The Best Way to Get Started with AI? Build Something for Your Kid.
Field Note #2
The number one thing that holds lawyers back from using AI isn't the technology. It's not knowing where to start.
I get it. So, start somewhere where the stakes are zero.
Start Personal. Start Free.
When I first got my hands on ChatGPT, I tested it on legal work — obviously. But I also tested it on personal things. Meal plans. Travel itineraries. Gardening in Denmark advice for two native Californians. Explaining things to my six-year-old. And that's where I actually learned how the tools think.
If you haven't started yet, here's my advice:
Download the free versions of ChatGPT, Claude, or Gemini and use them in your personal life first. Figure out what a good prompt looks like when the worst-case scenario is a bad dinner recipe, not a bad contract clause.
One important caveat: with free tiers, always turn off the setting that allows training on your data. And remember that free versions don't carry the same privacy and confidentiality protections as paid plans. Never put anything sensitive into a free tool.
From Chat to Code: Building My Son a Video Game
Right now I use several AI tools for legal work — Claude, Gemini, Wordsmith, and ChatGPT. But I recently started exploring Claude Code, which is a command-line tool that lets you describe what you want to build in plain English, and it writes the code for you.
I hadn't figured out my first legal use case for it yet (I suspect it'll be integrations and data pipelines). So instead, I decided to build a video game for my six-year-old son.
Here's the backstory:
My husband plays one of those home design games — the kind where you design homes, pick furniture, choose layouts. My son watched him play and desperately wanted to try. But he's six. He doesn't use a computer. He doesn't have a tablet. His entire screen experience is watching TV.
So we made a decision:
This would be his introduction to using a computer. But the game his dad plays is way too complex. We needed something simple — build a house, decorate the rooms, drag and drop. Fun, not frustrating. Enter Claude Code to save the day.
The Workflow: Chat First, Then Code
Here's where it gets interesting from an AI workflow perspective.
I didn't jump straight into Claude Code. I started in Claude Chat — the conversational interface — to think through what the game needed to be. I described the concept, the user (a six-year-old with zero computer experience), and what "simple" actually meant in this context. Chat helped me shape the prompt that I then fed into Claude Code to build the actual project.
This is a pattern I use in legal work too: use chat to think and develop prompts, use specialised tools to execute. The thinking layer matters as much as the building layer.
Real-Time User Testing (a.k.a. Watching a Six-Year-Old Get Frustrated)
We built the first version and put him in front of it. And immediately, we could see where it broke down. Not the code — the experience. Things that seemed intuitive to us weren't intuitive to a child who'd never used a trackpad before.
So I went back into Claude Chat with live feedback: "He can't figure out how to drag items." "The buttons are too small." "He wants it 3-D, not 2D" And then the question: should I scrap this and start over, or adjust what we have?
This is exactly the loop that makes AI-powered building so different from traditional development. The feedback cycle is minutes, not weeks. You watch, you learn, you iterate — and the AI helps you do it fast enough that your user (in this case, a very impatient six-year-old) doesn't lose interest.
Why This Matters for Legal AI
You might be thinking: Laura, this is a kids' game. What does this have to do with my legal work?
So glad you asked!
The skills I used to build this game are the same skills you need to implement AI in legal work:
Describing what you want in precise natural language — the core skill of prompting
Breaking a complex problem into simple steps — the same thing you do when structuring a deal or building an argument
Iterating based on user feedback — whether your user is a six-year-old or a client who hates lawyers
Knowing which tool to use when — chat for thinking, code for building, and knowing the difference
The only difference is the stakes. And that's exactly the point. Start where the stakes are low so that when they're high, you already know what you're doing. (Although some may argue building a video game for your impatient six year old son is higher stakes than a contract negotiation…)
Your Homework
This week, pick one personal task and do it with AI. Not legal work. Something for you. Plan a meal. Have AI advise you on your outfit. Build workout plan for the week. Recommend some podcasts or books based on what you’re currently listening to or reading. Build your kid a game. Turn off training on your data, stay on a free plan if you want, and just play. Or upgrade and get a better experience.
The legal professionals who are going to thrive with AI aren't the ones who waited for the perfect use case. They're the ones who started.
Laura Jeffords Greenberg is General Counsel at Worksome and writes about building an AI-first legal function in real time. She previously trained 3,000+ legal professionals on AI use for legal work at Wordsmith Academy.
If you have questions or want to follow up on anything, hit reply.
Want the Field Notes to hit your inbox each week? Subscribe here.
Want to dive deeper? Become a member of The Field Guide.
