ChamberOne

Building an AI training programme for law firm staff

By the ChamberOne team 5 minute read

Most firms start with a single lunchtime talk about AI, attended by the enthusiasts and the nervous, and then wonder why habits have not changed. People rarely learn to use a tool by hearing about it. They learn by using it on work they recognise, being corrected and trying again. A training programme that takes that seriously is more than a presentation, and it need not be expensive.

Key points

  • Define the behaviour you want, such as approved tools, careful input and verification, not just knowledge.
  • Use role-based sessions and real, anonymised cases, with planted errors to teach checking.
  • Teach verification as a short routine and record what was checked.
  • Refresh when tools change and keep attendance records.

Decide what the training is for

Start with outcomes, not content. After training, what should people be able to do, and what should they stop doing? A short list is enough. For example: use only approved tools, know what may never be entered, check every output against its source and know whom to ask.

Aim at behaviour. Knowing that a model can produce invented citations is useful. Having a habit of opening the source every time is what protects clients.

Different roles need different sessions

One session for everyone teaches nobody much. Group people by what they do.

  • Partners: risk, supervision, client conversations, insurance and what to ask of the team.
  • Fee earners: practical use on their kind of work, verification routines and confidentiality.
  • Trainees and juniors: how to use tools without losing the chance to learn the craft, and how to flag a doubt.
  • Support and finance staff: data handling, approved tools and spotting a suspicious request.
  • IT and knowledge staff: configuration, access control, records and vendor questions.

Use real cases, safely

Hands-on practice is the heart of the programme. Use realistic material, taken from closed matters and anonymised, or invented to resemble the work. Give each participant a task they would actually do, such as summarising a lease, drafting a client update or extracting dates from an order.

Then make the learning point of the exercise the checking, not the generating. Hand out a version of the output with planted errors, such as a wrong date or an invented reference, and ask people to find them. Most people find fewer than they expect, which is a more effective lesson than any warning.

Build the verification habit

The one habit worth the whole programme is verification. Teach it as a small, repeatable routine.

  1. 1

    Treat every output as a draft written by a fast, confident stranger.

  2. 2

    Trace each fact to the source document.

  3. 3

    Open each authority yourself and read it.

  4. 4

    Check numbers and dates independently.

  5. 5

    Read for what is missing, not only for what is wrong.

  6. 6

    Record on the file what was checked.

Cover the rules in plain language

Give staff a short policy they can read in ten minutes. Say which tools are approved, what may never be entered, when a client must be told, who approves new tools and what to do when something goes wrong. Explain the reasons behind each rule, since people follow rules they understand. Use examples: this is fine, this is not, and this needs a conversation first.

Make it easy to ask. A named person or a shared inbox for AI questions will do more than a long document.

Make mistakes discussable

People hide mistakes when they fear blame. A firm that wants to learn needs the opposite. Tell staff that reporting an error or near miss is expected and valued, and that the response will be about fixing the process. Share anonymised examples at team meetings. A story about how a fabricated reference was caught is worth more than a page of guidance.

A suggested first session

If you have only half a day, a workable shape is as follows. Spend the first hour on the firm's policy and the reasons behind it, with a few examples of what is allowed and what is not. Spend the next ninety minutes on hands-on tasks drawn from the participants' own areas, using an approved tool and anonymised material. Use the following hour for a checking exercise, with planted errors, and discuss what people missed. Finish with a short session on where to ask questions and how to report a mistake. Collect a few lines of written feedback before people leave, and use them to improve the next session.

Run the same session again for new joiners, as part of induction, so that the habit does not depend on who happened to be in the room on the first day.

Refreshers when tools change

Tools change often, and a programme that ran once is soon out of date. Plan short refreshers, perhaps twice a year, and whenever the firm adopts a new tool or a provider changes how something works. Keep them brief and practical, and focused on what is different. Keep a record of who attended and when, which is useful evidence if a client, an insurer or a regulator asks.

Measure whether it worked

  • Do staff use approved tools rather than personal accounts? The audit record will show.
  • In file reviews, is there evidence of checking?
  • Do errors found after the fact go down over time?
  • Do people ask questions early rather than late?
  • What do staff say, in a short anonymous form, that they still find confusing?

Adjust the programme using the answers. A training plan is itself a draft, and it improves with review. Share what you changed with staff, so they can see that their comments led somewhere and are more likely to give honest ones next time.

  1. 1Run the session
  2. 2Measure using the audit record and file reviews
  3. 3Adjust the programme using the answers
  4. 4Tell staff what changed
  5. 5Refresh when tools change
  6. Then back to step 1.
Improving the training programme

Start small

A small firm does not need a formal academy. A half-day session with real examples, a one-page policy, a named person for questions and a check-in after three months is a respectable programme. Larger firms can build from that, adding role-based tracks and a learning team. The thing to avoid is the single talk that is never followed up.

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

Building an AI training programme for law firm staff