Automation and crisis: Arguing the future

This is an adapted extract from A World Beyond Work? Labour, Money and the Capitalist State Between Crisis and Utopia, forthcoming with Emerald in January 2021.

The main question confronting analyses and debates about automation today is whether this time is different. There have been many predictions through time that automation will lay waste to the economy, displace jobs and tear apart the labour market, with all the social and political consequences that implies. Evidently, automation on a mass scale has not happened yet, despite automation being a consistent part of the organisation and reorganisation of work in capitalist society, but is there anything about our own time that potentiates a different outcome? And what might the economic fallout from the COVID-19 fallout hold for the fortunes of what Aaron Benanav terms the ‘automation discourse’?

The automation discourse is a way of talking about and thinking about the future of work in capitalist society that periodically rears its head at times of crisis and uncertainty. However, it is not so easy as to merely dismiss this discourse as a kind of false consciousness uprooted from reality. As Jamie Morgan argues, the way that people talk about the future has a performative effect, discursively constructing a reality that takes flesh in the actions of governments and firms. Both Sam Kinsley, as well as Andrew Sturdy and Glenn Morgan, have shown how the ‘fourth industrial revolution’ is summoned up by consultants and other bodies like the World Economic Forum, McKinsey, the Boston Consulting Group, Pricewaterhouse Coopers and Deloitte, its effects forecast through contested forms of statistical and anecdotal evidence and numerous airport bestsellers, and the expectations of an inevitable future used to structure how firms are advised to reorganise their workplaces and workforces, how investors are advised to allocate their wealth, and how governments create policies and incentives to regulate and promote the desired outcomes.

This is not a conspiracy theory, cooked up by consultants to sell clients problems and the solutions that follow – rather, the question is why this discourse should lie so close at hand now, at this specific moment in time. What are the material bases for the fact that capitalist development should present itself as an accumulation of technological potentialities leading seemingly inexorably to an eradication of jobs? Benanav suggests that it can be seen to relate to worries about a widespread low demand for labour manifested in unemployment and underemployment, and a decline in the share of income labour receives in the economy as a whole. But because these phenomena never quite disappear, and recur in different guises throughout time, so too does the automation discourse persist, assuming new forms every two decades or so. As we will see, this often has less to do with incipient technological possibilities as does it with contradictions in capitalist political economy, and in particular the longstanding pull of secular economic stagnation in the second part of the twentieth century and first part of the twenty-first. COVID-19 may have appeared to change everything, but its capacity to change this state of affairs is contingent on a range of political and economic factors.


The automation discourse has been around as long as industrial capitalism even began its first rumblings with the waning of feudal society. With this began what Brynjolfsson and McAfee call the ‘first machine age’, wherein work was typically organised around a strict division of labour with tasks fragmented and standardised, rendering it susceptible to automation. As we shall, whilst this did occur in some sectors and industries, it seldom lived up to the worst fears or best hopes of techno-dystopians and techno-utopians.

With the onset of the Industrial Revolution in the late 18th century onward, the rise of the factory system induced concern and trepidation for the short- and long-run effects of introducing new technology into the workplace, including among classical political economists like Smith and Ricardo. Hours of work reached their peak in capitalist society in the early nineteenth century, falling rapidly thereafter. This happened in a context of technological upheaval, as textile production was partially automated with new machines, drawing the apocalyptic vision of the Luddites who protested what they saw as impending technological unemployment by destroying machines.

In a manner typical of subsequent moral panics about automation, the Luddites’ worst fears were confounded by political-economic circumstances. The kind of mechanisation current in the early 19th century was only financially worthwhile to capitalists to the extent that it replaced only some labour, but not all. Meanwhile, the new technology made possible new forms of work to superintend it, such as mechanics to maintain machines, supervisors to oversee shop floor life, and accountants to deal with the complex new bookkeeping necessary to measure production in rising industries. Moreover, as Joel Mokyr has shown, the technology produced new and innovative commodities that in turn increased demand for new sectors of production, and the workers freed from labour in technologically superseded forms of work engaged in new activities and demanded new services that themselves created new job categories. Following the initial upswing of the Industrial Revolution, which witnessed high capital investment in productive industries, a downswing followed as capital clogged up elsewhere, resulting in a depression in 1820s which put paid to the more exuberant visions of automation and technological dynamism.

