does, not on a general demonstration. A fee earner needs three things: a clear picture of the tasks the tool is approved for, a habit of checking every answer against the source, and a note of what to record on the file. That can be taught in an hour and reinforced in short sessions afterwards, and it works better than a single long induction that nobody remembers.

Start with a real matter, not a feature tour

Most AI training fails because it shows the tool doing impressive things that nobody in the room needs to do. A fee earner watching a demonstration of a capability they will never use learns nothing they can apply on Monday.

Pick a closed matter from your own files, one that is typical of the department's work. Run a real task on it in front of the team: pulling the chronology out of the papers, checking what a piece of correspondence actually says, working out a date. Let people see the tool get something partly right and watch you catch it.

Use a closed matter, or one where client consent and your engagement terms plainly cover it. Training material is still client material.

What every fee earner needs to know before they touch it

Keep the taught content short enough to fit on one page. At a minimum:

1. Which tasks are approved. Name them. Research on official sources, questions about a matter's papers, first drafts, date calculations, whatever your firm has decided. Anything not on the list needs a conversation with whoever owns AI at the firm.
2. Which tool, and only that tool. If the firm has approved one product, say so plainly, and say that free consumer tools are not to be used for client work. People fill gaps with whatever is on their phone if you leave a gap.
3. What must never be pasted anywhere. Client names, papers, and anything confidential go only into the approved system.
4. That the output is a draft. Nothing leaves the firm, gets filed, gets served or gets sent to a client without a qualified person reading it properly first.
5. How to check. Every point traced back to the page or the source it came from, and read there.
6. What to record. Your file note standard, in one sentence.
7. What to do when it goes wrong. Who to tell, and that telling someone early is treated as good practice, not as a fault.

Teach checking as a skill in its own right

The part that needs real training time is checking, and it is the part most firms skip. Fee earners are used to checking a trainee's work. Checking an AI answer is a different exercise, because the answer arrives confident, tidy and complete in tone whether or not it is right.

Three things to drill:

• Open the source. Not the summary of the source. If the answer says a document says something, read the page. If it cites a section of legislation, open the section. An answer that cannot be traced to something you can open is not an answer you can rely on.
• Look for what is missing. A summary that reads well may have left out the document that cuts against your client. Ask what the answer does not mention, and go looking.
• Notice the silence. A good tool tells you when the papers do not say something, and flags a page it could not read. Teach people to treat those flags as work to do, not noise to scroll past.

A useful exercise: give two fee earners the same output, one containing a planted error, and ask each to say what they would need to see before signing it off. It teaches the habit faster than any slide.

Keep it short, and keep coming back to it

One induction session is not training. Skills built in a single sitting fade within weeks, and the tool and the firm's rules will both change.

What tends to work in a firm of four to fifty fee earners:

• A short session at the start, department by department rather than firm wide, so the examples are relevant.
• Fifteen minutes at an existing team meeting once a month, where someone shows one thing that worked and one thing that did not.
• A named person people can ask without feeling foolish.
• A short written note of the firm's rules, reissued when they change, and read as part of induction for new joiners.

Record who has been trained and when. If a client, an insurer or a regulator ever asks how you supervise the use of AI in your firm, a training log and a one page policy are a far better answer than an assurance that everyone is sensible.

Supervisors need training too, and slightly different training. A partner who has never used the tool cannot judge whether a junior's work with it was competent, and cannot ask the right questions when reviewing it.

Where Alesis fits

Alesis is an AI assistant for UK law firms, made by L25 Limited, used through the web browser, with one conversation for a matter. It answers questions about a matter from the matter's own papers and names the page each answer came from, and if the papers do not say, it says so, which gives training a concrete checking step: open the page and read it. It prepares drafts for a qualified person to review and sign off, and it does not file, serve or email anything. It assists qualified professionals and does not replace them, and it does not provide legal advice.