Every business contains two halves that obey different physics. One half gets cheaper to run each year, because software and scale keep taking work out of it. The other half does not, because the work is the people doing it and there is no version with fewer of them in it. Economists named this sixty years ago.

William Baumol and William Bowen were trying to explain why orchestras ran chronic deficits while playing to full houses. Their answer, published in 1966, was that a string quartet needs the same four musicians and the same half hour to perform a piece as it did a century earlier. Performance productivity is flat, permanently. But those musicians live in an economy where pay rises with productivity everywhere else, and if quartet players earn far less than their talents would fetch elsewhere, they stop being quartet players. So the cost of an hour of live music must rise relative to everything else, forever, with nobody doing anything wrong. They called it the cost disease, and it appears wherever the unit of output is a person's attention.

Your quartet is whatever work resists being automated out: the support case that needs someone to understand the customer's actual setup, the implementation call, the security questionnaire, the enterprise deal that closes in a room. As the engineering half accelerates, that human half does not merely stay expensive — it becomes a larger share of every dollar, and it looks worse in every review, because it is being judged against a half that keeps improving. The predictable response is to demand efficiency from it, quarter after quarter, and to read the failure as a people problem. It is not a people problem. There are two honest moves: change the product so the task stops existing, or price the line explicitly and stop planning as though it will fall.