Idea

• A stakeholder wants "AI for X" without a measurable outcome
• You need a one-page brief before allocating engineers or model budget
• Risk tier is unknown and must gate how much AI autonomy is allowed
• Multiple ideas compete and you need kill criteria, not a backlog dump

1. Write the problem in one sentence; reject ideas that only describe a model or a UI.
2. List users, jobs-to-be-done, and hard constraints (data class, latency, compliance, cost ceiling).
3. Define 2–4 success metrics that a human can verify in production within 30 days.
4. Write non-goals in the same document — what you will not build this cycle.
5. Assign an initial risk tier (internal tool / customer-facing / regulated data) using handbook/aidlc-roles-and-evolving-delivery.
6. Run story-kickoff.prompt to produce a structured brief; human edits until scope is stable.
7. Get sponsor sign-off before creating epics or opening a kickoff repo checklist.

• Build vs buy vs prompt-only prototype for the first slice
• Risk tier and whether AI agents may touch production data paths
• Which metric is the go/no-go for continuing past kickoff
• Owner accountable for scope changes (PM, not the coding agent)

• One-page brief with problem, users, metrics, non-goals
• Named sponsor and accountable AIDLC roles
• Initial risk tier recorded in the initiative change log
• Kill criteria if metrics are not met by a dated review

• PM: Own the brief, metrics, and non-goals; block coding until sponsor signs.
• Architect: Flag data-class and integration risks that change the risk tier.
• Engineer: Challenge vagueness; demand testable acceptance language before estimating.
• Engineering Manager: Confirm capacity and whether the initiative competes with committed work.

• Story kickoff

  Forces problem, users, metrics, and non-goals into a reviewable brief.

• Planning

  Structures initiative scope before any implementation plan exists.

None required at this stage — keep tools off by default.

• AIDLC roles & evolving delivery

  Maps who owns scope, risk tier, and sign-off.

• Skills vs prompts vs agents

  Stops teams from spinning agents when a prompt or skill is enough.

• Definition of Ready

  Exit bar for leaving Idea into Kickoff.

• Planner pattern

  Use only to structure the brief — not to invent product scope.

• Research agent

  Optional competitive or domain research with human synthesis.

• Sponsor signs scope, metrics, and non-goals
• Risk tier assigned before any AI coding or MCP enablement

• Generate brief scaffold via story-kickoff.prompt
• Link related cookbook examples once stack is known
• Create initiative change-log entry from the approved brief

• Map initiative to workload classification before model spend
• Record decision in an auditable change log, not chat
• If regulated data is in scope, involve security before kickoff

• Skipping non-goals and letting the agent expand scope
• Confusing a model demo with a product initiative
• Assigning risk tier after code already exists