Nothing Is
Inevitable But Class Struggle: A Workerist Perspective
on AI Regulation
Alexandros Minotakis
University College Dublin,
Centre for Digital Policy, Dublin, Ireland,
alexandros.minotakis@ucdconnect.ie
Antonis Faras
National and Kapodistrian University,
Athens, Greece, antfaras@phs.uoa.gr
Abstract: AI systems are often portrayed in narratives that exaggerate
their potential, either as a universal solution or as a looming disaster.
Particularly in the workplace, AI is depicted as an unstoppable force that
shapes an entirely novel labour process. Consequently, efforts to regulate its
introduction and deployment are dismissed as futile or even regressive. This
paper argues for a different approach. Firstly, it proposes engaging with AI
within the framework of Fordism-Taylorism, reflecting capital's long-standing
aspirations for labour intensification, opacity of decision making and
separation between conception and execution. Secondly, drawing on recent
workers' struggles, with a particular focus on the Efood
platform in Greece, the paper illustrates how labour resistance can influence
the deployment and regulation of AI. The case study of Efood
highlights the collective actions taken by workers against AI-powered
algorithmic management, demonstrating the potential for labour to contest and
reshape the introduction of new technologies. In the concluding sections, a
general outline of this perspective is provided as well as an analysis of how
it can potentially tactically benefit from existing regulation, such as the AI
Act.
Keywords: artificial intelligence, class struggle, AI regulation, unions,
AI Act
Almost 20 years before the current AI hype, which is mainly associated with the rise of products such as the LLM-based ChatGPT, Vincent Mosco (2004) criticised techno-optimist narratives surrounding the development of new business models based on emerging technologies. This trend intensified in the aftermath of the 2008 global economic crisis, where narratives of hope (or catastrophe) have focused on new technologies, starting from Web 2.0 and digital platforms and, past the 2020 pandemic, onto AI systems. Mosco foresaw that technological determinism related to the digital boom could also operate as an ideological current rooted in the impasses faced by capitalism and liberal democracy.
Simos et al. (2022) provided the framework of an intellectual history of AI, identifying it with four periods of hype and relating AI to highly transformative technologies such as steam power. In their efforts, the narratives surrounding AI seem to correlate with more persistent discourses related to technology. For instance, techno-determinism remains a dominant ideology. This approach has a long history, as Mosco (2004) has shown, and consists of treating technological systems as a force that is evolving independently of societal dynamics and comes to unilaterally shape and transform society (Fuchs 2008). On the other hand, AI is related to its own techno-solutionism, which posits that any issues can be addressed through the further development of AI systems, provided that they have access to sufficient amounts of data (Lindgren & Dignum 2023).
Ultimately, whether AI is conceptualised as a “power for good” or a “force for evil” is of secondary importance. In this context, Fuchs (2014a, 202) notes: “techno-optimism and techno-pessimism are the normative dimensions of technological determinism”. In that sense, the core argument remains the same.
Therefore, we argue that, in the context of AI,an overarching narrative emerges: “AI as an inevitability” which posits that the further evolution of AI systems, their access to more and more data as well as their deployment on an expanding range of spheres of public life, is an inescapable reality. This approach assumes that technological development follows a linear progression, overcoming any obstacles that are treated as bugs, able to be fixed through minor modifications or the collection of even more data (Elish & boyd, 2018). Furthermore, this narrative is also taken at face value by critical and Marxist theorists (Srnicek & Williams 2016, Bastani 2020) who stress the need for more and faster integration of AI systems and automation, as a means of overcoming capitalism and liberating workers from work. Within this technological deterministic context, social struggle is focused on the issue “which social class(es) will reap the benefits of AI”. Regarding AI in the workplace this approach asserts that “unions will not succeed in keeping AI out of their industries, but they can certainly succeed in forcing companies to spread the benefits of AI to all” (Nolan 2024).
Overall, a dominant narrative about AI in the workplace rests on two key hypotheses. The first suggests that AI’s integration is inevitable – the next step in the linear, unyielding progression of technology, as posited by techno-determinist perspectives. The second contends that AI represents a transformative leap, ushering into a futuristic transformation in the nature of work, comparable in scope only to the Industrial Revolution (Brynjolfsson & McAfee 2014).
