How We Reduced Our LLM API Costs by 60%: What Actually Worked

At some point in most of our production AI projects, someone looks at the monthly API bill and asks whether we can do something about it. The answer is always yes — but the specific answers vary a lot...