Get the framing right, and everything downstream gets easier.

I treat the brief as a hypothesis. Paradigm Health asked how AI could make their doctors' lives easier, and discovery found the problem was structural rather than technical: the care was excellent because specific people held it together, and it broke as soon as you added more. I learned that by becoming a patient, sitting inside the practice, and facilitating the workshops.

Workflows first, technology second. Before anyone argues about screens I want the shape of the service on one page, with everyone living inside it at once. The same question decides an AI product: the.garage answers with a decision, a trade-off or a plan the founder can act on, instead of a chat box.

Map every state, including the failures, then put the prototype in front of people and fix what they trip on. At Verizon that meant the whole eligibility system behind a limited card drop, and a rubric that scored the app portfolio so the apps that should not exist could be killed on evidence. I stay in the material too: I wrote most of the.garage's first production front-end.