The Digital Factory is a timely and thought-provoking book that addresses the spectre that haunts many studies of the digital revolution: the impending obsolescence of human labour. The book adds significantly to the larger contemporary debate about the implications of digital technology for the future of work. Moritz Altenried contributes to a growing body of writing that calls into question techno-fetishistic views of digital capitalism, as well as statements and forecasts about the onset of a new period marked by complete automation. Written to challenge assessments that are more concerned with the technological marvels of automation than with labour relations, this corpus of literature makes the critical point that human labour is still an essential prerequisite for modern global capitalism.

The book stands out in this growing body of literature by taking up the challenge of carefully researching specific labour sites under algorithmic management. Based on seven years of ethnographic research at several workplaces to investigate how platform business models are changing the work process and working conditions, the book advances a compelling combination of theoretical breadth and empirical depth, going beyond the field's mostly theoretical works. “Wherever possible, I tried to be present at the digital factories in question”, Altenried wrote, and the effort paid off because the end result is a highly readable account that reveals humans, who code and feed the algorithm, as the true agents behind the enigmatic control of algorithmic management. The book is organized around the labour process in the four areas of logistics, gaming, crowdwork, and social media. Altenried takes the reader on a captivating tour that visits work settings ranging from Amazon warehouses to online video games, from data centres to the gig economy to social networks, and from content control businesses to social networks.

Altenried advocates for a perspective on the growth of digital technologies that is anchored in an understanding of capitalism's political economy by highlighting the importance of power dynamics in influencing the structure, trajectory, and implications of these technologies. His analysis stresses continuities in work organization and labour management from the industrial workplaces of the nineteenth and twentieth centuries to the digital workplaces of the twenty-first century, on two grounds: the nature of the factory, and Taylorism. Altenried correctly remarks that the factory's most important role is often that of a counterexample against which the transformation of labour and capitalism is analysed (p. 4). However, rather than signalling the end of the factory age, he contends that digital capitalism expands the factory model of centralized control over vulnerable labour to encompass all aspects of the global economy, well beyond the physical confines of factories. By utilizing digital technologies, factory logic is able to explore novel spatial configurations through the mobilization, repurposing, and recombination of fundamental Taylorist concepts. Therefore, Altenried contends that we have to theorize “contemporary capitalism not by the factory's end, but by its transformation, multiplication, and generalization” (p. 160).

Both of these ideas have been circulating for some time. The term “digital Taylorism” has recently been used to describe a number of organizational approaches, such as quantification methods, microwork management, standardization, and technical algorithmic management, that regulate contact centres and platform-based labour processes.Footnote ¹ Labour historians have previously contended that the factory, as a paradigm of production organization, has been implemented in the service sector (namely call centres) and logistics and distribution (warehouses).Footnote ² Companies in the service sector (call centres) and logistics and distribution (warehouses) use the factory as a production model, adopting historic Taylorist practices such as the use of technology to control the pace of work and the fragmentation and mechanization of tasks to increase productivity while deskilling workers. In such work environments, workers are subject to rigorous regimentation and routine monitoring of their performance against metric norms, leading to claims of modern-day “Satanic mills”. Altenried's work is distinguished by the fact that he successfully incorporates historical underpinnings and a wide range of case studies to recognize the continuities between the long-standing dynamics of capitalism and its allegedly innovative dynamics.

Altenried's use of the digital factory as an integrative tool substantially enhances empirical analyses of various digital labour environments. With digital factory and digital Taylorization, he qualifies a specific way of work organization and method of labour control that regulates the labour performed within digital network spaces. At the same time, however, due to the high level of abstraction, the use of these concepts to describe exceedingly diverse work settings presents a number of drawbacks. The book's strength becomes its shortcoming in the sense that the factory metaphor and the Taylorism framework do not operate equally well across all of the case studies it discusses. In what follows, I examine the book from a labour history perspective to situate Altenried's argument within longer histories of control, both at work and more widely.

Firstly, Altenried's rescue of the factory from its physical shell logically identifies it as the quintessential capitalist workplace, characterized by strict labour control based on division of labour, product standardization, and time coordination of activities, as well as a social organization of labour that accelerated work and quantified the work effort. He thus follows the mainstream historiographies of modern labour management that locate the emergence of modern labour management techniques in the factories. Recent scholarship, however, has revealed earlier attempts at labour management methods in the context of non-industrial and physically coerced labour.Footnote ³ Crucial labour-management techniques associated with modern scientific labour management, such as standardization, minute division of labour, and even the stopwatch, were developed under colonial, unfree circumstances long before they were adopted in Europe. What Altenried calls “the logics and workings of past factories” (p. 5) could be found in historical terms, for example, in plantations. The “unexpected continuities” (p. 15) Altenried refers to can be traced back to earlier instances. But there is little doubt that the constellations of these logics and workings in digital factories are novel, and it is just this element that piques the curiosity of the labour historian. The questions that need to be asked are: what are the consequences in terms of labour control, political subjectivity, and resistance, once the factory's logic and working are removed from its physical structure? When approached with such questions, our discussion of “the digital factory” expands far beyond the issue of digital labour, taking on new theoretical dimensions, including our understanding of classical Taylorism, and, by extension, our understanding of capitalism.

