You can't learn AI from a list of prompts
The best teacher I ever had barely said a word. Here's what that taught me about helping lawyers use AI.
I spent time today with a lawyer in Los Angeles who wanted help with Claude. She had vague ideas about what she wanted to do, so I listened.
After a while, I realized I could help her with one specific thing. Not ten things. One.
We worked on updating a Claude skill together. I showed her the manual way, which is cumbersome, and then a shortcut for doing it automatically. Then we built a second skill for writing her emails. She saw how it was structured and how I put it together. I told her to paste it into Claude and say, “I want to create a skill like this, but tailored to me.” She got it right away.
The realization
If you want to learn how this stuff works, you have to strip out the conceptualization and the abstraction.
People don’t learn from abstraction. That’s what most CLE programs are. That’s what “here are nine prompts that will change your practice” lists are. They sound good. They make sense at one level. But they’re theoretical, and they have no specific application to your actual work.
My golf instructor figured this out decades ago
His name was Jimmy Self, and he was a renowned instructor. He was always booked, so it was hard to get an appointment.
I’d had other instructors before him. They were all conceptual — they’d video my swing, show me what I was doing wrong, explain the mechanics. It all sounded reasonable. Everyone agreed it made sense. I didn’t get better.
Jimmy didn’t work that way. When you finally got in, he wouldn’t really talk to you. He’d put the ball on the ground, place your hands on the club in the right grip, then stand behind you and move your body through the swing. Back then, though. He’d just say, “Now we’re going back. Now we’re on the other foot.” Cradling you through the motion.
It looked strange. But you’d instantly start hitting the ball straight. As soon as you hit a bad one, he’d check your grip and rock you through it again.
He was transmitting the information straight into your body by putting you through the motions. That’s why he had a waiting list.
One day I told him I’d read that Jack Nicklaus said you want a slight right-to-left draw because the spin makes the ball roll farther. Jimmy listened. Then he picked up my club, dropped a ball on the mat, and hit shot after shot — all dead straight. “See that?” he said. “That there is a straight shot. That’s what you want. Any other questions?”
I didn’t have any. He’d removed every bit of abstraction from the equation. He refused to let it in.
What this means for learning AI
The Jimmy Self approach is what works for teaching people to use AI. Here’s what it looks like:
Teach one-to-one, on a real task the person actually wants done.
Show them how to do that one thing, and let them get a real result.
Repeat with a few more specific things.
Once they’ve had real wins, they can start handling the more conceptual stuff.
That last step is where a group helps. People can compare what’s working, what they’re trying, and how they pulled it off.
But the reverse doesn’t work. Leave people on their own, or drop them in a group where the abstraction seeps in like a fog bank, and you get the same three things every time: confusion, uncertainty, and no clarity at all.
Start with one real thing. Get a result. Build from there.
;-)
Ernie
P.S. This is the whole idea behind the Inner Circle — lawyers learning AI on real tasks, together, instead of collecting prompts that never get used.
→
https://ernietheattorney.net/