What can a team practice before using AI on important work?
At the culture panel, the Cisco example was a playful mystery exercise, and I think it's a smart way to practice before AI touches real work. Give your team a made-up case where a wrong answer costs nothing. Then watch whether they check the AI's answer against the facts or just take its word for it.
Make up a small case with a handful of facts. A delivery landed in the wrong meeting room, say, and the team has to work out which room it reached. Give them everything they need to solve it, plus one detail that sounds important and isn't. The stakes are zero, since no customer or colleague has to absorb a wrong answer.
Have everyone use the approved AI tool to organize the evidence and propose an answer, and ask them to point to the supplied fact behind each conclusion. Then add a missing or contradictory detail. The thing to see is whether they ask for clarification or accept a confident answer the evidence can't support.
Talk it through afterward: which prompt helped, which answer sounded convincing but was wrong, and what would need to change before you'd try this on real work.
Keep it accessible. Nobody should have to share anything personal.