What Am I Really Paying For? The Hidden Cost of AI, Seen From the Bottom Up

Two years ago the question about AI was whether it could do the work. Today it is whether we can afford it. The cost of AI flows down a chain, from the companies building data centers to the companies building models to the businesses and people using them. Those at the end pay the bill but cannot see what they are buying: plans promise "5×" or "20×" more usage without saying more of what, and vendors earn more whether or not that usage produces anything useful. Using several AI systems from several vendors is now essential; the question is how to use them effectively and efficiently. I set out a framework for doing so, the Staffing Model for AI, which staffs models the way organizations already staff people. The measure that matters to me is useful, working results per dollar. The rest of the paper answers the three questions executives ask: what am I getting for my dollars, am I overpaying, and is there a better, cheaper alternative? It also covers how to train staff to use these systems without wasting money.