In 1865 a young English economist named William Stanley Jevons noticed something that did not make sense. James Watt’s improved steam engine used coal far more efficiently than the old ones. Common sense said Britain would therefore burn less coal. Instead, it burned more. A lot more. British coal consumption roughly tripled over the second half of the century.1
This is the Jevons paradox, and once you see it you cannot unsee it in your own working week.
Why does efficiency increase consumption?
Because making something cheaper to use makes it worth using in more places. The efficient engine made coal-powered work economical for industries that could never have afforded it before. So demand exploded, and the gain in efficiency per unit was swamped by the rise in units. The saving was real. It just unleashed an appetite larger than itself.2
Economists call the modern version the rebound effect.3 Make cars more fuel-efficient and people drive more, recovering part of the saving. Make computing cheaper and we do not use less of it; we invent entirely new things to compute, until data centres draw a measurable share of world electricity. Capacity, once cheap, gets used.
How does this eat the time AI saves?
This is the exact mechanism by which AI-saved time becomes AI-inflated workload. When drafting an email drops from twenty minutes to two, you do not send the same emails in less time and leave early. You send more emails, longer ones, to more people, because each now costs almost nothing. When a deck takes an hour instead of a day, the expectation quietly becomes more decks. The per-task saving is genuine. It vanishes into a higher volume of tasks.
A telling moment came when a cheaper AI model was released and a major technology chief executive remarked, in effect, that cheaper intelligence would not reduce demand for it but multiply it.4 That is Jevons, stated by someone who stands to benefit. Cheaper output does not mean less work. It usually means more.
So is efficiency pointless?
No. The gain is real and worth having. The danger is assuming the gain converts itself into free time. It does not. Left alone, efficiency feeds an appetite for more, and you end up running faster to stay in the same place, with a fuller plate and a vague sense that the tools were supposed to help.
How do you escape it?
There is only one defence, and it is deliberately counterintuitive: cap your own output. Decide, before you begin, what “enough” looks like for a given piece of work. Then produce that, and stop. Bank the saved time as time rather than letting it flow automatically into a longer report, a fancier deck, or a fourth option nobody asked for.
The Jevons paradox is what happens when you let efficiency run unchecked. An output cap is the check. It is the difference between a tool that buys you an afternoon and a tool that buys your employer more output for the same afternoon. Same speed-up. Opposite result. The only variable is whether you decided in advance where the saving would go.
Frequently asked questions
What is the Jevons paradox?
The Jevons paradox is the finding that making a resource more efficient to use can increase its total consumption rather than reduce it. Jevons observed in 1865 that more efficient steam engines led Britain to burn more coal, not less, because cheaper power created new uses for it.
How does the Jevons paradox apply to AI?
When a task becomes cheaper to do with AI, we tend to do far more of it rather than pocketing the saving. The efficiency vanishes into extra output, so AI-saved time turns into AI-inflated workload.
How do you beat the Jevons paradox at work?
Cap your output. Decide before you start what enough looks like for a piece of work, produce that, and stop, so the saving is banked as time rather than flowing into more volume.
About the author
Tom Goodwin is the author of Don’t Work Harder, a book about taking the time AI gives back as time rather than more work. He is a co-founder of GAMEPLAN and writes on productivity, technology, and the economics of the working week.
Footnotes
Footnotes
-
William Stanley Jevons, The Coal Question (London: Macmillan, 1865). British coal consumption roughly tripled over the second half of the nineteenth century. ↩
-
Jevons, The Coal Question. ↩
-
For a survey of the rebound effect, see Steve Sorrell, The Rebound Effect (UK Energy Research Centre, 2007). Economists distinguish partial rebound from full backfire, where consumption ends up higher than before. ↩
-
Satya Nadella, post on X, 27 January 2025, following the release of the DeepSeek model; reported by Fortune and NPR’s Planet Money, February 2025. ↩