Support agent from documentation
Your support team wants an agent that answers common customer tickets straight from your product documentation before a human ever sees them. Design it for me: what does it read, what can it send, and when must it hand off to a person? Then tell me how you'd prove it's safe to trust — what would you measure before letting it reply to real customers? Close with a short pitch: why should the support lead want this?
route_support_ticket(ticket_text: str, retrieval_confidence: float, has_citation: bool) → str["How do I export a CSV report?",0.93,true]out"auto_reply"["I need a refund for last month",0.95,true]out"escalate_sensitive"["Where is the settings page?",0.42,true]out"escalate_low_confidence"Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.