From a useful question
to a working prototype.
Build one functional flow with AI, test it thoughtfully, and explain your decisions. Work at your own pace; the time ranges are planning estimates.
Frame one useful task
Write who is trying to do what, their starting context and the observable finish. Choose one uncertain assumption.
Use the Decision Checklist to separate evidence from a guess. For a booking flow, “choose an available appointment and understand the next step” is narrower than “build a scheduling platform.”
Your deliverable
A one-paragraph brief, a success condition and a list of exclusions.
Map the flow and its states
Draw the shortest credible route and the moments where it can fail.
For the booking example: choose slot → review → confirm. Add unavailable slot, missing required information, loading, failure and cancellation. Mark which behaviors will be simulated.
Your deliverable
A state map and screen/content inventory.
Build a functional slice with AI
Use the prototype prompt with your brief, states and fictional data. Ask for a small runnable implementation and setup instructions.
Start with local state. Simulate availability and confirmation rather than connecting a real calendar. Label the whole experience as a demo. Review generated code and dependencies before running or sharing it.
Your deliverable
A runnable flow with an explicit simulation notice.
Test the experience
Verify behavior yourself, then observe intended users doing a neutral task.
Check the keyboard, visible focus, labels, narrow viewport, enlarged text and recovery. Ask what participants think happened after confirmation. Get appropriate consent and keep research notes private.
Your deliverable
A task guide, observations with sources, prioritized issues and revised flow.
Tell the story and plan the handoff
Explain the problem, evidence, alternatives, your role and next decision.
A working prototype is not a production-ready application. Document simulated behavior, data/security needs, remaining accessibility evaluation and engineering decisions. Do not turn expected benefits into invented outcome metrics.
Your deliverable
A short case study plus a production-readiness question list.