Why My Best AI Workshops Have Thirteen People, Not Five Hundred
In May 2026, I led two workshops for the Asia arm of Europe's largest pet-supplies retailer -- a productivity session and a creativity session, thirteen participants in each. The kind of engagement that doesn't make a headline. But both sessions produced something that a five-hundred-person webinar never has in my experience: every person in the room built a working output with AI tools, and I could see every screen while they did it.
The industry measures training in reach -- how many people attended, how many sites were connected, how large the audience. Those numbers scale well and report well. It makes sense: reach is easy to count and looks decisive in a deck. But most of my work that results in actual workflow change happens in rooms this size. The mechanics of why are specific enough to name.
Everyone is visible
At thirteen, there is no back row. Nobody opens a second tab while the demo runs. Nobody watches for forty minutes, concludes "interesting," and leaves without touching a tool. Every participant has a name the trainer knows, a screen the trainer can see, and an output they're expected to produce before the session ends. The social accountability of a small room keeps people working without being large enough to create performance anxiety. In a large-format session, most people are watching. In a room of thirteen, most people are building. That gap between watching and building determines whether anything changes on Monday.
The session can stop for one person
When a participant hits a real problem -- a tool produces something useless, a prompt pattern doesn't transfer to their actual task, a workflow step breaks -- the session can pause. At thirteen, that pause costs thirty seconds. Walk over, look at the screen, adjust the approach. The rest of the room watches the troubleshooting, and that troubleshooting often teaches more than the exercise that was working. In a webinar, the same problem goes into the chat and gets a generic answer, if it gets one at all. The stop-and-fix moment is the most educational thing that happens in a hands-on AI session. It only works when the room is small enough to stop.
Participants use their own material
Thirteen is small enough to ask each person to bring a real task from their job -- a report they write every week, a supplier email, a brief they're working on. The exercise uses that material as input, not a fictional scenario about a made-up company. The output is something they take back to their desk as a working first draft, not a demonstration artifact. Both sessions I ran used this approach: participants applied AI tools to their own productivity and creative workflows, in their actual roles. What they built during the session was theirs to keep using.
I run sessions at both scales. Firmwide webinars have a real purpose -- shared vocabulary, permission signaling, awareness at a scope nothing else matches. But awareness and workflow change are different outcomes, and conflating them is how companies buy the wrong format. The sessions where someone actually starts using a tool differently the following Monday, in my experience, happen in rooms closer to thirteen than to five hundred. The tradeoff is real: thirteen is not a scalable number, and two sessions for one team doesn't produce a reach metric worth reporting. But every participant produced a real output, on their own material, with problems solved in real time. Whether that tradeoff is worth it depends on what you're buying the training to do.