In contrast to this narrative and its influence on critical scholars, this paper argues two key points. First, that the future of work is already unfolding, exhibiting both significant continuities with the past and notable transformations; a dialectical approach is thus essential to illuminate these continuities and shifts within the labour process. Second, the paper offers an alternative perspective on AI, grounded in the idea that class struggle encapsulates and reflects the varying hopes and anxieties of contrasting class interests surrounding technology. Consequently, it needs to be examined whether class struggle remains the crucial factor that shapes whether, and in what ways, new technologies are introduced into the workplace or other areas of society.
To address this question, this paper will examine in sections 2 and 3 the qualitative transformations in the workplace driven by AI. This analysis requires a broad interpretation of the Fordist-Taylorist transformation of the labour process, alongside an exploration of how new and traditional elements are combined through algorithmic management. Platforms will be at the forefront of this inquiry as they function as laboratories for capital to test and introduce AI (Woodcock 2022).
Then, in section 4 this paper will examine workers’ resistance to AI, focusing on a case study of platform resistance – the strike at Efood, a food delivery platform in Greece. The collective organising and resistance of platform workers, protesting the deterioration of working conditions under algorithmic management, illuminates the relationship between workers’ struggles and the introduction of AI in the workplace. In this regard, the paper aligns with the perspective of Inhuman Power, which “sees issues of AI, jobs and joblessness as matters of class power inseparable from the chronic conflict of capital and its workers” (Dyer-Witheford, Kjøsen & Steinhoff 2019, 69).
In conclusion, in sections 5 and 6, these points are drawn together to provide an initial outline of what a “workerist perspective” on AI might involve. Building on this, the paper proposes a preliminary vision for a “workerist” approach to AI regulation, informed by lessons from labour struggles as well as for potential opportunities for workers and their representatives with the existing regulatory framework (such as the EU AI Act). This approach envisions labour as a regulatory force that can mitigate AI's impacts, guiding its development and deployment in ways that open new possibilities and align with workers' interests.
To revisit perspectives related to contemporary discussions regarding new technologies and labour, we need to situate them within the framework of their emergence. It has been argued that the global financial crisis of 1973 resulted in what has been characterised as a neoliberal attack and a crisis of the Fordism paradigm. Mainly, the standardisation of working relations and the deregulation of labour legislation has been attributed to the reconstruction of capitalist production based on micro-electronic technologies (Hirch & Roth 1986). Computational technology and workspace automautomation considered the driving factors for the transformation of labour. As such new working conditions briefly emerged that were later identified as flexicurity, flexiprecarity, hidden labour, task-based work, etc.
Technology-fueled transformation of work seemed largely inevitable. Shortly, historians of technology began comparing this new fixation on micro-electronic technologies with the techno-deterministic endeavors associated with previous landmark technological assembles such as the steam or the railway (Edgerton 2007). Proponents of a revived and digital technological determinism began arguing for the end of the work (Rifkin 1995) or even the end of capitalism (Hardt & Negri 2020). Drawing mainly on Marx’s Fragments on the Machines, proponents of digital technological determinism argued that the growing dynamic of the displacement of living labour (which Marx describes as labour-power in action) from the production process, in parallel with the inability to receive the necessary surplus value from machinery, would lead to the overcoming or alteration of the capitalist system.
Moreover, in examining algorithmically managed platform labour, scholars have questioned the analytical relevance of the Fordism-Taylorism paradigm (Wood et al., 2019). Fagioli (2021, 135) argues that within platforms where various spheres – economic, political, and social – intersect to shape the labour process, “it is not useful to apply conceptual tools developed to analyse modern industrial capitalist to a capitalist horizon that has mutated”. In the context of microwork platforms such as Amazon Mechanical Turk, Wood and Lehdonvirta (2021) highlight that workers enjoy a significant degree of autonomy, both in job selection (choosing tasks and clients) and in execution (arranging their own schedules and methods of task completion). In this sense, they characterise such work as resembling self-employment (for a different perspective on microwork and crowdwork, see Jones 2021).
