# AI training for law firms and legal teams Source: https://tpiros.dev/power-user-gap/legal/ AI training for law firms and in-house legal teams, on their own matters. AI training for law firms and in-house legal teams, taught to one firm's own people on its own matters and documents. I look at how your lawyers and the teams around them work, find where AI can do real work in it, and teach your people to use it there, with the tools and the AI policy you already have. For whoever is responsible for getting non-technical people using AI. Nothing to code. The Power User Gap is the split EY's Work Reimagined Survey measured in August 2025: 88% of employees use AI at work, mostly for search and summarising, and 5% use it to change how they work. EY's Work Reimagined Survey (August 2025, 15,000 employees and 1,500 employers across 29 countries) found 88% of employees using AI at work "primarily limited to basic tasks such as search and summarization", and only 5% "using it in advanced ways to transform the way they work". ## Where the gap shows in legal work The Power User Gap for law firms and in-house legal teams is legal AI training for one firm's own lawyers and support staff: they learn to use ChatGPT, Claude or Gemini on their own matters and documents, within the firm's AI policy, and a lawyer stays responsible for every output. These are the kinds of work I'd look at first, for fee earners and for the paralegals, secretaries, knowledge, business development and operations teams around them. Which of them we start with comes out of your team leads and the firm's process documents. - **A first pass through long documents.** A contract bundle or a long lease cut down to the points that matter, with clause and page references, so the lawyer reading it starts in the right place and still owns the conclusion. - **Research starting points.** A first map of a question: the issues to look at, the terms to search on, the questions to put to the firm's own research tools. Every authority gets checked at source before anyone relies on it. - **First drafts of letters and clauses.** Client letters, chasers, standard clauses and the covering email, drafted from the matter notes in the firm's style, for a lawyer to review, correct and send. - **A document against your playbook.** The other side's draft run against the firm's own checklist or playbook, listing where it departs and what is missing, as a first pass before the lawyer's own review. - **Attendance notes to next steps.** A call transcript or a rough attendance note turned into the agreed points, who does what by when, and the follow-up email, in the format the team already uses. - **Finding what the firm already knows.** Knowledge and support teams field the same questions about precedents, templates and internal guidance. The tool drafts the answer from the documents it is given and quotes the passage it relied on. **What the teaching looks like there.** First the people attending say how they do this work today, which is the baseline. Then they practise on their own kind of matters with the general assistant the firm already licenses, inside its AI policy: what to give the tool, what to keep out of it, and how to check what comes back. A lawyer stays responsible for anything that goes to a client, a court or the other side. Afterwards the same people say what they now hand to AI and what they still do by hand. ## How it works You see what changed, team by team. The people attending say how they use AI before anything is taught, and what they hand to it after. It works with whatever tools you've already bought. Remote or on-site, so teams in different countries get the same teaching without anyone flying anywhere. No pre-work for attendees. **For when someone asks what the licences bought.** You hear from the people attending how they use AI before anything is taught, and from the same people afterwards what they now hand to AI and where they still don't. That's the before and after, team by team, to take to whoever approved the spend. 1. A 30-minute call. Where your rollout is, and which teams or workflows you'd like to start with. 2. I look at how your teams work. I talk to the team leads and go through the process documents for the workflows you've chosen, and suggest others where I see a better place to start. How long it takes depends on the scope we agree. 3. A baseline before anything is taught. The people attending say how they use AI on those workflows today and how that work gets done. That is the mark everything after is measured against. 4. I teach your people to use AI there. Remote or on-site, for the people who haven't worked it out yet. They learn to use AI on those workflows, so they do that work more efficiently and AI becomes the first thing they reach for. Your power users are welcome to join. 5. What changed. After the teaching, the same people say what they now hand to AI and where they still don't. You see the change against the baseline, team by team, and we decide where the next cohort starts. The teaching comes as named courses, one per group of teams, and the call picks which. They are listed at https://tpiros.dev/power-user-gap/#courses. ## Who is in the room The people who have not worked it out yet, whatever their job: customer success, sales, support, operations, HR, finance, legal. No technical background, and nothing to code. Your power users are welcome, and it works with the tools you have already bought. If your rollout has not started yet, the teaching works just as well before the licences go out. An engineering team is different work: the AI Engineering Workshop at https://tpiros.dev/ai-engineering-workshop/ teaches them to work with coding agents and to build agents, by the same person, so one conversation can cover both. ## What they leave able to do They use AI on their own workflows to do that work more efficiently, and reach for it first in their day-to-day