sent it to engineering the way you’d send a Figma file. Engineering read it as a proposal for how to build the thing. They spent most of a week on component structure, state handling, and why it diverged from the architecture. She meant a sketch. They received a spec. Nobody was wrong and it still cost us a week. THE PROBLEM
the system. It reproduces the inconsistency faster, at scale, in production. Levels three and four depend entirely on how mature your design system is. That problem gets its own piece next time. HANDS OFF TO PART THREE
evidence sooner? Did review time drop? Did engineering rework go down? Did quality improve? What did the tools cost? What can the team reuse next quarter? SIX QUESTIONS
HYPOTHESIS FIRST Genuine customer Small enough that Design, product and Written down before value, manageable risk. failure is not a crisis. engineering together. anyone starts. "Everyone should be using Cursor this quarter" is a press release. HOW TO BEGIN
holding. A shared process Better design system coverage Clearer decision rights Reusable context for agents Known failure patterns An agreement about what design owes engineering WHAT SURVIVES THE PILOT