Will AI trigger a Jevons paradox in the job market? Maybe.
William Jevons was the 19th century economist credited with starting the mathematical method in economics. He concluded that by making a process more productive, people use it more, not less.
You may reasonably be asking, will work, being made more efficient by the arrival of artificial intelloigence (AI), reduce the need of humans (contraction of the workforce), or do the opposite? mWilliam Jevons published “The Coal Question”, where he documented the effect of James Watt’s steam engine, which led to what is now known as Jevons paradox.
Nowadays, if AI makes cognitive work dramatically cheaper, demand for it will grow so much that total employment will rises, not fall. It means that as companies will pay less for talent that currently goes at a premium, job opportunities for the trained will be plentiful. For example, an hisotrical episode is the introduction of Automated Teller Machines / Cash Points in the 1970s significantly reduced the number of human tellers required per bank branch. However, because operating a branch became dramatically cheaper, banks opened many more branches, leading to a net increase in total number of tellers employment over the following decades. Also, digital search tools made sifting through legal evidence vastly cheaper. Rather than reducing legal jobs, the lower cost allowed legal teams to examine millions of documents instead of thousands, increasing overall demand for paralegals and legal analysts. This is called “augmentation”. It means that cheaper technology augments human labor needs. However, economic history suggests that it isn’t automatic. The Jevons paradox requires that demand for the output (goods or services) to be price-elastic (|E_d |>1). If demand for the output is inelastic, lower costs of production may lead to workforce reduction rather than job growth. This is because at some point demand will reach saturation. For example, the widespread mechanization of the 19th century that made the textile industry more efficient eventually shed most of the old jobs once demand for textiles saturated. In agriculture, mechanization made food far cheaper that ever before, but people can only eat so much. Therefore, farm employment collapsed. Jevons works when demand does not saturate, and stops working when it does.
Get Dr. Hernán López-Schier’s stories in your inbox
Join Medium for free to get updates from this writer.
Salary dynamics, however, are disconnected. This is because salaries depend on both wage elasticity of labor supply and labor market competition, and not in supply and demand for goods or services. If a efficiency boost triggers a massive Jevons rebound in demand for a service, firms will hire more workers. Therefore, depending on labor availability, total employment increases while wages remain relatively stable, or the sudden surge in demand for specialized talent can actually drive salaries up, despite the reduced cost per unit of work produced. Thus, whether AI expands the job market will likely depends on three questions. First, is there an unsaturable appetite for goods and services? Second, will AI handle tasks within jobs? If so, more output means more jobs. Third, how fast things will change? Even if transitions initially generate joblesness, they will create new work over years and decades. Early evidence on AI and employment is still limited and mixed, but an International Labour Organization’s June 2026 review finds meaningful productivity effects but still limited evidence of large aggregate employment displacement. Effects differ considerably by occupation and worker group
My guess is that AI will not produce one economy-wide Jevons paradox. Instead, it will generate thousands of sector-specific experiments with currently unpredictable outcome. Wherever AI lowers unit costs faster than it substitutes human labor, and demand expands faster than productivity, employment will increase. Wherever demand saturates or AI directly substitutes for labor faster than markets expand, employment will contract. The defining labor-market question of the AI era therefore is not simply how many tasks AI can automate. It is whether AI can create demand, as well as new human tasks, faster than it destroys the old ones. Which sector do you think has the most unmet demand waiting for cheaper expertise?
