Automation Will Create More Jobs, But Most Will Be Shit

by Matthew Cole

5 October 2020

A black and white photograph of people sitting at a bank of computers
Clemson University Library / Flickr

In 2013, the academics Carl Frey and Michael Osbourne predicted that up to 47% of American jobs were at high risk of automation over the next two decades. Their paper provoked automation hysteria, reflected in pronouncements of the “second machine age”, “fourth industrial revolution” and “industry 4.0”. Seven years on, the reality appears far less dramatic.

Predictions of mass job losses as a result of automation have, it transpires, greatly overstated the job-destructive impact of new technologies, and understated their ability to augment existing jobs and create new ones. 

Critics of the so-called “automation discourse” point to the stagnant productivity and lowered labour demand. However, not only does slowing productivity and growth alone fail to explain low labour demand; they tell us virtually nothing about the actual qualitative effects of automation on work and society. Up until the coronavirus pandemic, there had not been an overall fall in labour demand. Despite the fallout from the global financial crisis, from 2010 and 2015 total employment among member states of the Organisation for Economic Co-operation and Development (OECD) grew by 4.9% or roughly 27 million jobs. Canada, Mexico and the United States added 12.9 million jobs; the European Union, 3.6 million. While increases in labour demand have slowed in the EU, they have risen significantly in non-EU economies.

Nor should we be surprised by the sluggishness of productivity, increases in which generally require the diffusion of new technologies, which takes time; computers took nearly 25 years to reach their plateau of 5% of non-residential equipment capital. Recent OECD analysis shows that the pace of digital transformation varies hugely between countries and sectors: firms in the real estate, hospitality and construction industries are investing in software and other information and communication technologies (ICT) much less intensively than advertising, finance and law, for example. Just as diffusion varies across sectors, there are significant differences in digitalisation among firms of different sizes. On average, large firms are more digitally intensive than their small and medium-sized counterparts, even in the same sector. For example, 28% of large firms utilise big data analysis, while only 16% of medium-sized and 9% of small firms do so. The divergence in the uptake in digital technologies corresponds to a divergence in productivity between leading and lagging firms.

If we are not facing a jobs apocalypse from overproduction or automation, what impact is technological change actually having on work? Much the same as it has always had. While computers have automated tasks since the 1950s, very few of the 271 occupations listed in the 1950 US census no longer existed in 2010 (though some were grouped differently). Labour demand in some occupations declined due to both technological obsolescence (telegraph operators to typewriter servicers) and full automation (elevator operators); yet very few occupations were destroyed. What this shows is that automation tends to be partial, changing the task composition of most jobs rather than eliminating them entirely. 

The management consultancy research that has dominated “automation discourse” tend to portray this in a positive light, assuming that routine, low-skilled tasks will be replaced by high-skilled, high-quality jobs. However, the idea that technological change favours high-skilled labour obscures a more complex relationship between technology and skill. While AI- and data-driven digital automation in advanced capitalist countries have generally created more higher-skilled and higher-income professional jobs, they have also created more in-person service jobs that require lower levels of education and training. Meanwhile, routine, rule-based, middle-income work has been automated leading to job polarisation, hollowing out the demand for middle-skilled jobs. Evangelists of the “fourth industrial revolution” tend to ignore the negative effects technological change can have on workers, particularly those at the bottom of the pile.

The transformational impact of automation on jobs has more to do with changes to the task composition of jobs and the quality of work. Most jobs are comprised of a variety of tasks with varying degrees of susceptibility to automation. Many include a significant component of non-automatable tasks; for example, while AI can outperform humans at large-scale image analysis, these technologies often fail miserably at tasks that require common sense or emotional intelligence. 

An automation discourse that focuses on a workless future is a red herring. In the long term, the crisis isn’t one job destruction – it’s job transformation, and of a kind that is detrimental to (particularly low-income) workers. It’s the incentivisation of cost-cutting approaches to labour, including deskilling, surveillance and intensification, all of which result in lower worker discretion and poorer quality of work. Low-income and lower-skilled workers will bear the brunt of such changes as they are at the highest risk of automation displacement. If they lose their jobs, they will unlikely be able to retrain, and so will move into other similarly routine jobs, which they will lose again.

The pandemic will likely only accelerate these trends. Managers have been forced to adopt platform technologies and remote working they have historically resisted. As a result, the pandemic has accelerated the digitisation of all sectors, opening up an unprecedented threat to workers from datafication, digital automation and surveillance technologies. At the same time, the labour market has been decimated both in the US and EU, with real unemployment rates nearly double the official statistics. The full effects of such technological changes will reflect how power is distributed in society and thus depend as much on the increasing sophistication of automation as on the actions of workers or policymakers. In this context, it is all the more imperative to end the proliferation of low-waged work, increase social protections, invest in social infrastructures and fight for collective ownership of the means of production, platforms and all.

Matthew Cole is a postdoctoral research fellow at the University of Leeds, and will soon be joining the Oxford Internet Institute. He is also a research associate at Autonomy, a think tank focusing on the future of work.

  • The Future of Work focus is part of Novara Media’s Decade Project, an inquiry into the defining issues of the 2020s. The Decade Project is generously supported by the Rosa Luxemburg Foundation (London Office).

We’re up against huge power and influence. Our supporters keep us entirely free to access. We don’t have any ad partnerships or sponsored content.

Donate one hour’s wage per month—or whatever you can afford—today.

We’re up against huge power and influence. Our supporters keep us entirely free to access. We don’t have any ad partnerships or sponsored content.

Donate one hour’s wage per month—or whatever you can afford—today.