The depression created the space for nascent industrial capitalism to renew itself around the advent of machinery. By the mid-19th century the prospect of imminent automation of vast swathes of industry, with unmanned factories and attendant unemployment on a mass scale, once again entered the frame. Charles Babbage’s On the Economy of Machinery and Manufactures, John Adolphus Etzler’s The Paradise Within the Reach of All Men, Without Labour, Andrew Ure’s The Philosophy of Manufactures – the 1830s were awash with the equivalents of today’s airport bestsellers. A financial crisis and long depression followed, temporarily depriving such imaginaries of their material basis, but out of which the ground was set for new technologies to solve the social and economic problems that resulted from the turmoil.

The aftermath of the crisis set in stead the impressive gains in capitalist production identified with the Industrial Revolution and the long boom that followed it, propelled by heavy industry and electrical engineering. Out of this sprang a new production system through which work was organised and managed, Taylorist scientific management. In the wake of the recovery and the rise of this new paradigm labour productivity skyrocketed from 1870 onwards without let up, increasing 15 times in the US and 18 times in Europe in the century or so up to 1998. But, confounding attempts to see in this productive dynamism the potential for a reduction in human labour, over the same period hours of work were actually only cut by around a half. As David Spencer has shown, machines may well have had a transformative impact on productivity in industry over those years, but did little to alter the underlying tendency of long work hours. Where shorter working hours were secured, this rested less on technology than on state policy spurred on by worker struggle through unions and new political organisations. Growing worker power was tied to capital’s dependence on human labour. David Autor demonstrates that, into the 20th century, the employment-to-population ratio actually rose, and, cyclical fluctuations aside, there was and has been no longer-run increase in unemployment – although, as we shall go on to consider, underemployment is another issue.


Utopian and dystopian expectations of workerless automated factories once again reappeared in the 1930s. Work hours fell in capitalist economies and the Great Depression sparked high unemployment. The economist John Maynard Keynes saw in these conditions a future whereby machines would reduce work to a minimum – he predicted only a 15-hour week – by 2030. But such visions were reading into the present worklessness something that was not necessarily there. Economic conditions were dictating a reduction in employment due to the failure of the market to provide for the creation of new jobs amidst the wreckage of the Great Depression. This did not stop some from seeing technological development to blame – just as, Mokyr observes, commentators have seen in the aftereffects of the Great Recession of 2007-2009 a wave of unemployment owing to digitalisation and mechanisation when macroeconomic factors and aggregate demand better fit the frame.

But the damage wrought by the Great Depression in the wake of the Wall Street Crash was repaired by the mass state investment in industry in the Second World War, whereby the war machine paved the path for reconstruction once peace was won, taking advantage of new technological developments like transistors and synthetic materials and the advent of mass consumer goods underpinned by the ‘mass production’ model of Fordism.

In response, the 1950s and 1960s saw anticipation of automated factories and so on percolate once again. Spencer suggests that post-war worker power granted bargaining heft, producing a tendency in some unionised sectors towards shorter working hours. General Motors installed the first industrial robot in 1961. In the wake of such developments, fears of technological unemployment reached such a pitch that in 1964 President Johnson commissioned an inquiry into automation and the widely-accepted prospect that productivity increases were at such a pace as to rapidly deplete the demand for labour. At the same time as the commission was appointed, another set of academics and public intellectuals issued the ‘Triple Revolution’ report, which predicted mass job loss in the years to come. However, as the presidential commission concluded, such fears were misplaced. As Benanav suggests, in the 1960s, it was precisely in those enterprises where technological innovation was most rapid that employment grew the most – because, as successive predictions of imminent automated job loss have missed, productivity increases decreased prices creating more demand for the products. The shorter working hours that workers had bargained for did not stabilise in longer-run reductions because of the massive expansion of consumer desire stoked by investments in advertising and product development. Workers increasingly prioritised the capacity to consume which led to longer working hours as unions bargained for better pay rather than shorter hours.

These boom years busted with the Oil Shock of 1973. A long downturn followed, within which underlying profitability was so weak as to deter any substantial investment in new technology. Meanwhile, whereas in the post-war period a strong welfare state and trade union power granted workers high wages in support of expanded consumption, the 1970s instigated a retrenchment in the welfare state and the gradual decline of trade unionism, leading to a situation where workers turned to credit to sustain consumption and worked longer hours to service the debt. As Spencer charts, the rise of neoliberalism witnessed also the interweaving of the financialised infrastructure supporting this economy of debt and credit with an increasingly consumerist society, at the same time as workers were deprived of the means to weather the storm in their workplaces and seek out bargains around productivity, time and wages.

