How do you forecast what an AI feature will actually cost — before you build it?

While building (at drawing board) you have some per-call estimate, then you compose and test. Then staging gives you estimates per run (optimistic). Basically, if subjective complexity factor (0-5) * cost per run * 50 < customer cost per run… is the point where you ask your self 2 questions: 1. Can you divide the costs at least by 2 (better 4)? 2. Is it worth bothering?