The course
In five Saturdays, you take one problem from your team, redesign how it is handled, and build a working first version of a solution with Claude.
No coding or tool knowledge is needed. Claude writes the code.
5
Saturdays
Live and remote, with short recordings to watch or listen to.
90 min
Each session
Live with both facilitators.
10–16
Groups
Of two to four, with a buddy for each group.
1
Working first version
Saved to GitHub, with a place in the builders' directory.
Why start with the problem
You know the problems in your own work better than any software vendor does. Vendors are technology companies that build for common tasks, and many problems in legal work are not covered by any commercial product.
Many AI projects fail because they start with a tool and look for a problem to use it on. In this course you choose a problem your team has, decide how it should be handled, and then build a first version of a solution.
You use Claude, an AI coding agent, which writes the code. You describe what you want in plain English and review what it builds, the way you review a junior's work. Once you have written down how the work should be done, Claude follows those rules every time. Lawyers already work this way with precedents and checklists.
What you'll be able to do by the end
By the last Saturday, you can do all of this on your own.
- See where AI can change how your team works, beyond helping one person do a task faster.
- Choose a problem worth solving, with help from the group.
- Redesign how it is handled: which steps a tool takes over, which stay with the lawyer, and where a lawyer must check.
- Turn the redesign into a brief that Claude can build from.
- Build a working first version with Claude, and improve it with clear feedback.
- Save it to GitHub and show it to the group.
- Keep going on your own: fix problems with Claude and plan the next version.
What you do, week by week
You work in a group of 2 to 4. Each member brings problems from her own team, and the group chooses one to redesign and build together. Building starts in session four.
The five weeks at a glance
| Week | Focus | What you do | What you learn | What you'll have |
|---|---|---|---|---|
| Session 1Change your perspective | How AI can change the way work is done | Women lawyers who have redesigned work in their own teams explain how they did it: the thinking that came first, before choosing any AI tool, and the value the change created for their organisations. We look at why many AI projects fail when they start from a tool looking for a problem. Before session two, you list five problems from your own week. | Where AI helps a team, beyond one person doing a task faster. | A list of five problems from your work. |
| Session 2Pick your problem | Choosing with the group | You share your five problems. The group helps you narrow them to three, looking at how often each one comes up, how long it takes, and how much a mistake would cost. During the week you choose one. | How to choose a problem worth solving, and keep it small enough to finish. | One chosen problem. |
| Session 3Redesign it | How the work should run | You write down how the problem is handled today, why it should change, and how it will run: which steps a tool takes over, which stay with the lawyer, and when a lawyer must check. A partner from another team asks questions. Then we show what a simple build looks like. | How to redesign a piece of work, and what a build involves. | A redesign, and a brief Claude can build from. |
| Session 4First draft | Building with Claude | Claude builds a first version from your brief. It will be rough, like a first draft of a contract. You review what it builds and tell it what to change, one thing at a time, while the facilitators watch shared screens and help. | How to direct a build and check the result. | A working first draft on your computer. |
| Session 5Improve, save, show | Finishing version one | You improve your first version, save it to GitHub, and show it to the group. We cover what a tool needs before it is used with real client data, and how to keep going on your own. | How to keep improving a tool, and what real use would require. | A working first version on GitHub. |
The technical steps
You do not need any of this until week four. Claude writes the code for each step; you give the instructions and check the result.
| When | Step | What you do | What you'll have |
|---|---|---|---|
| Session 0, optionalSetup clinic | Accounts and Claude Code | A live clinic on the Saturday before session 1, recorded, for anyone who wants help with the technical set-up: create Claude, GitHub and Vercel accounts, install the Claude desktop app, and open a project in Claude Code with the standing instructions and memory files we provide. You can also follow the written setup guide in your own time. A drop-in call before session 4 helps anyone who is still stuck, and you can message the team at any point in the course. | Claude Code running on your computer before session 4. |
| Session 4First build | From brief to first draft | Save your redesign as a brief in the project. Ask Claude to plan the build, check the plan, then let it build. Review the result and give feedback. | A first draft that runs on your computer. |
| Session 5Save and share | GitHub | Save your work to a GitHub repository with Claude's help, and write a short description of what the tool does. | Your tool on GitHub. |
| OptionalNext steps | Going further | Publish the tool on the internet with Vercel, add a database and sign-in, or add an AI step. We cover these in an optional clinic after session 5 for anyone who wants to go further. | A tool you can open on your phone. |
How the course works
Sessions one and two involve no tools. Building starts after the problem is chosen and the redesign is written down.