The recovery from the downturn saw some signs of an upswing on the back of information and communication technologies and digital networks, in the wake of which the next iteration of the automation discourse arrived in the 1980s and ran into the 1990s. Aided by new information and communication technologies, labour came to incorporate flexibility, multitasking, teamworking and individualised payment systems to a much greater degree. Worker bargaining power having declined with the rise of neoliberalism, and with it the infrastructure through which shorter working hours and higher wages had been achieved, the ‘automation discourse’ of the time was auspiciously driven by technological trends, namely advances in computerisation.

Most notably, the discourse found favour in the work of Jeremy Rifkin, whose End of Work traded in the same anecdotal evidence base as many bestsellers today. In the mid-1990s, the economy was booming around the eventually inflated promise of the internet and new information technologies. Whilst the effects of information technology may have switched some jobs between statistical categories, as Kim Moody asserts, the accounting concealed that little changed, and if anything new technologies had served to intensify and extend the working day for many workers. In this context, it was not so much the technology itself that failed in bringing about the anticipated shock to the society of work, but rather economic conditions related to the contingencies of a model of production and consumption that rested not on standardised goods as in previous iterations of capitalism but a more complex and segmented markets. Underlying contradictions in the character of contemporary capitalist accumulation manifested in the Dot.Com bubble bursting and, with it, investment in new technology, sending the techno-utopian expectations of the age up in flames. But, whilst feeding the vast over inflation of tech stocks that led to the Dot.Com boom and eventual bust, the late-nineties/early-noughties overinvestment in servers and millions of miles of fibreoptics and submarine cables did lay the material and technological foundations for the next upswing, which saw the early rise of what later became known as ‘platform capitalism’.

As Nick Srnicek describes, the financial foundations were also set in government and central bank responses to the crisis, with technological overcapacity accompanied by easy monetary policy- ‘asset-price Keynesianism’- that sought to boost the economy on the back of stock market surges and thus provided cheap capital to new ventures. These favourable financial conditions continued even in spite of the Great Recession following the 2008 financial crash, with the low interest rate environment that trailed in its wake (and persists today) having a buoyant effect on the nascent digital economy by reducing yield of financial assets and forcing investors, in the absence of viable options for investment in a sluggish productive sector, into risky and often unprofitable platform businesses that employ very few people. Whereas the top firms in 1962, like AT&T, Exxon and GM employed hundreds of thousands of workers, Google and Facebook’s workforces number in the tens of thousands and companies like Whatsapp and Instagram have been sold for billions whilst employing tens of workers. In so doing capital flowed into the production and circulation of a new kind of commodity – data, around which a spate of large monopolistic firms – platform giants like Google, Amazon, Facebook and so on – have sprung. The novel production system of this age of enterprise, the platform, represents, Srnicek explains, a “digital infrastructure that enables two or more groups to interact” by intermediating between different users – customers, advertisers, service providers, producers, suppliers. Firms then profit from these interactions by monopolising, extracting, analysing and using the data produced. This is not confined to the service sector, but increasingly shapes industrial production through the nascent ‘internet of things’, deployed through platforms by firms like GE and Siemens who build and own hardware and software to transform manufacturing and other industries into internet-connected processes into lower production costs and turn goods into services. This includes embedding servers and computer chips into the production process linked together over the internet. Components can therefore communicate independently of workers and managers.