Secondly, Taylorism is not a static system; on the contrary, it has exhibited ongoing historical variability and adaptability, shaped by the previous models of production organization, the timing of the Taylorist principle's introduction into a national setting, its reception by labour, and the broader political landscapes. Beneath the surface of apparent similarities resulting from the adoption of specific labour management strategies, there exists a vast panorama of political and economic factors that influence both internal and external labour regulation. Historians have documented these variations, focusing on their social and political implications.Footnote ⁴ The rise of Taylorist labour control in Germany is a significant chapter in the study of these historical divergences. Unsurprisingly, a notable illustration of this variation in The Digital Factory is in the chapter on labour and the political economy of video games, where Altenried examines the quality assurance workers in the German gaming industry and the professional video game players known as “gold farmers” in China. Although “both groups of workers are part of the gaming industry's complex and transcontinental value chains” (p. 65), there are significant disparities in work organization and external labour control. The notable contrast is particularly evident in the “literally one piece of paper” (p. 80) discovered in the Berlin office of the Chinese-owned gaming corporation. In these highly digital work settings, that one piece of paper belongs to the works council, the first of its kind in Europe, as Altenried acknowledged, in an industry where strikes and unions are unheard of, yet a standard phenomenon in many “old industries” throughout Germany. What established that “standard” were the significant advances made by German workers after 1918, which gave the labour movement a strong voice in debates over methods of increasing productivity. Taylorism's historical embeddedness, as seen in this case, is an important component of what Altenried refers to as “a complex economic geography” (p. 18).

My third point concerns a central component of Taylorist work organization: managerial labour. Where does algorithmic management sit in the long history of Taylorist management? At the heart of Taylorist work organization was managers’ monopoly of knowledge and expertise. By concentrating knowledge in their hands, Taylorism aimed at completely separating work conception from work execution. The digitalization of management challenged this separation by changing the role of management from responsibility “for the performance of people” to responsibility “for the application and performance of knowledge”. While, in traditional factory settings, management extracts information from workers in order to control workflows and intensify production, this process is now at least partially automated, as software systems capture workers’ activities to datafy and analyse them. In 2015, The Economist described the changing role of managers in the age of digital capitalism as follows: “In Taylor's world, managers were the lords of creation. In the digital world, they are mere widgets in the giant corporate computer”.Footnote ⁵ The truth, however, appears to be more nuanced than that.

Over fifty years ago, Harold Wilensky, an American organizational sociologist, observed that “white-collar automation entails both upgrading and downgrading”, wherein top-level management gains authority over middle-level management, similar to how scientific management in the past enabled supervisors to exert control over workers.Footnote ⁶ Wilensky projected that the top management would experience a greater degree of separation from the rest of the organization, resulting in a more distinct division between those who make decisions regarding “what is to be done and how” and those who carry out those decisions. In time, this division would be more pronounced and elevated, and the men who previously imposed Taylorism on the working class would themselves have been Taylorized. It appears that his predictions were accurate. The persistent increase in vertical division of labour between managers in contemporary work organizations also means a growing separation between workers and management as a whole. The ability of capitalists and their functional elites to conceal managerial authority behind algorithmic management techniques frees them from the obligations previously imposed by the Fordist capital-labour nexus, building on a class compromise between organized industrial capital and labour.

But what exactly happened to managerial labour itself? How is automated decision-making hidden from both workers and managers? Altenried's primary focus is on microtasks, which are repetitious, dull, and have clear boundaries. These tasks require low investment in training and development, and they are carried out with increased supervision and workforce utilization via algorithms. However, his analysis does not go into detail on managers’ first-hand experiences with these technical improvements. As algorithms have assumed managerial tasks, I have been left wondering about the fate of managerial labour and the limits of control at work. To what extent did platforms outsource the middle-manager position? Can we talk of the process of dehumanizing managerial roles? To what extent do these different work settings show similarities in the reconfiguration of labour-management relations?