Jarrett (2022) provides a comprehensive critique of the Fordism-Taylorism paradigm in digital labour, including platform labour. She challenges the relevance of the “alienation/commodification thesis,” arguing that in digital labour, life becomes intertwined with labour. This is evident in social media users who create content and data for corporate platforms while building networks, socialising, or even earning income (e.g., influencers). Thus, what emerges is a “both/and scenario where commodification and its productive logics coexist with authentic selfhood and its reproductive dynamics” (Jarrett 2022, 158). Discussing platform labour, Jarrett acknowledges the relevance of commodified labour power but introduces the concept of “assetized workers”. She argues that “a worker who is marketing their own assets is not selling their labor power to a platform to do with as it wills but opening those assets to a valuation by the marketplace” (Jarrett 2022, 161). In this framework, part of the labour power remains inalienable and the work itself is meaningful to the worker. While they sell their labour power, they simultaneously view themselves “in terms of human capital” (Jarrett, 2022, 165), speculating on the value of their “embodied subjective assets in a workplace” (Jarrett 2022, 197).
Gandini (2019) observes that within labour process theory, such approaches view labour process research as relevant only for traditional, waged workplaces, where deskilled employees work under stable contracts in fixed locations and schedules, performing repetitive tasks that result in tangible products. A thorough critical examination of these approaches is beyond the scope of this paper. However a wider perspective on what constitutes the essential characteristics of Fordism-Taylorism may be helpful in identifying continuities between the historical Fordist-Taylorist factory and the modern, AI-driven workplace.
We argue that at the heart of Fordism-Taylorism lies the transition from the formal to real subsumption of labour to capital, as well as the shift from the extraction of absolute to relative surplus value. The notion of formal subsumption aligns with the early stages of the capitalist mode of production, where capital (often in the form of mercantile entrepreneurs) employs artisans and craftsmen for commodity production without dictating how production is organised. Essentially, the pre-existing knowledge and skills of workers shape the labour process (Vercellone 2007). To increase productivity and establish greater control within factories, capital moves toward the real subsumption of labour, assuming a managerial role to restructure production into a capitalist labour process. For this purpose, capital transforms labour from a subjective element into an objective element of production, fragments skills into simplified tasks, and transfers knowledge from workers to managers (Braverman 1998).
This process aligns with the shift from a focus on absolute surplus value to relative surplus value. Absolute surplus value is extracted through the extension of the working day, while relative surplus value is generated by increasing labour productivity, either through the rationalisation (from capital’s perspective) of the labour process, the introduction of new machinery, or a combination of both; In this process, new technology is deployed in the workplace both to maximise profits by increasing the extraction of relative surplus value and to reinforce capital’s control over the labour process, thereby extending its dominance over workers. The shift toward the production of relative surplus value is associated with a deeper subordination of labour to capital and an increased reliance on machinery and “scientific management” of production.
To quote Marx:
“The production of absolute surplus-value turns exclusively on the length of the working day, whereas the production of relative surplus-value completely revolutionizes the technical processes of labour and the groupings into which society is divided […]. It therefore requires a specifically capitalist mode of production, a mode of production which, along with its methods, means and conditions, arises and develops spontaneously on the basis of the formal subsumption of labour under capital. This formal subsumption is then replaced by a real subsumption” (Marx 1976,645)[1].
Τhis transition is doubly necessary. On the one hand, the extraction of absolute surplus value is confronted with the insurmountable limit of the working day’s length. This barrier is not strictly a natural one but it is pertinent to workers’ struggles that limited the working day. Therefore, the turn to relative surplus value relates to an increase in productivity and labour intensity.
On the other hand, the shift to relative surplus labour relates to a “political project”. As Braverman (1998, 39) notes, labour power, as a commodity, carries a lot of potential as well as a degree of indeterminacy; “what he [the capitalist] buys is infinite in potential, but in its realization it is limited by the subjective state of the workers”. This “indeterminacy” greatly aggravated Taylor who realised that “although he was foreman of the shop, the combined knowledge and skill of the workers who were under him was certainly ten times as great as his own” (Braverman 1998, 70). In that sense, “scientific management” can be conceptualised as a restructuring of work that begins by severing the ties between the working class and control of the labour process. Braverman (1998,90) notes that the ties between “working population and science are more or less completely broken”. In that regard, Pasquinelli (2023, 94, author’s emphasis) speaks of the “originary accumulation of technical intelligence as the dispossession of knowledge from labour”.
