Economics & Markets
Why Your Service Costs Rise When Your Software Gets Cheaper
The parts of your business that never get more efficient get more expensive every year, and efficiency elsewhere is the cause.
2026-08-091 min read
A string quartet playing Beethoven needs four musicians and the same span of time the piece needed at its first performance. No process improvement in two centuries has reduced either number. Every other sector of the economy, meanwhile, has become dramatically more productive across those same two centuries, and the wages those sectors can pay have risen with their output. The quartet's players are not competing with 1800's labor market. They are competing with today's, and they have to be paid enough not to leave for it.
William Baumol and William Bowen named this in 1966, studying why performing-arts organizations ran perpetual deficits that no amount of management ever closed. Their answer was that the deficits were not a management failure at all. In a sector where output per hour cannot rise, wages still have to track the wider economy, so unit costs climb in real terms year after year while the product stays identical. Baumol spent much of the rest of his career tracing the same arithmetic through education, healthcare, live performance, and public administration — which are, not coincidentally, the sectors people describe as chronically and mysteriously expensive.
The business version is closer to home than it looks. Whatever inside your company still takes a person the same length of time it took five years ago — the implementation call, the escalated support case, the enterprise onboarding — is getting more expensive in real terms whether or not anyone is running it badly. Automating the software around it does not relieve the pressure; it adds to it, because it raises what your people could earn elsewhere. The line item that never improves is the one that eventually decides your gross margin.
Key insights
Wages in a sector with flat productivity are set by what those workers could earn in sectors with rising productivity, not by what their own sector produces.
This makes the real cost of unimproved work rise permanently — without inefficiency, mismanagement, or anyone being slower than they were before.
Productivity gains elsewhere inside your own company do the same thing internally: they raise the opportunity cost of every hour still spent by hand.
Why it matters
A cost line that rises every year without anyone doing anything wrong will be met with management pressure that cannot work, and the team blamed for it will be the one performing correctly.
Pricing built on the assumption that delivery costs fall the way infrastructure costs fall will erode margin quietly for years before it surfaces as a problem.
Use this tomorrow
1Pull your three largest human-delivered services — onboarding, implementation, escalated support, whatever they are for you — and write down the average person-hours each consumed this quarter alongside the same figure from two years ago, then count how many of the three have not fallen.
2Take next quarter's plan and count how many of its cost-reduction targets sit on tooling versus how many sit on hours a person actually spends; if the second number is zero, the plan does not touch the line that is actually growing.
Go deeper
Baumol and Bowen's 1966 study, Performing Arts: The Economic Dilemma, was commissioned to explain what looked like bad management — orchestras and theater companies whose deficits grew regardless of who ran them or how carefully. The finding reframed those deficits as a structural feature of any labor-intensive activity whose output per hour is fixed by the nature of the work itself. Baumol returned to it late in life in The Cost Disease, published in 2012, extending the argument to healthcare and education. That book makes a point usually missed in the popular version: cost disease is a symptom of an economy getting richer, not poorer, because the sectors that are racing ahead keep growing the total from which the stagnant ones are paid.
The sharper managerial question is which of your activities are genuinely stagnant and which only look that way. The test that follows from Baumol's framing is whether labor is the product or merely the current means of producing it: a musician's time is the thing being sold, whereas a bank teller's time was a way of moving money and could be replaced. Activities where the human presence is what the customer is actually buying — the trusted advisor, the in-person training, the account manager who carries the history — resist substitution by definition, and their price should be expected to rise. Activities where labor is merely today's implementation are the ones automation genuinely removes, and confusing the two categories is how companies automate the wrong half of their cost base.