The intensification of labour control through the use of performance management tools and the increasing levels of work intensity have been captured in terms such as “an assembly line in the head”, “quantified workplaces”, and “algocratic systems of governance”.Footnote ⁷ Empirical studies have shown that these new work organizations rely on a blend of technologically driven measurements and discretionary human management.Footnote ⁸ How might different variations of this blend play out in the many case studies that Altenried investigated? For instance, how would the experiences of workers dispersed throughout space, like social media workers or gold farmers, vary from those of warehouse workers in the e-commerce industry? What are the limitations of algorithms and automation in labour control, particularly in terms of ensuring worker commitment? What about scenarios in which capital employs algorithmic management to obscure managerial decision-making processes and eradicate accountability?Footnote ⁹ And, last but not least, how might the mix of technology-driven metrics and discretionary human management impact workers’ access to knowledge about the work organization and their resistance to management in the labour process? This final query brings us to the subject of worker subjectivity. Understanding the influence of digital technology on workers’ perceptions of control is essential for grasping capital's current supremacy and possible disruption. In the lengthy and complex history of resistance to technology and the use of technology in worker resistance to employers, where do the new forms of resistance under digital labour fit in?

The connection between technology and workers’ subjective perceptions of control has been a key component of capital's control of labour. A look at some of the key evaluations of that connection may help us contextualize the contemporary debate over digital technology and automation, which are sometimes portrayed as novel and unprecedented interventions in the labour process. One crucial example of such analysis pertains to the identical type of work that Altenried begins his book with: the reproduction of texts. The Digital Factory opens with the story of “ScanOps”, subcontracted workers at Googleplex, Google's Silicon Valley headquarters, who are employed in a project to digitize every book in existence. Without the employee benefits that other Googleplex employees receive, the ScanOps’ job is pressing the scan button on a machine that instructs the worker to turn the pages in a rhythm. The ScanOps, reduced to a mere extension of the machine and subject to a tight quota system, serve as our entry point into a world of labour regimes that bear little similarity to the creative, communicative, or glamorous image of employment at Google headquarters.

The labour involved in the reproduction of texts is the example Gramsci chose to explain his views on Taylorism and the mechanization of the worker as well. Taylorism supposedly produces a gap between manual labour and the “human content” of work, Gramsci observes. When detailing the differences between printers, compositors, and scribes in the past, he does not become nostalgic over the scribe's loss of craftmanship. The compositor, unlike the scribe, has to ignore the intellectual content of the text he is reproducing, be much quicker, and “keep his hands and eyes constantly in movement, and this makes his mechanization easier”. But rather than argue that this latter kind of work is conducive to spiritual death, Gramsci sees it as potentially liberating in that automation permits another form of thought dissociated from the specific task undertaken: “Once the process of adaptation has been completed, what really happens is that the brain of the worker, far from being mummified, reaches a state of complete freedom”. The unexpected forms of labour intellectualism enabled by new industrial methods were recognized by American manufacturers, who knew that “the ‘trained gorilla’ is just a phrase, that ‘unfortunately’ the worker remains a man”, who not only thinks but also acts.Footnote ¹⁰

Almost three decades after Gramsci, in her analysis of automation and new humanism, Raya Dunayevskaya wrote about the 1947 West Virginia mine workers’ strike, dubbed “the first Automation strike”, against a continuous mining machine known as the “man-killer”. The continuous miner, frightening in an “entirely new, total way”, generated ghost towns throughout Pennsylvania and West Virginia, resulting in the longest strike in mine workers’ history since the establishment of the Congress of Industrial Organizations. Unlike Gramsci, Dunayevskaya saw automation as threatening not only the worker's body and livelihood (she quotes a woman auto worker as saying, “All Automation has meant to us is unemployment and overwork. Both at the same time [emphasis in the original]”), but also her basic humanity. For her, the conflict over automation is a “struggle for the minds of men” and an attack on work pride (she quotes a young worker who questions, “What pride can you have in your work if everything is done electronically and you are there just to blow the whistle when the machine breaks down?”).Footnote ¹¹

While automation is an actual heightening of the alienated labour in Dunayevskaya's analysis, Herbert Marcuse states in his letters to Dunayevskaya and in One-Dimensional Man that the form of drudgery is “expressive of arrested, partial automation”. While “restricted automation saves the capitalist system”, Marcuse argued, “consummated automation would inevitably explode it”.Footnote ¹² The 1950s also saw the emergence of the first wave of debates on digitalization, which characterized computerized automation as a radical departure from traditional technology. This sparked political discourses that promised a significant restructuring of labour, including the deconstruction of traditional jobs and increasing leisure time for workers. Similar to Marcuse's claim on consummated automation, these discourses predicted that digitalization would release workers from work as productivity continued to rise.Footnote ¹³ In these assessments, automation was seen as a solution to problems associated with the crises of Taylorism and Fordism.

