A pilot should do more than demonstrate that a new technology can produce an interesting result. It should help a team make a better decision about what comes next.
The smallest useful pilot has a clear owner, a defined workflow, and a short feedback loop. It creates enough contact with reality to reveal where the idea helps, where it creates new work, and what support people need to use it well.
Start with the smallest experiment that can change your mind. That is often where the most useful learning begins.