Thus far this incarnation of capitalism has not produced substantial displacement of work or workers through the implementation of new technology. Just as some saw in the aftermath of the Great Depression a technological cause to unemployment, some too saw the same in the aftermath of the Great Recession, but after 2008, just as in the 1930s, the job loss owed more to macroeconomic pressures like aggregate demand than technological factors, and, as Peter Fleming argues, in the case of those countries where unemployment was felt most acutely like Greece and Spain, was often the outcome of macroeconomic issues like sovereign debt crises and severe trade deficits. In this context, new technology has been as likely to reshape as replace jobs. As Paul Thompson explains, from 1989 to 2017 there was an increase of 118 million jobs in the US. OECD figures suggest that in the world’s largest and most advanced economies employment-to-population ratios are at a high. To the extent that working hours have fallen since 2000 – 75 hours a year in OECD countries – the decline has been distributed unevenly among an increasingly ‘hourglass’ shaped labour market of vast disparities in work and worklessness, as Mokyr has shown. There has been particular growth, Thompson suggests, in interactive and personal service roles as well as higher-level IT and systems jobs and associated professional services, reflecting the new forms of work produced by technological change. Since the Great Recession, employment in the UK has increased with hours remaining static and real wages plummeting, as a range of new precarious forms of employment have populated the labour market statistics, often owing the new kinds of economic activity produced or facilitated by the rise of the platform. Specifically, Thompson argues, we have seen a rise in warehousing and logistics employment, such as in Amazon’s ‘fulfillment centres’. Moreover, as Ursula Huws has shown, the platform economy is sustained by the proliferation of a range of physical jobs including mining, mineral extraction and waste disposal that create and recycle the hardware, and the sales and marketing jobs that help circulate them. Hence, technology and new forms of employment – often exploitative and of low quality – appear to go hand in hand in the platform economy, with technology creating new forms of labour supply and flexibly specialising or deskilling work with a commensurate impact on the wages workers receive in return.

Specifically, logistics and warehousing have witnessed a considerable boom not only due to the massive supply chains that connect the global economy today, but because of the facilitation of the buying and selling of goods by the platform firms. As Moody recounts, predictions in the early 1990s suggested that warehouse labour would reduce by 25% in the US due to automation, but it grew by 27% in the decade up to 2000 and after that took off even further, rising 83% by 2017 in spite of recessions and new technologies – the latter actually propelling employment by making possible the intensified and more efficient tracking and tracing of goods in what have become known as ‘logistics clusters’ servicing major cities. Indeed, transportation is one of the few sectors in which investment in information processing and industrial equipment has risen since the 1990s. The expansion of this sector and the labour it employs has perhaps been the most notable impact of the rise of the platform as an organisational form.


This brings us up to the present day. What would Keynes, who suggested in 1930 that, a century on we would work only 15 hours a week in a state of automated luxury, have made of the way things panned out? Our brief detour through the history of the automation discourse in theory and practice shows us that, in spite of technological developments, work is a persistent part of everyday economic life. Technology has not replaced work, but created new forms of work and augmented others, often with negative consequences for those who perform it. Working hours around the world are often more than double the 15 per week Keynes foresaw. As Spencer argues, all the historical instances of anticipated automation above were confounded by factors Keynes seemed to miss, such as the impact of the relative strength and weakness of worker bargaining power on the pursuit of shorter working hours, or the expansion of a consumer society as a driving force behind an increase in the amount of work and working hours individuals performed to sustain their lifestyles.

But what does the future hold? Whilst there has so far been no evidence that the current ‘wave’ has increased tendencies towards automation, this has not stopped the discourse developing once again. It is here that we enter what Brynjolfsson and McAfee call the ‘second machine age’, characterised by the rapidly unfolding development of digital and algorithmic and robotic technologies, including machine learning, natural language coding and programming, artificial intelligence, robotics, sensors, connectivity, cloud computing, nano-technology, 3-D printing and the Internet of Things, which work alone and in tandem to reshape production. Specifically, advances in AI – formally defined as ‘the capability of a machine to imitate human behaviour’ – excite futurists, who see it as something akin to ‘the new electricity’. Some see in this new ‘wave’ of capitalism the potential for what did not happen in previous iterations to happen now – automation and the replacement of human labour on a mass scale. We should not, the argument goes, read off from past appearances of the automation discourse such as those discussed above any inevitability about how automation will reshape employment in the future. The coincidence of the new technologies listed previously is seen as the basis for a decisive break with the history of dreams and nightmares outdone by reality recounted above.

The main difference is perceived to be AI and its application in automating production. As Judy Wajcman notes, ‘Moore’s law’ – the idea that digital processing power doubles every 18 months – is often cited to suggest that there is an inexorable upwards bend in the technological possibilities before us. Specifically, unlike the machines typically deployed in factories, it is proposed that robotics is now capable of replicating non-routine physical, cognitive and emotional labour. ‘Smart’ machines use a combination of software, sensors and robotics to emulate complex actions performed by humans, such as self-driving cars that are capable of perceiving and responding to live traffic situations. Moreover, such technology is capable of machine learning, so that it develops and betters itself autonomously. Meanwhile, the greater affordability of robots able to perform complex work tasks – rendered cheaper by the use of sensors rather than extensive pre-programming – is making the prospect more attractive to companies.

