# [Feature or product change] — Product requirements document (PRD)

Owner: [name]    Status: [draft / in review / approved]    Updated: [date]

## 1. Problem and evidence
- User and situation: [who is trying to do what, and when?]
- Observed problem and consequence: [what happened, and why does it matter?]
- Evidence (source and date): [interview, support issue, product behavior, or other signal]
- Current workaround: [what do people do today?]
- Facts and assumptions: [separate what is known from what is inferred]

## 2. Desired outcome
- Customer outcome: [what should become easier or better?]
- Success measure and baseline: [metric, current value, and target if justified]
- Review window: [when will the team inspect the result?]

## 3. Scope and constraints
- In scope: [smallest useful end-to-end change]
- Non-goals: [what this change will not solve]
- Constraints and dependencies: [technical, privacy, business, or operational]

## 4. Requirements and acceptance criteria
- R1 — [observable user or system behavior] ([must / should / could])
  Acceptance: Given [context], when [action], then [observable result].
- R2 — [observable behavior]
  Acceptance: Given [context], when [action], then [observable result].

## 5. Risks, edge cases, and safeguards
- Empty, error, and recovery behavior: [what should happen?]
- Privacy, permissions, accessibility, and performance: [relevant constraints]
- Rollout, support, or operational risks: [what needs a guardrail?]

## 6. AI behavior and evaluation (if this change uses a model)
- AI responsibility and approved context: [what may the model do, and which sources may it use?]
- Uncertainty and recovery: [when should it ask, abstain, or hand control back?]
- Evaluation: [representative cases, failure modes, and pass thresholds]
- Human review: [which outputs or actions need approval?]

## 7. Open questions and riskiest assumptions
- Question or assumption: [what could change the decision?]
  Owner: [name]    Evidence needed: [signal]    Decide by: [date]

## 8. Launch and learning
- Rollout: [how and to whom will this ship?]
- Events or checks: [what will be observed?]
- Guardrails: [what must not get worse?]
- After [review window], decide to continue, revise, or stop based on [evidence].
