In short: Good generative AI training for lawyers is built around legal tasks (summarising documents, first drafts, client updates, research orientation), makes verification of every authority and fact automatic, and sets clear rules on confidentiality and privilege. It should also explain what your regulator and the courts expect. A generic "intro to AI" session rarely changes how a legal team works; practice on your own matters, with sensitive data removed, does.
What generative AI training for lawyers should cover
The short answer: five things. Practical legal tasks in the tools your firm has approved, a prompting method suited to legal work, a strict verification routine, clear rules on client data and privilege, and a plain explanation of professional duties. Miss any one of these and the gap tends to show up later, either as licences nobody uses or as an error that reaches a client or a court.
Lawyers are a demanding audience. The work depends on precise wording and reliable authority, and a confident but wrong answer can cause real harm. People do not need a history of machine learning. They want to know what they can safely do with the contract, the bundle or the email thread in front of them, and what they must never do.
Why lawyers need more than general AI literacy
A general AI literacy programme is a sensible foundation for a whole firm. Fee earners need more on top of it, for three reasons.
- Invented authorities. General chat tools can produce case names, citations and quotations that do not exist. The best-known example is Mata v. Avianca in the US District Court for the Southern District of New York in 2023, where lawyers were sanctioned after filing a brief that cited fictitious cases generated by ChatGPT. Courts in England and Wales have also warned lawyers publicly about citing fake authorities.
- Confidentiality and privilege. Putting client material into the wrong tool, or a personal account, can breach duties to the client. Training must be specific about which tools and accounts are approved.
- Professional duties do not change. Competence, supervision and accountability still sit with the lawyer, whatever tool produced the first draft.
Legal tasks worth building a workshop around
Tool features and availability change, so check what your licence includes on the vendor's website. The tasks below are where legal teams commonly see early value, whether the firm uses Microsoft 365 Copilot, ChatGPT, Claude, Gemini or a specialist legal platform.
| Task | How AI helps | What to watch |
|---|---|---|
| Summarising long documents | Structured summaries of contracts, witness statements, policies or disclosure, with a list of questions to follow up | Summaries can miss or distort key clauses. Read the original for anything that affects advice. |
| First drafts | Client letters, internal memos, standard clauses, engagement letter wording, plain-English explanations | Drafts can invent facts or terms. Check every name, date, figure and reference. |
| Contract review support | Comparing a draft against a playbook, flagging unusual clauses, suggesting fallback positions | The playbook and the judgement stay with the lawyer. AI flags, it does not decide. |
| Research orientation | Mapping the issues, suggesting search terms, explaining an unfamiliar area at a high level | Never rely on a cited case or statute until you have checked it in a reliable legal database. |
| Email and matter admin | Summarising long threads, drafting chasers, turning meeting notes into action lists | Check tone and commitments before sending. Check consent and retention rules for recordings. |
Example: a first-pass contract summary
An associate receives a 60-page supply agreement and needs a briefing note for a partner by the afternoon. Training shows how to ask for a summary organised by term, liability, termination, indemnities and governing law, plus a list of clauses that look unusual against the firm's standard position. The associate then checks each point in the original and adds the judgement. The time saved is in the first pass, not in the legal analysis.
Example: explaining advice in plain English
Clients often struggle with dense advice letters. A lawyer can ask AI to rewrite a section for a non-lawyer finance director, then check that nothing has been lost or overstated. This is low risk, easy to review and a good first exercise for sceptical colleagues.
A prompting method that suits legal work
Results improve when every prompt includes the same elements:
- Role and audience: for example, "You are helping a commercial solicitor brief a non-lawyer client."
- One clear task: "List the termination rights for each party."
- Source and limits: "Use only the attached agreement. If something is not in it, say so."
- Format: a table, numbered clause references, British or US spelling, a word limit.
Asking the tool to quote the clause it relies on, and to say when it is unsure, makes review faster. Teams that keep a small shared library of tested prompts for recurring tasks get more consistent results than teams where everyone improvises.
Verification: the habit that protects you
Treat every output as a draft from a capable trainee who has never seen the matter. A workable checking routine:
- Check every case, statute and quotation in a reliable legal database or the original source.
- Tie every fact, date, party and figure back to the documents.
- Remove any statement you cannot trace to a source.