The reason that this combination of AI and robotics is expected by some to automate away jobs to an extent previously unseen is that the ‘smart’ technologies to which companies have access today are capable of substituting not just routine unskilled labour – which is often low cost enough to not merit substitution with more expensive technology – but skilled professional work as well. Automation of the past, where it did have an effect, typically impacted routine, repetitive assembly and clerical work. These roles are still very much at risk. But, as Morgan explains, advances in AI make possible the substitution of tasks that are mobile, discriminating, multi-functional, linguistic and even those that involve complex decision-making. On paper, Spencer suggests, the previously unavailable combination of new technologies afforded by the present moment potentiates the replacement of everyone from taxi drivers, truckers and warehouse operatives to doctors, translators and journalists. It is this expansive array of automatable jobs (or tasks, as we will go on to see) that differentiates the ‘second’ machine age from the first.

So what fate do these technologies hold for work and workers today? The most apocalyptic predictions from economists Frey and Osborne portend that 47% of the employed population in the US work in occupations that could be replaced “by computers and algorithms within the next 10 to 20 years”. The occupations Frey and Osborne identify include retail workers and financial traders. Some countries, like Germany, are predicted to lose as many as 60% of jobs, a similar proportion in India, and up to three-quarters in China. In a particularly comprehensive study that goes further than Frey and Osborne in breaking down roles into capabilities, McKinsey project that 51% of jobs in the US economy, routine and semi-routine, manual and cognitive, are susceptible to replacement by robots. Occupations identified are as diverse as accountants, lawyers, butchers and waiters, with “interfacing with stakeholders” one of the capabilities most susceptible to automation. Closer to home, extreme estimates from the likes of Andy Haldane suggest that some 15 million jobs in the UK are at risk of automation.

Based on an algorithm detecting the liability of different occupations as a whole to automation, Frey and Osborne is the cornerstone evidence base for much of the contemporary automation discourse. But their study conflates tasks – some of which can be substituted for or reshaped by technology – with jobs, which independent of whether certain tasks are automated may still persist, and ‘occupations’, which is a more general category altogether. Frey and Osborne award waiters a 94% automatability rating, for instance, but this abstracts the routineness of the occupation from the context of social interaction in which it takes place. As Thompson observes without social interaction, the service encounter on which a lot of hospitality rests would be meaningless. Moreover, this work is often based on a business model resting on the extraction of maximum effort for minimum cost, which raises the question as to why employers would invest in labour-substituting technology – especially when the evidence shows that wage-raising measures like minimum wages make more likely that low-skilled work will be automated.

The leading counterpoint to Frey and Osborne is the OECD study by Arntz and his co-authors, which goes much further than McKinsey in bringing a task-based perspective to bear against the susceptibility of employment to replacement by automation. This approach shows that the heterogeneity and adaptability of tasks within individual occupations means that rather than half of occupations being vulnerable to automation, only 9% of total employment in the 21 OECD countries is at risk, measured by a prevalence of 70% of automatable tasks or more within a given occupation. Even the McKinsey study suggests that, whilst around half of jobs could on paper be automated, only 5% of occupations themselves could be automated in full, with the great amount of jobs devoted to people management, expertise, planning, creativity and personal interaction the most resistant to any automation whatsoever.

Like Arntz et al, the McKinsey study is also distinguished by giving due consideration to wider factors than the technological possibilities alone. The study dwells on the broader context within which technology is implemented, such as costs of the technology, the existing cost and availability of labour, the impact of automation on productivity and quality, and regulatory, ethical and socio-political issues around implementation. Placed in this context, due to the lack of sophistication that robots still exhibit in performing even the most menial factory work and the costs of programming them, the McKinsey study anticipates that rather than wholesale replacement of human labour the new technologies are more likely to result in the augmentation of human labour, with its flexibility and ‘soft’ skills, with robotics in pursuit of greater speed and precision. In this way, as Autor has argued, un-automatable tasks are typically complemented by automation. Indeed, even Brynjolfsson and McAfee acknowledge that irreducibly human qualities of ideation and creativity will resist automation and actually generate new job opportunities in the wake of widespread job wastage elsewhere.