Once this has been achieved, the labour process is fragmented, broken down into repetitive tasks, and the work rate is intensified and placed under supervision while labour power is deskilled. Overall, labour is displaced as its potential as a subjective factor within the labour “is removed to a place among its inanimate objective factors” (Braverman 1998,118).
In a similar vein, Sohn-Rethel (2021,128), discusses “Taylorised labour” as labour that “is not only subsumed economically to capital […] but also physically and technologically”. Sohn-Rethel points to Taylor's focus on time study as the process that illuminates “how long the studied job should take” (Sohn-Rethel 2021,127, author’s emphasis). In that sense, scientific management is build on the pretence of accuracy while it produces a “fictitious norm of labour timing” that is applied upon labourers; an abstract-”synthetic timing” which “is construed without consulting or watching the workers, even for new jobs which have never yet been practised” (Sohn-Rethel 2021,128). Therefore, the rate of the labour process becomes an alien force that exerts its power over the workers.
In the following paragraphs, we argue that AI-driven workplaces (re)produce hierarchies that devalue labour, separate conception from execution, and intensify the alienation of workers from the labour process. In this sense, we recognize continuities with the past, allowing us to speak of "digital Fordism-Taylorism" or "AI Fordism-Taylorism." In the Fordist assembly line, workers could roughly discern how the various parts were integrated into the final product, as well as the number of workers employed and the ways that they were organised within the labour process. On the other hand, in AI-driven platforms, workers have a view on the “task at hand”, with no clear indication provided on the scope of operations. Moreover, algorithmic management entails the potential of unilateral decisions on pay rate, zones of operation, surge hours and many more aspects of the labour process within platforms (Cant 2020, Wood et al. 2019). Algorithmic opacity is a crucial part as platforms do not disclose the way payment is shaped (Gandini 2019), and this is expressed as a fundamental “information asymmetry” by workers that contest mystifying narratives (Rosenblat & Stark 2016, Tassinari & Maccarrone 2020, Wood & Lehdonvirta 2021). The features mentioned above represent further alienation of workers from both the labour process and the final product, rather than a transcendence of alienation.
Even when forced by labour struggles to make certain concessions, platform companies sternly refuse to disclose any parts of the proprietary algorithm (de Stefano 2020), creating a new mystified version of “they-the algorithms” that design and impose demands upon labourers. Discussing algorithms, Cant (2020,62) discusses the “antagonistic relationship with the black box” that all riders shared but goes on to note that also coders were “artificially divided from one another”; therefore, while in a privileged position comparatively, they also had no oversight over the design of the production process. In that sense, AI proceeds in fragmenting manual as well as intellectual labour, fostering proletarianization of intellectual labour already predicted by Braverman (1998, 234).
In expanding upon this tendency, we need to understand that what is automated and replaced through the introduction of AI is not labour itself but its management that reproduces hierarchies, mystification and alienates labour from control over the production process. Pasquinelli treats AI as a tool “for imposing standards of mechanical intelligence that propagate, more or less invisibly, social hierarchies of knowledge and skill… AI does not simply replace workers but displaces and restructures them into a new social order” (2023,246, author’s emphasis). Regarding algorithmic management in platforms, Pasquinelli (2023,250) notes that algorithms “replace management and multiply precarious jobs […] against Alan Turing’s prediction, it was the master, not the worker, that the robot came to replace first”.
Overall, capital-labour relationship is embedded in specific choices regarding technology. As Dyer-Witheford, Kjosen and Steinhoff note (2019, 149): “The real subsumption of labour by capital means that capital develops and adopts technologies that fit its systemic requirements of valorization; this imperative can be baked into the very design of technology”.
Thus, we can claim that AI-driven workplaces, in accordance with the Fordist-Taylorist paradigm, engage in the intensification and fragmentation of the labour process. Intensification is expressed in two main management practices: surge pricing and emotional labour. Surge pricing refers to the process when pay rate is modified upwards either for a certain amount of time or a specific number of deliveries (often in adverse weather conditions – that is described by Cant (2020,54) as a “continual pressure to increase the risks you take”). Through surge pricing, algorithms take advantage of precarity and wage instability that are embedded into platform’s production processes (see section 4) to adapt the workforce to fluctuations in demand, by rewarding increased productivity (Woodcock 2021). Simultaneously, through AI-facilitated surveillance, blue and white-collar workers are subjected to intensive supervision, effectively limiting their autonomy (Aloisi & Gramano 2019, de Stefano 2019). Crawford (2021, 219) concludes by noting that “the future of work looks more like the Taylorist factories of the past, but with wristbands that vibrate when workers make errors and penalties given for taking too many bathroom breaks” (Crawford 2021, 219).
