Why Your AI Training Isn't Working (And What to Do Instead)
Situation
I've watched the same AI training get built more than once now, in more than one place.
The content is usually fine. Here's the tool. Here's what a good prompt looks like. Here's an exercise where you build something. People finish it and can tell you the difference between a vague prompt and a specific one. Some of them even enjoy it.
Then Monday happens.
What people are missing isn't knowledge. Almost everyone I talk to already knows the tools exist and already knows roughly what a good prompt looks like. What they don't know is what they're supposed to do differently on a Tuesday afternoon, with their actual work sitting in front of them. Working that out takes mental time, and mental time is the thing nobody has right now. Days are already full. Not everyone has the instinct to go poke at something new for an hour because they like learning. Plenty of people are just trying to get through the week.
The build-something exercises don't close it either. What you build in a training is generic by design, because it has to work for the whole room. So you walk out having built something you'll never open again.
The AI Move
At some point I stopped trying to design a better version of that and asked AI a different question.
Something close to: instead of creating another training about how to use AI and why we should use AI, what are some other creative ways I could reach people and actually get them excited about it?
I wasn't looking for a curriculum. I was looking for a way to hand people my own enthusiasm for this without putting it in a deck.
That brainstorm is where the Slack channel came from.
The other half of it wasn't AI at all. It was listening. When you actually listen to people, instead of assessing where they are, they tell you exactly where they are. What I kept hearing, in different words every time, was some version of: I know I'm supposed to be doing this. I know what tools I have. I know how to write a prompt. I still don't understand what I'm supposed to do differently in my job.
You can't train your way out of that.
The Shift
I was running a set of professional development sessions, and one of them was on problem solving. Not an AI session. I had a single slide in it about ways you could point AI at the problem itself. One of the examples was asking it to help you find the underlying problem rather than the one sitting on the surface.
One slide. A few examples on it. That was the whole intervention.
Multiple people came back to me afterward, and none of them came back to tell me it was interesting. They came back to say: I tried that, it worked, and here's what I did next. Every one of them had bent the example into something for their own work. A couple of those versions had almost nothing left in common with what was on my slide.
They weren't replicating what I showed them. They were using it as a place to push off from.
The Pattern
The thing that moves between people isn't the technique. It's the evidence that it worked for somebody.
The posts that land in the channel are short and unglamorous. Somebody names the tool, says what they were trying to get done, and shows what came out the other end. Often they'll paste the actual prompt right in the thread so you can take it without asking. Sometimes it's a two-minute video of someone talking through how they set the thing up. There's nothing theoretical anywhere in it.
And what happens next is almost never replication. Somebody reads it and thinks: huh, what's my version of that? The use case itself is frequently irrelevant to their job. That turns out not to matter. What they got was a starting point, and a starting point is the one thing training can't hand you, because training has to stay general enough to serve everyone in the room at once.
It also helps that it came from a peer. Somebody with a job like yours, in a week like yours, who isn't being paid to be excited about this.
I want to be careful here, because I'm not saying the foundational training is worthless. It's needed. People do have to learn how to think about this and how to ask for something well, and that should keep happening.
It just can't be where it stops. Training gets people to the edge of relevance and leaves them there, and relevance is where the behavior actually changes.
The Implication
Pick one thing you did with AI this week. Something small and specific enough that it would sound boring to say out loud.
Write it down in four lines. What you were trying to do. What tool you used. What came back. The prompt, if there's one worth copying.
Then put it somewhere your team can see it. A channel, a team meeting, an email thread, a slide in a deck you were already going to present.
You're not teaching anyone anything. You're giving one person a place to push off from.
What I'm Testing Next
I've been trying to figure out how to tell the difference between the people who are one example away from moving and the people who are further back than that. Right now I'm reading it off how people engage, and I want to know if that read is actually accurate or just the one I find flattering.
Comments