The question of whether artificial intelligence will primarily destroy jobs or create new ones has become one of the defining economic debates of the decade. The answer, as with most genuinely important questions in economics, is deeply uncertain and depends critically on the pace of technological change, the adaptability of workers, and the institutional environment in which these transitions play out.
Historical Analogies and Their Limitations
Optimists frequently invoke the agricultural and industrial revolutions as evidence that technological disruption ultimately generates more employment than it destroys. The mechanisation of farming displaced millions of agricultural workers over the course of a century, yet living standards rose dramatically and new industries absorbed the displaced workers, albeit after periods of significant hardship and social dislocation.
The key difference this time, pessimists argue, is that AI threatens cognitive labour in a way that previous technologies did not. Agricultural mechanisation replaced muscle power; the assembly line replaced routine manual dexterity. But if AI can replicate complex reasoning, analysis, and even creative work, the range of tasks susceptible to automation expands far beyond anything seen in previous technological revolutions.
Knowledge workers in a modern office environment. Photo: tommao wang / Unsplash
"The worry is not that machines will replace all human labour, but that the transition will be faster than institutions can adapt, leaving millions of workers stranded between the jobs that have disappeared and the jobs that have not yet been created." - Daron Acemoglu, MIT, 2024
What the Evidence Actually Shows
Recent empirical work has complicated both the optimistic and pessimistic narratives. A 2024 study by Acemoglu and Restrepo found that AI adoption has so far been concentrated in a narrow set of tasks and sectors, with limited aggregate employment effects. However, the same study noted that the pace of adoption has been accelerating rapidly, and that historical patterns may not be a reliable guide to the current transition.
A separate analysis from the McKinsey Global Institute estimated that between 75 million and 375 million workers globally, roughly 3% to 14% of the global workforce, may need to change occupational category by 2030 due to automation. The enormous range of that estimate captures how uncertain the projection genuinely is, and how sensitive it is to assumptions about adoption rates and policy responses.
The Policy Question
If the pessimists are even partially right, the policy response matters enormously. The historical record suggests that technological transitions produce better outcomes when active labour market policies, retraining programmes, and social insurance systems are strong enough to support workers through the adjustment. Countries with more developed welfare states and more responsive education systems have generally managed previous technological transitions with less inequality and less political disruption.
For the UK in particular, the challenge is acute. Decades of underinvestment in further education and adult retraining mean that the institutional capacity to manage a rapid technological transition is weaker than in many comparable economies. The question is not whether AI will change the labour market. It will. The question is whether policymakers act early enough to shape the transition.
Sources
Acemoglu, D. and Restrepo, P., "Tasks, Automation, and the Rise in US Wage Inequality." Econometrica, 2022.
McKinsey Global Institute, "Jobs Lost, Jobs Gained: Workforce Transitions in a Time of Automation." mckinsey.com, 2023.
Autor, D., "Work of the Past, Work of the Future." AEA Papers and Proceedings, 2019.
OECD, "OECD Employment Outlook 2024: The Net-Zero Transition and the Labour Market." oecd.org, 2024.
Goldman Sachs Global Investment Research, "The Potentially Large Effects of Artificial Intelligence on Economic Growth." March 2023.
