The guillotine was designed to make execution faster and more efficient. But the experience of revolutionary Paris offers an unexpected analogy for the economic effects of artificial intelligence: greater efficiency does not necessarily reduce demand for a service. It can increase it. The comparison comes from an analysis published by the Observer UK, drawing on Hilary Mantel’s novel A Place of Greater Safety and its depiction of the French Revolution’s executioners.
Mantel describes the first execution by guillotine, that of the highwayman Nicolas Jacques Pelletier, as taking only about five minutes. The new machine was convenient and apparently humane, but its efficiency disturbed the executioners because it appeared to remove the skill from their profession. “It looks so easy; people will be thinking that there is no skill in it, that anyone can be an executioner,” the executioner reflects in Mantel’s account. Later, as executions multiplied during the Reign of Terror, the profession’s transformation became still more pronounced. “It is not an art any longer, it is more like working in a slaughterhouse,” the executioner says.
The historical point is not simply that technology can replace labour. The guillotine dramatically reduced the time required for an execution, yet Parisian executioners did not disappear. The Revolutionary Tribunal continued sending large numbers of people to their deaths, meaning that the increased efficiency of the process helped accommodate a much greater volume of executions. The example illustrates the Jevons paradox, in which a technology that makes a resource or activity more efficient can lead to increased rather than reduced consumption of it.
The same mechanism could apply to artificial intelligence. If AI enables software engineers, video producers and legal professionals to produce services much more efficiently, lower costs could stimulate enough additional demand to expand the overall market. If demand is sufficiently responsive to price, the result could be increased employment in sectors widely expected to suffer from automation. The argument, developed in greater detail by economists and writers including Alex Imas, Nathan Witkin and David Oks, provides one reason to question straightforward predictions of an AI-driven collapse in white-collar employment.
But the Observer UK analysis identifies a different and potentially more immediate consequence of automation: the effect on the character of work itself. The guillotine did not merely accelerate execution. It changed what an executioner needed to know and do. Skill, craftsmanship and professional prestige were diminished because the machine performed the central technical task.
A similar transition is already visible in some modern professions. Kazra Kyanzadeh, who briefly worked as a software engineer for OpenAI, left the company saying that “building commercial software doesn’t feel fun anymore”. Writing on X, he argued that when AI produces most of the code, software engineers spend less time collaborating with colleagues and more time reviewing machine-generated work. For him, the change made the job more mentally draining. The modern software engineer, in this account, risks becoming less a creator of code than a supervisor of code produced by a machine.
Translation provides another example. Sarah O’Connor of the Financial Times has written about how AI can remove some of the intellectual work involved in translating television subtitles. Previously, a translator had to consider culturally specific jokes, appropriate forms of address and other linguistic problems requiring judgement and creativity. When AI produces the initial draft, the translator may instead spend much of the time scanning rapidly for mistakes. Productivity increases, but some of the satisfying and intellectually demanding elements of the profession are reduced.
The argument does not amount to a case against technological progress. Rather, it shifts attention from the number of jobs affected by AI to their quality and composition. A profession can survive while becoming substantially different, with machines taking over the tasks that previously provided workers with expertise, autonomy or creative satisfaction.
The Observer UK analysis places that debate alongside two significant developments in the UK economy. The Bank of England’s Monetary Policy Committee voted 6–3 to keep its policy rate at 3.75%, while announcing a major change to its quantitative tightening programme. Instead of setting a new annual sales target each September, the Bank plans to reduce the stock of gilts held for monetary policy purposes to zero by 2034, while retaining £120bn of the longest-dated gilts to back banknote issuance. Active sales will fall to about £20bn a year from an average of £32bn, with the bonds sold to the Debt Management Office rather than private investors.
The Office for National Statistics has also revised its assessment of British labour productivity. Its new estimates suggest productivity grew by an average of 1.3% a year between 2009 and 2019, rather than the previously estimated 0.7%. A major reason is a revised assumption about annual leave: the ONS is now assuming workers take their full entitlement, reducing estimated hours worked and therefore increasing calculated output per hour. The change alters the picture of Britain’s post-financial-crisis productivity performance, while leaving open a separate question at the heart of the AI debate: not simply how much more efficiently people can work, but what happens to the work itself when machines take over the tasks that once made it skilled.