You bring five problems. The group helps you narrow them to three, and you choose one, checking that it is small enough to finish in the course.
Each group is paired with a woman lawyer who has built with AI. She checks in during the week and helps with questions and errors.
Six kinds of help, so nobody is left stuck on accounts or installation:
- Setup clinic (Session 0): a live, recorded session on the Saturday before session 1, where we set up Claude, GitHub and Vercel together.
- Set up with Claude's help: paste one prompt into Claude and it walks you through each step. Participants find it at Set up with Claude's help after signing in.
- Written setup guide: step-by-step instructions with screenshots, for anyone who prefers to work alone.
- Drop-in call: before session 4, for anyone still stuck.
- Message the team: at any point in the course, in your cohort group or by email.
- Glossary: plain-English meanings of the terms you will meet. See the glossary for beginners.
Guest speakers are women practitioners from law firms, in-house teams and innovation roles who build with AI in their own work.
Each week has one recording of about seven minutes, with an audio version. The recordings cover the material if you miss a Saturday.
Auditors join the live sessions and the recordings to watch and listen, for a lower fee. There is no limit on the number of auditors. Auditors do not join a group, present, or receive a certificate.
At the Session 0 setup clinic, or in the written guide, you receive a standing instructions file, following the approach shared by Boris Cherny, who created Claude Code, and a memory folder of linked notes: your problem, your redesign, each decision, your sources and a log of each session. Claude reads the memory at the start of every session, adds to it as you decide things, and tells you when something has not been recorded, rather than guessing. You can open the folder in Obsidian, a free notes app, to see every note and how they connect, and correct anything Claude has wrong.
How the sessions run
Remote, once a week, with both facilitators on every session.
Five weekly sessions of 90 minutes, on Saturdays at 13:00 Bangkok and 16:00 Brisbane. The first cohort starts in early 2027.
You arrive having watched or listened to the week's short recording. A facilitator or guest explains the week's step and shows it on a worked example. You apply it to your own problem, get feedback from a partner or the facilitators, and some of you share your work with the group.
Ten to sixteen groups of two to four, with both facilitators on every call, and a buddy for each group. Join with colleagues, or on your own and we will match you with others. The fee is per person.
A short recording for the coming week, a check-in from your buddy, and a group chat that we check between sessions.
What it costs and what you need
The course fee is US$199. Bring a colleague free: for each paid place, one colleague from your organisation joins at no cost when you register together. Sponsors support the running and growth of the programme. Scholarships of 50% or 100% of the fee are available for anyone who cannot afford it. Lawyers working in the public sector, including government, courts, regulators and legal aid, receive a full scholarship automatically. Register with your government email address (for example one ending in .gov.au, .govt.nz, .gov.ph, .gov.my or .gov.sg) and choose "Government or public sector"; you do not need to apply.
What you'll pay for
| Item | When | Amount |
|---|---|---|
| Course fee | On joining | US$199 |
| Team offer: a colleague from your organisation | On joining | Free with each paid place |
| Claude Pro subscription | From week four | ~US$20 / month |
| GitHub | Week five | US$0 |
| Optional: publishing, database, AI step | After the course | US$0 to ~US$5 |
- A computerYour own laptop, Windows or macOS, from week four.
- Five problemsThings that go wrong, wait or repeat in your team's work, described with invented examples. They need not be legal tasks: running the team, work for the wider business and your own development count too. You list them after session one.
- Each weekThe 90-minute session, the short recording, and about 30 to 45 minutes on your own.
After the course
Participants whose tool is on GitHub are listed in the directory of women lawyers who build with AI.
Each entry shows the builder, what she built, a link to the GitHub repository and a link to her LinkedIn profile. Each cohort's work is shared when the course ends. Past and current participants stay in one group chat.
Everyone in a group that presents its redesigned workflow receives a certificate of completion from AI Women in Law, with the course hours, which you can add to your LinkedIn profile. Abigail Serrano and Lana Kelly review each tool; those ready to publish go on GitHub and into the directory, and the directory entry is the public record of what you built.
If you want to go further and build more advanced tools than the course covers, you can join LegalQuants, a community of lawyers in over 30 jurisdictions who build their own AI systems. Its coaching programme is four weeks of one-to-one coaching with an experienced member. The founders of AI Women in Law met at LegalQuants, and within it is LadyQuants, a group for women members.
Who you'll learn from
The course is designed and led by Abigail Serrano and Lana Kelly.
Both are practising lawyers who build their own tools with Claude, and they run every session between them.