Start with a problem.

Pick a time and walk us through the part of the operation that keeps getting stuck. The first conversation is a working session, not a demo. We look at the shape of the problem, what evidence already exists, and whether there is anything worth examining further.