Practical guides and field notes from sessions I have run and systems I have shipped. Start with the planning guide, or go straight to one sharply argued part of the work.
How to define the changed work, diagnose the real constraint, choose the right format, set safety boundaries and measure whether anything survived after the room.
Read the guide →Every training I've watched fail gave the same early signal. People could describe the tool accurately and couldn't name one thing they'd do differently on Monday. That gap shows up long before the feedback form says it went well. What it looks like from inside the room, and what still works once you've spotted it.
Practical test: ask every participant what they will do differently on Monday.
Supervised failure is the mechanism, not an unfortunate side effect. What changes when people break the tool in front of each other, instead of alone at their desk at eleven at night and quietly concluding it does not work.
Senior people are careful because the cost of being wrong lands on them personally, and often because they've already watched it produce confident rubbish. Treating that as ignorance and explaining harder is the most reliable way to lose a room in its first hour.
The positions underneath the work are deliberately explicit, so your team can challenge them.
This stopped being an information problem a while ago. Nearly everyone in an office job has tried a chatbot by now. What's missing is a version that fits their actual work, permission to use it, and enough practice that it survives a deadline.
So the interesting question is never what to teach. It's what's stopping this particular group, and whether the days are spaced far enough apart to change anything.
The useful line isn't between what AI can and can't do. It's between where a wrong answer costs a minute and where it costs a client. Most company policies draw that line in the wrong place and then wonder why nobody follows it.
Judgement is the part that stays expensive. Reading a room, carrying a decision, knowing which clause the other side will actually fight over. That's fine, and it's exactly why judgement has to be taught rather than assumed.
Thirty minutes, no deck. Who's in the room, what they're stuck on, and what you need them able to do by when. If a day isn't what you need, I'll say so.