What a weekly AI show-and-tell looks like
Half an hour, the same slot every week, three or four people each showing one thing they did with the tools. No slides about AI. It's the cheapest training a business can run and the quickest way to find the next job worth automating.

Why a short slot every week
A monthly hour gets cancelled when the month is busy, and turns into a presentation when it does happen. A weekly half hour at the same time survives, because it's too small to cancel and too soon to prepare for. People show what they actually did this week, which is the point. In our experience Friday morning works best – the week's work is fresh, and nobody has to present on a Monday.
The format
| Minutes | What happens |
|---|---|
| 0 to 2 | The lead reads out what happened to last week's items: built, parked, dropped. |
| 2 to 22 | Three or four turns of 5 minutes. The job, the brief, the result, the change, the number. Questions in the time, not after. |
| 22 to 28 | One failure, from someone who tried something that didn't work, and what they learned. The most useful slot of the lot. |
| 28 to 30 | The lead adds the week's items to the list and says which, if any, go to the gate. |
No slides, unless the slide is the thing that was made. The presenter shows their screen.
What people bring
- A brief that worked: the exact words they gave the tool for a job other people have too, so others can copy it.
- Something that went wrong: the invented figure, the wrong tone, the answer that looked right and wasn't, and how it got caught.
- A prototype: something built in the prototype lane, shown for 10 minutes, with the four gate questions asked in the room.
- A question: "has anyone got it to do this?", which usually somebody in the room has.
- A number: what a tool that's been running for a month actually saved or changed, from its owner.
What doesn't belong: a vendor demo, a talk about where AI is going, or anything over 5 minutes.
The three rules
Anyone can present, including the newest starter and the most junior role, because they often have the best small wins. Failures are welcome and get the protected slot, because a business where only the successes get shown learns nothing. And nothing shown is a commitment – showing a prototype doesn't mean it'll get built, and the person who showed it isn't responsible for it afterwards. The gate decides that, separately.
What the lead takes from it
The running list. After a quarter it shows which teams use the tools and for what, which briefs have spread, which prototypes keep coming back in different forms, and what the sceptics have found. The next agent to build, the next training session to run and the next rule to write are all on that list. It's probably the cheapest piece of research in the whole programme.