Included:
• Workflow and risk discovery
• Small production-ready interface and API
• Prompt and structured-output design
• Human review, confidence, and failure states
• Logging, evaluation cases, and handoff
• Deployment with clear model/provider boundaries