Keeping a mutual help community alive

After the momentum of the launch, usage drops off if nothing feeds it. Etienne Wenger described communities of practice: people who share a domain, exchange regularly and build together a repertoire of ways of doing things. That is exactly what the use of AI needs.

• Champions: one or two volunteer colleagues per team, whom others can ask for help.
• A ritual: fifteen minutes a week, where everyone shows a request that worked or an instructive failure.
• A library: the requests that work, kept in a shared project, with their usage tip.

The ritual must stay short and concrete. People show a real result, the request that produced it, and what they had to correct. A failure well told is worth as much as a success, because it saves others from making the same mistake.

1. Create an “Our requests” project and invite the whole team.
2. For each request that works, note the task, the full request and the tip that makes the difference.
3. When a request is used often, suggest turning it into a shared assistant.
4. Once a quarter, remove the requests nobody uses any more.