Use cases

Optimize a support agent only after checking action and escalation quality

OpenAI's Cookbook provides a synthetic support-agent optimization sprint comparing prompts, tool controls, model routing, caching, and offline follow-up; its default simulation is not a production performance benchmark.