Modelling of the future aside, what do we actually know about what is happening right now? Bureau of Labour Statistics figures for the US suggest very little employment churn prior to the coronavirus crisis, and, as Thompson shows, recent firm-level case studies in the US and Asia-Pacific suggest that implementation of AI was “notably absent” in most companies studied. This is supported by the World Economic Forum’s report on The Future of Jobs, which investigated what executives saw as the main drivers of change in the present time. As Moody highlights, only 9% suggested advanced robotics and fewer still AI and machine learning, whilst some 44% saw “changing work environment and flexible working arrangements” as a key issue for the future, from which we can infer that executives are focused more on raising productivity through management practices rather than labour-saving technologies. Looking out upon this slightly underwhelming scenario when measured against the claims of the automation discourse, the Roosevelt Institute even went so far as to state that “while it is challenging to know what the future holds, the data are clear. We are not in the middle of a labor displacing technological boom, nor are we on the verge of rapid technological change in the near future”. Indeed, as research by Ian Clark has shown, the prevalence of hand car washes in the UK would appear to mark a step backwards from where technology was twenty years ago – regulatory grey areas supporting the exploitation of migrants and other precarious workers at low wages that disincentivise investment in productivity-raising or labour-saving technologies.

The seeming clash between the expectations projected in the mainstream modelling and the actuality of work on the ground owes to the absence in the former – aside from notable exceptions like Arntz et al and the McKinsey study – of any real engagement with the political economy of contemporary capitalism, or, in other words, how politics, the state, social relations, class and so on impact upon the functioning of the economy and the decisions made by actors within it. Placing automation in its political economic context shows that the introduction of AI and robotics is far from inevitable and will ultimately be determined by their value and cost to capital. By contrast, the mainstream modelling, by focusing on a narrow understanding of the susceptibility of different tasks and occupations to automation, places emphasis on technological opportunity rather than the profitability criterion which represents the decisive factor in whether firms find automation practical or not – and in an age of persistently low profitability, the prospects are not promising for those who perceive utopian potentials in the rollout of labour-substituting technologies.

As Fleming argues, in this sense automation is ‘bounded’ by social, political, cultural and geographical factors that influence how it plays out. We have seen in our journey through the history of the automation discourse numerous examples of where these factors come into play. One factor that ‘binds’ automation is the pricing of labour. As Thompson argues, in low-margin business models presupposed on flexibilised and intensified forms of labour exploitation, outlay on labour-substituting technologies would be an unnecessary extravagance. The deployment of industrial robots is still extremely costly, with substantial programming needed to overcome what they lack in sophistication owing to what is called ‘Moravec’s paradox’, or, in other words, the difficulty of translating into the functioning of robots the full spectrum of impressive achievements in the spheres of computing and AI. Most companies thus confront costly but clumsy robots, making their existing workforce a safer bet to exploit.

Another factor we have seen enter the frame is organisational power relations. If workers are more militant and unionised, employers might be more keen on labour-substituting technologies to deal with the issue – a notable case being Uber who, as Fleming describes, declared the intention to invest in self-driving technology at the precise time its drivers were mobilising. This exposes the extent to which the implementation of new technology does not happen in a vacuum and is not propelled simply by the agency of individual executives or entrepreneurs. Rather, capitalist development takes place in response to the shifting stakes of the relationship of conflict between employers and employees, and the need of the former to exploit the latter to turn a profit. Workplace technology as we know it is not a result of engineering brilliance alone but also the need for capitalism to reproduce itself through expanded productivity and profit. The forms it assumes, the purposes to which it is put and the outcomes it achieves are all subordinated to the unequal power relations between labour and capital and the latter’s search for surplus value extracted from the former. Workplace technologies arise and are implemented only insofar as they contribute towards this aim, and capital will seek out other means of generating surplus value – low-cost labour, for instance – where a margin is not facilitated by technological means. Moreover, with infrastructures of collective bargaining hollowed out, workers have little basis to lay claim to some of the gains generated by technological impositions through productivity, wage and time agreements.

