Fears about machines destroying jobs are almost as old as machines. Since the Industrial Revolution every major new technology, from the steam engine to agentic AI, generates new worries.
In an influential paper written in 2013, two Oxford academics, Michael Osborne and Carl Benedikt Frey, estimated that up to 47% of US jobs were at a high risk of automation over the next 10 to 20 years. Their paper speculated that the application of machine learning, mobile robotics and big data would cause significant job losses. The introduction of ChatGPT in 2023 has prompted new and alarming forecasts of job losses to come.
Pessimism is the norm in discussions about jobs and machines. But if the forecasts are gloomy, what do the hard data, on job creation and destruction, tells us?
My colleague, Tom Avis, has analysed the changing shape of the UK job market between 2004 and 2024, mapping the expansion and contraction of over 400 separate job types. This period, encompassing the introduction of the smartphone, cloud computing, social media, e-commerce and an accelerating trend in automation and AI, enables us to gauge how the labour market responds in times of technological change.
Machines have destroyed large numbers of jobs. 700,000 manufacturing jobs and 600,000 roles in secretarial occupations were lost over this period. Technology has cut a swathe through work in banks and post offices, betting shops, libraries and travel agents.
Yet, far from shrinking job opportunities, technological change in the last 20 years has been accompanied by rising employment. 3.5m jobs have disappeared in the last 20 years and 8.5m jobs have been created. Roughly 4.5m new jobs were added to the UK labour market between 2004–2024, an increase of 16%. (The official data enables us to track jobs by employment type, but not to track the birth and death of each job.) Crucially, the new jobs were, on average, more skilled than the jobs that have been lost.
Estimates of the future impact of technology tend to underplay the role technology plays in creating work.
The number of ‘information technology professionals’ in the UK, for instance, rose from 0.6m to 1.4m between 2004 and 2024. Many technologies augment, rather than replace, humans, raising their productivity and value. Rapid growth in numbers of ‘knowledge workers’, from scientists to engineers and researchers, reflect such augmentation effects. Regardless of whether technology destroys or creates jobs, it raises productivity which, in turn, raises incomes. The broadest measure of welfare, real household disposable income per capita, increased by 19% between 2004 and 2024 in the UK. Rising incomes have facilitated the growth of new demand and new jobs. This can be seen in the tripling in the number of coffee shop workers, the doubling in the number of fitness instructors and beauticians and a 160% increase in the number of vets since 2004.
Technology is but one factor shaping the jobs market. Discussions about the future of work often underplay the role of globalisation, demographics, competition and politics in the creation and destruction of jobs.
Globalisation, coupled with low-cost communications, have led to the offshoring of service, sales and back-office jobs. Demographic change has driven growth in the number of social and care workers. Competition and the search for economies of scale in supermarkets have been major factors in the shrinkage of the High Street. The NHS has been a political priority for successive governments, helping lift health employment to record levels. Now, with defence seen as an urgent priority for the first time since the end of the Cold War, the size of the armed forces seems set to expand after four decades of shrinkage.
The story of work is too often seen as one of machines taking jobs. The reality is infinitely messier, with continuous job destruction and creation caused by a multitude of factors that interact and change over time. Amid this tumult two enduring and related trends are discernible - the contraction of routine and manual work and the rise in work involving complexity and judgement.
Manual, less skilled service and manufacturing jobs have been in retreat since the War. We can see it in the rapid shrinkage of occupations such as labourers, agricultural workers, telephonists and data entry workers in the last 20 years.
The growth has been in jobs involving complexity, uncertainty and judgement. Since 2004 the number of cyber security experts has risen ninefold; the number of actuaries, economists and statisticians has nearly tripled; the number of psychotherapists and cognitive behaviour therapists or conservation and environmental professionals has more than doubled. Work in PR, the arts, as nurse practitioners, and in social care, has boomed.
Over time machines have liberated humans from dangerous, physically exhausting, repetitive work. Machines have enabled humans to stop behaving as machines – and to take on more unstructured, creative work, often involving human contact and judgement. It has been a good exchange.
Many experts think that AI will be different, that its capacity will reach so far up the cognitive scale that it will take over the unstructured, creative work which has been the engine of job creation this century. I, however, am disinclined to bet against the lessons of history, and of the last 20 years.