A plan you can actually follow
From observation
to working change.
Specific actions, task-level work allocations, deliverables, dependencies, and decision gates. Expand any task to see exactly what to do.
Planning assumption: one designer/developer working 20 focused hours per week. These are allocations, not guaranteed durations or hours already completed. Recruiting, approvals, and retesting may extend elapsed time. Includes a coded portfolio prototype—not a production AI system or full company-site rebuild.
Why eight weeks? The earlier six-week plan covered design/research. This expanded plan adds a coded website concept and its interaction QA. Plan, prototype, research, and production are distinct stages.
Keep delivery realistic
What could change the schedule?
| Dependency / risk | Response |
|---|---|
| Participant availability | Start outreach in Week 1. Do not replace sessions with invented personas or results. |
| Content owner unavailable | Use source-linked summaries and clearly marked gaps. Do not invent cases, guarantees, or product relationships. |
| Testing reveals a blocker | Reduce decorative scope; spend the revision budget on the failed task. Carry unresolved work into a visible backlog. |
| Production AI is requested | Re-scope backend, security, privacy, evaluation, and ongoing operations. They are not included in this prototype budget. |
Your first working session
Start with Task 1.1 for two hours: define the exact pages and three visitor tasks. Then begin Task 1.2 for four hours: capture and label the current journey. Review the initial concept only as a hypothesis—not as evidence that the proposed changes work.
Track actual hours and deviations beside this plan. Never relabel the 160-hour budget as completed work. This page’s task details are ordered within each week; finish each decision gate before proceeding.