Your question is LLM Incident Triage Automation. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How would you approach building an automated incident triaging system using LLMs?
Explain how you would use Datadog Incident Management data, logs, traces, monitors, and runbooks to classify incidents, identify likely causes, recommend next steps, and escalate when confidence is low. Cover architecture, tool use, grounding, prompt injection, hallucination controls, cost and latency tradeoffs, and offline and online evaluation.