Starting with problems, not models
Good AI consulting begins with the work, not the technology. We interview the people who do a process, look at volumes, error costs and exceptions, and write each candidate use case as a testable statement: what goes in, what comes out, who checks it, and what a wrong answer costs. These interviews usually surface hidden steps and workarounds that never appear in official process documents.
That framing removes many ideas quickly. Some turn out to need only a rule or a better form. Others look promising but depend on data that is scattered or unlabelled. What remains is a short list where AI plausibly beats the current approach and where success can be measured within weeks. We also note who owns each process, because a use case without a committed business owner rarely survives past the pilot.