The following decade, however, saw a decrease in productivity and an increase in automation. Writing in the context of the “cankerous problem [of] the spreading disenchantment of the American worker with his workplace” after ninety years of “markedly successful” Taylorism and shortly after the 1972 General Motors strike, “the most dramatic instance of worker resistance since the Flint sit-downs in 1937”, the editor of the journal Mechanical Engineering pointed at automation as the possible remedy for “antiwork syndrome and declining productivity”. After diagnosing the problem as an excessively authoritarian industrial structure that resulted in endless division and subdivision of duties that deprived workers of decision-making powers, the editor saw the remedy as “the replacement of autocracy by democracy”. In an unexpected twist, the author contends that the introduction of automated technology rang “the death knell for scientific management” because it promised to end the sense of powerlessness and meaninglessness that characterizes the work relationship in machine-tending and assembly-line technologies. To fully exploit the productivity potential of automation, management must extend this promise to labour, rather than simply gifting it to them. Automation, in this scenario, represented a promise to reverse the dehumanizing qualities by shifting the role of the worker from providing skills to assuming greater responsibility in the production process, hence enhancing worker control over the immediate work processes.Footnote ¹⁴ The shift here reflected a broader trend of challenging the concept of skill specifically and its relationship to labour control in general.

The 1970s also witnessed the rise of the Labour Process (LP) debate, which has mostly focused on the impact of technology on workers’ scope of action and on labour control in general. In his seminal study Labour and Monopoly Capital, Henry Braverman contended that capitalist dominance over the labour process has been gradually cemented through deskilling and degradation of work, and that this same transformation of work will eventually lead to revolt by alienated workers. As the issue of subjectivity continued to feed subsequent LP debates, the automation-skill debate, as it related to long-term trends, shifted Braverman's singular focus on craft autonomy and expertise to a more comprehensive framework for empirical investigation, including dimensions such as responsibility, abstractness, and interdependence. The resulting discussion has argued that skill and worker control over the labour process have multiple dimensions, and long-term automation trends seem to encourage distinct shifts along each of these, which may be supported or inhibited by other developments, especially developments at the broader societal level.Footnote ¹⁵

How can these historical analyses improve our understanding of the current conflict between the autonomy of labour and the logic of capital? What insights can we gain from them regarding the relationship between measurement, surveillance, and worker subjectivity? First, the diverse historical assessments I reviewed show that there is always a dual character to automation in the workplace. There is a spectrum between machinic subordination and autonomy at work. The contemporary discussion on “digital Taylorism”, including Altenried's analysis, echoes Braverman's thesis in that it refers to a style of control that uses coercion, limits labour's breadth of action and subjectivity, and separates conception and implementation. This tremendous control scenario is enabled by technical applications such as real-time tracking and tracing, digitally assisted assistance systems, and automation, which provide new opportunities for labour deskilling, fragmentation, and standardization of labour processes. The problem with such analyses is two-fold. On the one hand, they appear to overlook the fact that artificial intelligence at work seems to be failing to deal with “the indeterminacy of labour power”. The significance of algorithms used for measuring and supervising work may be overstated due to little understanding of their practical application.Footnote ¹⁶

On the other hand, they assume restrictive labour control as the sole or main interest of management, thus overlooking the possibilities of workers using the same technological tools of labour control to “break” the system.Footnote ¹⁷ Rather than fulfilling “a historical wish of scientific management”, as Altenried contends (p. 158), digital measuring of the labour process appears to be opening up new possibilities for workers to push the frontier of control in the labour process while also engaging in regulatory and distributive protests.Footnote ¹⁸ Recent studies have explored the emerging methods that workers are employing to oppose the implementation of algorithms and automation in the workplace, examining the possibility for new collective capabilities in the realms of data collection, aggregation, and curation. The crucial questions to ask about this latest wave of protests are: (1) How much continuity do they have with previous instances of labour protests, and how do these new forms of labour organization differ from the project of social control contained within the standard Taylorist model? And (2) What are the building blocks for forging collective solidarity under conditions of digital labour?

The historical assessments also warn us against overgeneralizing the effects of technology on labour control and worker autonomy. As the proliferation of research on various digital labour forms illustrates, even a predominantly digital economy will have considerable diversity. The impact of digital technology is significantly determined by the technology utilized and the type of control applied: direct and restricting or indirect and activating. Digitalization takes many different forms, depending on the industry's technology and economic situation before automation. We should strive to explain the greater diversity of outcomes through finer-grained analysis of specific industries, firms, or tasks. The Digital Factory contributes significantly to this effort by demonstrating once more what historians of labour, management, and technology have repeatedly demonstrated over time and space: technology is not a neutral, exogenous factor for change, but rather a tool employed by capital to enhance competitiveness and exert control over labour autonomy. Although digitalization, artificial intelligence, and robots are all examples of different sorts of technological inputs in the manufacturing process, their implementation is still hindered by a variety of obstacles, including those of a technical, economic, social, and political nature.