- Record AI use on the matter in line with firm policy, and with any court or client requirements.
The lawyer who signs or files the work is responsible for it, whether or not AI helped. Training should say this plainly and early.
Confidentiality, privilege and approved tools
Before any training session, the firm or legal department should agree and write down:
- Which AI tools are approved, and on which licence or account.
- Which categories of data may be used with each tool, including client data, privileged material and personal data.
- Whether engagement letters or client terms need to mention AI use.
- Who decides when a case is unclear.
Business and enterprise plans from the main vendors generally come with stronger data commitments than personal accounts, but terms differ and change, so your risk or IT lead should read the current documentation. Our guide on whether Claude trains on your data shows the kind of questions to ask, and our AI usage policy template is a starting point for written rules.
What regulators and courts expect
Training should summarise the guidance that applies to your lawyers, and point to the current versions, because they are updated:
- United States: the American Bar Association's Formal Opinion 512, published in July 2024, covers competence, confidentiality, communication with clients, supervision and reasonable fees when lawyers use generative AI. State bars and individual courts may have their own rules or standing orders.
- England and Wales: the Law Society has published guidance on generative AI for solicitors, and the judiciary has issued guidance on AI for judicial office holders. The SRA's existing standards on competence, confidentiality and supervision apply to AI use as they do to any other tool.
We are trainers, not your regulator or your lawyers. The point of this section in a workshop is awareness, so people know the rules exist and where to find them. Your risk and compliance function should own the detail.
How to run the training so it sticks
- Use real, sanitised work. Ask participants to bring a recent task with client details removed or approved for use.
- Group by practice area. Litigation, corporate, employment and in-house teams use AI differently. Small, role-based groups keep examples relevant.
- Include partners and heads of legal. Supervisors need to know what good AI-assisted work looks like. A short executive briefing for leadership helps set policy and expectations.
- Train on the tools you have. If the firm runs on Microsoft 365, Copilot training on Word, Outlook and Teams is usually the fastest win. If you have chosen Claude or ChatGPT, train on that.
- Follow up. A check-in two to four weeks later, reviewing what people actually use and where verification caught errors, is where habits form.
How to tell whether it worked
- Active use of approved tools by role, compared with licences bought.
- Time spent on two or three named recurring tasks, before and after, as estimated by the team.
- Errors caught in review that came from AI output, tracked openly rather than hidden.
- Fewer people using unapproved personal tools for work.
We do not promise a particular saving. It depends on your matters, your data and how consistently people practise.
Next steps
We deliver practical AI training live online in UK and US time zones, or in person in London, built around your lawyers' own workflows and the tools you have approved. Published rates are on the pricing page. To see where your team stands first, try the AI readiness scorecard, or contact us to plan a programme for your firm or legal department.
Frequently asked questions
Is it safe for lawyers to put client information into generative AI tools?
It depends on the tool, the licence, the account settings and your firm's policy. Business and enterprise plans from the main vendors generally offer stronger data commitments than personal accounts, but terms change, so check the vendor's current documentation with your IT, risk or compliance lead. Training should make clear which data may go into which approved tool, and personal accounts should not be used for client matters.
Do lawyers need specific AI training or is general AI literacy enough?
General AI literacy is a useful base, but lawyers need more: verifying citations and authorities, protecting privilege and confidentiality, and understanding what their regulator and the courts expect. Training built on real legal tasks, such as summarising a bundle or drafting a first-pass letter, changes behaviour more than a general overview.
Can generative AI be used for legal research?
It can help with orientation, issue spotting and summarising, but general chat tools can invent cases and quotations. Every authority must be checked in a reliable legal database or the original source before it is relied on or put before a court. Specialist legal research tools reduce this risk but do not remove the need to check.
What do regulators say about lawyers using AI?
In the US, the American Bar Association's Formal Opinion 512 (2024) addresses competence, confidentiality, communication, supervision and fees when lawyers use generative AI. In England and Wales, the Law Society and the judiciary have published guidance on generative AI, and the SRA expects solicitors to remain competent and protect client confidentiality whatever tools they use. Check the current versions, as guidance is updated.
How long should generative AI training for a legal team take?
A focused half-day workshop on real tasks, followed by a few weeks of supported practice and a short review, usually works better than one long session. The right length depends on how many tools and practice areas you want to cover.
