AI power skills for lawyers, paralegals and law students

Build the AI workspace your practice runs on.

Most lawyers use AI as a chat window. The useful version is a folder: standing instructions, version history, a searchable bank of your sources, and a project manager that tracks your deadlines. I run law school this way. I teach lawyers to do the same.

Or see the January cohort.

A chat window is not a system

You paste a document in, get an answer, and the work vanishes when the tab closes. Next week you explain the matter again from scratch. Nothing is versioned or searchable, and nobody can check what the AI was told.

Among solo practitioners, 71% use AI and 57% have no AI policy (Clio, 2026 Legal Trends Report for Solo and Small Law Firms, summarized by theNorth Carolina Bar Association). Adoption is not the gap. Most firms have the tool and no system around it.

A workspace fixes that. The AI reads the same instructions every time. Every draft has a history you can show a partner. Your research lives in a bank that answers with page cites. Client data stays where your firm's policy says it stays.

Sharper, not lazier. That is the whole stance.

What you will build

Four pieces, and each is a folder or a file. No new platform to log into. Plain folders and files on your own machine, and you decide what any tool sees.

A matter folder, as the AI sees it
smith-v-acme/
├── CLAUDE.md            # standing instructions the AI reads every time
├── sources/             # public law and fictional facts only
│   ├── cases/
│   └── statutes/
├── research/
│   ├── databank/        # searchable, answers cite pages
│   └── sources-log.md
├── drafts/
│   └── motion.md        # every change in Git history
└── admin/
    ├── deadlines.md
    └── weekly-review.md

1. The workspace

One folder per matter or article. A CLAUDE.md file holds the thesis, the style rules and the data rules, so the AI starts every session already briefed.

2. The history

Git records every change to every draft, with a note on why. Redlines, audit trail and backups in one tool, and the cohort covers the commands you need.

3. The databank

A searchable bank of your sources. Ask a question, get an answer that cites the page, then verify it against the original. The verification step is part of the build.

4. The project manager

Deadlines, a task tracker and a weekly review the AI runs for you. Claude Code and Cowork do the busywork. You keep the judgment.

Two commands from a normal Tuesday
$ git log --oneline drafts/motion.md
a41f0c2 Tighten the standard-of-review section
7d93be1 Add the three Michigan cases from the databank
c02e7aa First full draft

$ claude "Which of our sources discuss the business judgment rule? Cite pages."

Three minutes, start to finish

Demo video coming with the first YouTube episode. Captions included.

A matter folder, a versioned draft and a databank query, built live on screen.

Three ways to learn it

Free workspace kit

A folder structure, a basic CLAUDE.md template, one skill and a safety checklist. Free with the newsletter.

Get the kit

The January cohort

Four live sessions. You build a workspace, a versioned draft and a databank around an article you want to write. Founding price $249, students $99.

See the syllabus and dates

Work with me

A two-hour setup on your own machine, a half-day firm workshop, or a full workspace build for your firm.

See the options

Who is teaching this

Harrison Huijskens, founder of Tenth Hour Labs

I am Harrison Huijskens, a 2L at Michigan State University College of Law. Every class, every paper and this business run from folders, Git and Claude Code. I am not a licensed attorney, and nothing here is legal advice. What I can show you is a setup that works, with the mistakes left in.

More about me and how I work

The Tenth Hour, weekly

One email a week from my own setup: a file tree, a workflow, a safety step. The free workspace kit comes with the first issue.