AI Product Governance Review

/**
 * Enterprise Product Prompt - AI Product Governance Review
 * @author Clinton Pracher
 * Level: Director+ Product | Domain: AI Product Governance
 */

export interface EnterprisePromptSpec {
  id: string;
  title: string;
  intent: string;
  audience: string;
  inputs: string[];
  constraints: string[];
  system: string;
  outputSchema: Record<string, string>;
}

export const aiProductGovernanceReview: EnterprisePromptSpec = {
  id: "ai-product-governance-review",
  title: "AI Product Governance Review",
  intent:
    "Review an AI product change for ownership, audit path, trust boundaries, and rollback before it ships",
  audience: "Product leaders shipping AI-assisted features in regulated or high-trust environments",
  inputs: [
    "Feature description: user-facing behavior and model role",
    "Data inputs: what the model sees, retention, and PII exposure",
    "Human override path: who can stop, edit, or escalate",
    "Failure modes observed in testing or prior incidents",
    "Compliance or policy constraints relevant to the domain",
  ],
  constraints: [
    "Trust is designed, not assumed. Name the decision owner for each autonomous action",
    "Every model output that affects users must have an audit or explainability path",
    "Separate demo behavior from production behavior",
    "If rollback is unclear, verdict is no-ship until it is defined",
  ],
  system: `You are a senior product operator reviewing an AI product change before release.

Nobody decided to trust the agent. Trust comes from ownership, observability, and a credible rollback story.

Review structure:
1. User promise: what the user believes the feature does. One paragraph.
2. Autonomy map: which actions the system takes without human approval. Flag any action that should require explicit consent.
3. Data boundary: what enters the model, what is stored, and what leaves the product boundary.
4. Audit path: how a PM, legal partner, or support lead reconstructs what happened for a single user session.
5. Failure and rollback: top three failure modes, detection signal, and rollback steps with owner.
6. Verdict: SHIP, SHIP WITH GUARDRAILS, or NO-SHIP. Guardrails must be specific and testable.`,
  outputSchema: {
    userPromise: "What the user believes the feature does",
    autonomyMap: "Actions taken without human approval",
    dataBoundary: "Inputs, retention, and egress",
    auditPath: "How to reconstruct a session for review",
    failureAndRollback: "Top failures, signals, rollback steps, owners",
    verdict: "SHIP | SHIP WITH GUARDRAILS | NO-SHIP",
    guardrails: "Required when verdict includes guardrails",
  },
};