I use AI every single day in my litigation practice.
I use AI every single day in my litigation practice. But I know a lot of attorneys who tried it once, got a bad result, and wrote it off entirely. Hereβs what theyβre doing wrong:
They're using a crappy model. The free versions of ChatGPT and Gemini are not built for serious legal work. Theyβre garbage. Use the smartest, paid versions available.
They're writing bad prompts. AI needs the same thing a junior attorney needs: context, a clear ask, and some sense of what good work looks like.
In the past couple years, Iβve developed a method called DRAFT to fix both problems.
D β Dread is your trigger. The moment you're dreading a task is the moment to hand it to AI.
R β Recruit it. "How can you help me with this, and what do you need from me?β
A β Ask well. Give it context, a clear instruction, your specs, and an example of what a good output looks like.
Fβ Fix it. If the first output isn't right, don't start over. Tell it what was wrong and go another round.
T β Template it. Once you've landed on something that works, save it. Ask the AI for a reusable prompt so you never have to build it from scratch again.
Thatβs it.
When you build the habit of engaging with AI, you'll finally walk away with more time back in your day rather than more frustration.