The argument I make in Smart Machines and Service Work takes a different tack altogether. I make the case that there are three primary reasons for the relatively insignificant impact new forms of automation are having on labor productivity growth rates in the world’s advanced economies. First, the types of labor processes many automation theorists suggest are vulnerable to replacement by smart machines in fact require an intuitive, embodied, and socially mediated form of knowledge or skill that even the most advanced machine-learning programs cannot master. This is especially the case with so-called “personal services,” the fastest-growing segment of today’s job market. These activities require in-person interactions between providers and consumers that pose technological, moral, and even legal limits to their replacement by machines. These limits are, I contend, reinforced by an even more powerful disincentive to automate certain lines of production: the prevalence of cheap labor in the advanced economies. When workers will take jobs that are poorly paid because nothing else is available, there is little reason for business owners to invest in expensive and soon obsolete machinery to do their work. In the pages that follow, I show that this prevalence of cheap labor, indexed by decades of stagnant wages for workers, is itself an effect of technological stagnation. Wage gains for workers depend primarily on prior gains in labor productivity, which in turn require elevated rates of business investment to develop and implement ever-newer technologies that will continually reshape labor processes. As I demonstrate in detail in Chapter Five, however, the u.s. economy since 1980 has demonstrated a steady decline in private-sector investment; since 2000, the downward trend has accelerated dramatically. This collapse in investment, occurring against the background of the rise of the tech giants, points to a still more profound disorder at the core of the advanced capitalist economies of North America, Europe, and Japan: a crisis of profitability whose roots lie in a decades-long expansion of what I call, following the classical political economists, “unproductive” labor.

[...] In 2007, only one of what are now called the “Big Tech” firms made the list (Microsoft); today, the top seven slots are occupied by such firms, the first five nominally u.s. companies, the other two, Chinese. Apple entered the list in the last quarter of 2009, as the “recovery” began, its market cap less than $200 billion. Today, holding the top spot, it is worth five times that.23 The iPhone was first marketed in late 2007, as the crisis set in. By 2014 Apple was worth half a trillion dollars on paper; a year later, three-quarters of a trillion. Throughout this period, its core products remained a set of “smart” devices (iPhone, iPad, iPod), an expansive network of retail stores, and a lucrative service-provisioning division. Its recent triumphs in equities markets, like those of its peer “Big Tech” companies Facebook and Google, have come despite very little innovation in either product development or organizational efficiencies. Throughout the crisis period, unlike firms across the economy as a whole, Apple has piled up profits; by February 2018 the company held cash reserves of $300 billion.24 Yet instead of investing these profits in product innovation or expanding existing capacity, Apple has chosen to spend an astonishing $210 billion since 2012 on stock buybacks, a full $100 billion of it in 2018, as a windfall from the $1.5 trillion dollar tax cuts passed by Congress flooded into their coffers. Many other cash-rich corporations followed suit. According to the Roosevelt Institute, corporations spent $3 out of every $5 of their net profit on stock repurchases between 2015 and 2017; some $1.1 trillion in corporate profits (with Apple leading the way) was spent on buybacks by the end of 2018.25

In the first half of 2018, as stock market indexes (especially the tech-heavy NASDAQ) blew through historical highs, a total of six companies, all usually lumped in the category of the “tech”—or, more broadly, the “Internet”—sector, generated a full 99 percent of the S&P 500 index’s gains for the year.28 The simple subtraction of the FAANG companies (Facebook, Amazon, Apple, Netflix, and Alphabet’s Google), along with Microsoft, from the index would leave a group of 494 large-cap companies that have generated net zero growth in equity prices over the first six months of 2018.

Marx’s approach allows us to understand that the rising proportion of the labor force working in circulation and supervision represents an increasing cost to the system as a whole. This introduces an added complication to Marx’s theory of the tendency for the average profit rate to fall. What if a significant portion of the wage bill includes personnel who perform activities that do not produce value, as is the case with circulation and supervisory labor? Since these workers do not produce surplus value or, a fortiori, sufficient value to provide for their own reproduction, their wages must be paid out of surplus value produced by productive workers elsewhere in the economy, thereby drawing down the total surplus value available to capitalists for new investments. Since profits must be shared between productive and unproductive enterprises, the rising ratio of unproductive to productive labor represents an additional downward pressure on the profit rate. The increasing productivity of labor, in Marx’s sense of a rising rate of exploitation, must therefore compensate not only for the reduction in the total demand for labor relative to the capital mobilized, but for the increasing costs of circulation and supervision, as more and more labor is allocated to non-productive activity.

Here, though, we must be careful not to confuse jobs or occupations with tasks. What a particular occupation “does” is never one thing. A job is only ever a specific grouping of discrete tasks, some of which are more repetitive and easily mimicked by machines, others of which are more intuitive, subjective, contingent. Automation replaces tasks, not jobs. When the principles of automation are introduced into a particular branch of production, this tends to compel a restructuring of the prevailing division of labor rather than simply suppressing this or that “occupation.” The fate of the bank teller is a well-known example. The introduction of the now-ubiquitous automated teller machine (atm) did not spell the disappearance of the human teller; it merely shifted the responsibilities of those employees dealing directly with customers away from handling deposits and withdrawals and toward (say) the marketing of credit cards, consumer loans, and other banking services. The atm, in turn, represents a highly visible example of what Jonathan Gershuny identifies as the rise of a “self-service” economy, in which tasks formerly performed by paid employees are imposed on consumers: the automated teller replaces the bank employee’s labor not with a machine, but with the free labor of its user.

The bottom tier of the job market made up in large part of service occupations is resistant to wholesale automation because labor-saving innovation is happening elsewhere in the economy: labor-process refinements in other sectors shed workers who, because they must work to live, are shunted into low-wage, low-skill occupations. This surfeit of available labor creates intense competition for otherwise undesirable jobs, holding wages down. There can be an oversupply of labor in this sector, though, because these jobs are deemed to require little to no pre-existing skill, requiring no special training or aptitude to perform the tasks associated with them. In many cases, workers who hold these positions are trained on the job, or at most subject to short introductory training sessions or programs before they begin.

These sorts of activities, however necessary they may be for the reproduction of capitalist class relations, are always the last to be rationalized, that is, made more efficient, and less onerous, by means of labor-saving innovations. I have already discussed in detail the reasons for this. In some cases, the labor process itself is hard to reproduce mechanically. If you think robots have a hard time driving cars, imagine the calamities simple tasks like folding clothes or giving baths to the elderly would entail. But just as often, the pressure to automate activities is obviated by the sheer availability of human labor-power, which cheapens the cost of labor and therefore discourages business owners from investing capital in expensive machinery that often becomes obsolete well before it fully depreciates. An abundance of labor means a dearth of machines.