How would you answer an interview scenario involving LLM observability and cost control?

For an interview scenario involving LLM observability and cost control, I would first clarify the business goal, scale, constraints, and the failure or quality attribute the interviewer wants to explore. LLM observability tracks prompts, model responses, latency, errors, token usage, quality signals, and application-level outcomes. In this scenario, for a rapidly growing AI feature, explain how you would detect rising cost per request and identify which prompts or user flows are responsible. I would then explain the main alternatives and tradeoffs, identify likely failure modes, and describe how I would validate the solution through testing, observability, security controls, and recovery or rollback planning.