Most AI training opens with a definition. We open with a blank wall and a marker.
The wall
Everyone writes down one task they did last week that felt like a waste of their life. Not a task they think AI could do — a task they resent. The wall fills up in about twelve minutes.
Then we sort. Which of these are repetitive? Which require judgement? Which require a relationship? Which require a signature?
By the end of ninety minutes the room has produced its own curriculum, and it is always better than mine, because it is about their actual Tuesdays.
Why this matters
Because a definition of AI is forgettable and a list of your own resentments is not. The rest of the two days refers back to that wall constantly — "remember the invoice one" is a much better teaching tool than any framework.
It also does something subtler: it makes the room responsible for the content. People defend what they wrote. They argue about whether a task needs judgement. That argument is the learning.
The cost
It is slower. A definition takes four minutes; the wall takes ninety. If you are buying a lecture, this will annoy you.
And it requires a facilitator who can let silence sit, which is harder than it sounds when forty people are looking at you.
What comes after
Only once the wall exists do we touch a tool. By then the first thing anyone builds is a task they already hate, which means they have a reason to care whether it works — and a way to tell whether it did.
That is the whole design. Everything else is logistics.