There are wider macroeconomic features of the turbulent contemporary political economy that confound expectations of widespread automation. As Spencer argues, tech evangelists such as Brynjolfsson and McAfee present the proliferation of digital technologies as an inevitable fact of nature without considering the financialised character of business in the present age, and the fact that firms make decisions about investment less with productivity in mind than shareholder value. It is widely recognised in government at a national and international level that capitalism has been stricken by stagnant productivity growth for some time now – exacerbated by a financialised economy wherein firms allocate funds to a range of goals and purposes other than productive investment. Whilst dedicating not insubstantial sums to investment in and acquisitions of risky start-ups, tech firms themselves, like Apple, as well as the platform giants like Google, Facebook and Amazon, have otherwise amassed vast financial reserves sequestered away in tax havens. Most leading firms now redistribute profits upwards to shareholders in the form of dividends rather than downwards to productive investment in the firm itself.

Indeed, in the aforementioned WEF report The Future of Jobs, the executives surveyed pinpointed ‘pressure from shareholders’ as a key barrier to investment in new technology. Whilst some evidence points to increasing sales of robots, with the Institute for Robotics reporting a 16% increase in sales in 2016 and a tripling of units sold since 2006, placed in wider economic context, other evidence from the US suggests a notable decline in both investment in information processing and industrial equipment as a proportion of new private investment as a whole, and the ratio of capital stock to GDP. Driven by the need for quick profits to service rising corporate debt, what little gains in productivity capital has gleaned have been achieved largely on the back of intensified or flexibilised labour regimes rather than technical investment. The ease with which firms can exploit workers combines with the short-termism of shareholder value to disincentivise the kind of investment and R&D expenditure necessary to the scenarios of mass worklessness painted in the automation discourse. Indeed, as Paul Mason suggests, in the first decade of the millennium, productivity growth declined and “pure technological innovation drove global growth by precisely minus 0.2 per cent”. The evidence suggests that, contrary to utopian and dystopian visions of the unfolding future, the degree of automation in the UK economy has actually slowed in recent decades due to plummeting real wages deterring employers from investing – a point recognised even by the ‘fully automated luxury communist’ Aaron Bastani. Profitability determines investment, and the requisite conditions of profitability on which productive investment could be justified are lacking even in the latest technologies. As Moody notes, investment in new computer capacity is stalling as the superfast chips that represent the current state-of-the-art far exceed the needs and purposes of an economy based on other means of raising surplus value. As companies seek other means to accumulate reserves and satisfy shareholders in the absence of viable routes to invest otherwise, these hardly seem favourable conditions for the kind of technologically augmented reduction of labour some foresee.

All in all, this adds up to picture of unprecedentedly low innovation and opportunities for growth, rather than the age of technological dynamism portrayed by advocates of the automation discourse. The one thing that might serve to change the course of events is a sharp uptick in investment in automation as the result of a period of stable economic growth following a vast and catastrophic depression, taking advantage of the creative destruction that trails in its wake putting less productive firms out of business. In this sense, the economic catastrophe sparked by the coronavirus may have propulsive consequences for the automation tendencies discussed – and largely dismissed – in the first few chapters of our book. Its unpredictable effects on economies and labour markets may well lay waste to the former and severely tighten the latter, in a way that might well precipitate moves towards a greater degree of automation in a range of different industries. The crisis set in train by COVID-19 raises a number of questions about the future of work: will social isolation diminish the centrality of the human to service encounters, making possible previously unviable technologically enabled efficiencies in service work? Will tighter labour markets make it more feasible for firms to automate production without regard to the underlying cost of labour? And will the crisis spark a wave of creative destruction that clears the way for the fulfilment of the current potentials for productivity increases and technological dynamism that some see buried within the present? There is already some evidence, in a new EY report surveying nearly 3,000 executives about their upcoming decision-making in the wake of the crisis, that around 40 per cent of executives are planning to up their spend on autonomous technologies. How these tendencies play out in what is likely to be a rocky period politically and economically remains to be seen. But whatever the case, economic ruin on the back of a pandemic is probably not a price worth paying for a fully automated future.

Charting the history of the future as we have done here, one can see that the best laid plans hardly ever come to pass. As the first Futures of Work editorial pointed out, we have a choice between the possible futures before us, none of which are predetermined. As another crisis breaks and shakes up the pieces, it is time to walk an unfashionable line between abstract utopia and abject dystopia. That way, we can actively intervene in the unfolding of these futures and create the world of work we want to see, rather than the one we believe is coming.


Frederick Harry Pitts is Lecturer in Management at the University of Bristol

Ana Cecilia Dinerstein is Reader in Sociology at the University of Bath

Image credit: Adi Goldstein on Unsplash