work. ## What the gap looks like Same task, same tool, two different jobs. What most people send: > Write a summary of this report for the leadership team. What your top 5% send: > Here is the report, and the three slides the board saw last quarter. Summarise it for our leadership team, who already know the background and want what changed. One page. Lead with anything that moves a number. Flag anything you're unsure about rather than smoothing it over. Nobody taught anyone the second version. The people who write it worked it out on your documents and your approvals, which is why no vendor course covers it. I teach it on your own work. ## Why the gap is still there You have probably tried some of these already. This is what each one leaves you with. - The vendor's own course: teaches the product, the features, the modes, what the assistant can do. Your power users didn't learn there. They worked it out on your documents and your approvals, which no vendor has seen. - An e-learning platform: produces completion rates. Everyone finishes the module, and on Monday the same one-line prompts go out as before. - Your own power users: they can show what they do, and they often will. They can't see where their habits came from, and they have their own jobs, so the rest of the team gets a lunchtime demo and a link. - A generic AI trainer: runs the same day for every organisation, on prepared examples. Here the teaching starts from your workflows, so what people practise is the work they go back to. ## Where the gap shows up first The same split appears wherever the licences went out to everyone. This is the work I'd start with. Each setting has its own page. - Legal (https://tpiros.dev/power-user-gap/legal/): first-pass reviews of long documents, research starting points, first drafts of letters and clauses. The 5% already hand these over and check every line. Everyone else still reads and drafts from scratch, fee earners and support teams alike. - Professional services (https://tpiros.dev/power-user-gap/professional-services/): the first draft of a client report, the research behind a recommendation, a proposal that has to read as if a partner wrote it. The 5% already hand these over and check the result. Everyone else still starts from a blank page. - Financial services and insurance (https://tpiros.dev/power-user-gap/financial-services-and-insurance/): policy documents, client communications and regulatory reporting, where the tools have to be used inside the rules. The work is finding what can be handed over without moving the line on accuracy. - Technology and telecoms (https://tpiros.dev/power-user-gap/technology-and-telecoms/): engineering got the training and the rest of the company got the licences. Sales, support, operations, HR and legal are where the gap is widest, and where most of the hours are. - Luxury and heritage (https://tpiros.dev/power-user-gap/luxury-and-heritage/): client follow-ups, product and heritage copy, the same story adapted for every market. The 5% already hand over the first draft and keep the house voice. Everyone else writes it all by hand, or keeps away from the tools. ## Who is teaching 74 talks and 77 workshops in 35 countries since 2014. Google Developer Expert, Agentic AI Foundation ambassador. ## What else the engagement can include - A talk for the whole company: the split, shown to everyone at once, so the teams being taught know why before the teaching starts. See https://tpiros.dev/speaking-for-leaders/. - More cohorts: the same teaching for the next teams. - Briefings for managers: for the managers of the teams being taught. - Follow-up clinics: time with the people who attended, on what they run into back at work. - Change management: working with whoever owns the rollout on the change around it. - Agents in the workflows: putting agents into the workflows we reviewed, for one person, one team or the organisation, as extra work alongside the teaching. ## Before you book **Will the teaching follow our firm's AI policy and approved tools?** Yes. It uses the general assistant the firm already licenses, ChatGPT, Claude or Gemini, and follows the firm's own AI policy on every workflow: which tools are approved, what can go into them and what stays out. Writing that policy stays with the firm, and nothing here is legal or regulatory advice. **Is it for fee earners or support staff?** Both. Partners, associates and trainees work through the same documents as paralegals, secretaries and the knowledge, business development and operations teams, and the workflow review decides who goes first. No technical background, and nothing to code. **Does it replace legal review?** No. Everything in it assumes a lawyer reads, checks and is responsible for whatever the tool helped produce. Much of the teaching makes that review easier to start: asking for clause and page references, and for a flag wherever the tool is unsure. **We already licensed a legal AI tool. Is this still useful?** The teaching works on the general assistants a firm pays for: ChatGPT, Claude or Gemini. If your people have one of those beside the specialist tool, that is where the teaching applies. It does not cover specialist legal AI products. **What do attendees need to prepare?** Nothing. There is no pre-work. The scoping happens between us before the teaching. **What does it cost?** It is quoted against the scope we agree on the call. Nothing is priced on this site because no two organisations need the same amount of work. Questions about the tools, who attends and recording are answered at https://tpiros.dev/power-user-gap/. ## Next step Book a 30-minute call at https://calendly.com/tamas-piros/30min, or email hello@tamaspiros.com. Tell me when the tools went out and which teams you'd start with. It's one quote, for the workflow review and the teaching of the teams we agree on the call, so the figure you forward covers all of it. Nothing here is priced.