Thus far, we have argued that AI in the workplace exhibits several continuities with Fordism-Taylorism: the dispossession of workers' control and knowledge over the labour process, the intensification and fragmentation of tasks, the separation between conception and execution, and the overall degradation of labour. However, it is essential to examine the dynamics of AI-driven workplaces more closely to identify discontinuities and differentiations. This section will focus on precarity, surveillance and modes of incentivisation of workers as new elements that emerge due to algorithmic management.
Precarity is embedded in AI-ruled production processes and it is not simply an issue of misclassification. It is also manifested in the ways that surges are utilized by platform algorithms and unilateral changes in pay rate are normalised (Cant 2020), shaping a context where “actual incomes become impossible to predict” (Jarrett 2022,53). While stability characterised Fordist workers, a sense of perpetual impermanence appears to play a comparable role for neoliberal subjectivities in the workplace. Precarity is “weaponized” into narratives by platforms that frame it as mutually beneficial “flexibility” but research shows that workers do not necessarily consent to this view (Rosenblat & Stark 2016).
However, precarity should not be misconstrued as something forcefully imposed upon workers. Briziarelli (2019,826) observes that algorithmic management and platform labour depend on, and consequently contribute to, the post-Fordist subject, which has internalised characteristics such as “fluidity, variable intensity, and the extensive use of informational communication technology” – precarity is experienced inextricably from a certain degree of autonomy. There is a dialectic of coercion and consent, as precarity (alongside the breakdown of barriers between work-play, labour time-free time) is an integral part of ideological and cultural shifts within contemporary capitalism during the past 30 years (Pleios 2017). Without the potential for further elaboration at this point, we argue that the AI-driven, flexible and precarious labour process within platforms would be inconceivable without the prior “contribution” of social media and free-unpaid/gamified labour in the early days of Web 2.0. (Fuchs 2008, 2014a). On this regard, Cant offers an account of platform labour stating that it was a welcome alternative to conventional employment due to the absence of the manager and a certain degree of autonomy within a gamified work environment; “the black box mighty be an authoritarian system which ordered around workers like a miniature dictator, but the user interface was shiny” (Cant 2020,85). This provides a valuable insight on how surveillance, intensification, hierarchical management and autonomy/playbor (Woodcock 2019), freelancer identities and potential for organisation (Wood & Lehdonvirta 2021) co-exist in AI-ruled workplaces.
On this point, we need to elaborate on the role of surveillance within algorithmic management. Crawford (2021) understands AI surveillance as a continuity of pre-existing trends as supervisors are now equipped with tracking algorithms to detect employees’ (in)action, movement, break times, etc. On the other hand, Cant (2020) notes that algorithmic management emerges as the “partial automation of supervision and labour process coordination”; therefore, the overall number of supervisors is diminished leaving only a “driver-lead, an almost manager, who is relegated to acting as a problem solver” (Cant 2020,44-45). Discussing surveillance, de Stefano (2020,435-436) highlights a “qualitative leap in the domination and subordination of workers”, reflecting an unprecedented empowerment of employers. This arises from the immense amount of data collected through surveillance, coupled with a shift allowing machines to make decisions independently or, “at the very least, to suggest those decision”" (see also, Aloisi & Gramano 2019).
However, algorithmic management is a double-edged sword that AI-capital is holding without necessarily fully understanding its repercussions. On the one hand, algorithmic management is not only cutting costs but is actively shaping a new form of subjectivity. Tassinari & Maccarrone (2020, 38) posit that individualism of labour, alongside the “invisibilisation” of management, potentially “constrains the emergence of solidarity among workers”. On the other hand, in the absence of managers, a form of soft control is exercised that lacks physical supervision (Woodcock 2021,38). Once employees feel the need to express their discontent, they can benefit from it to engage in direct, unsupervised communication with their coworkers (Maffie 2020). For instance, Amazon workers have showcased a paradigm of AI-workplace resistance, mobilising against invasive surveillance and algorithmic management practices, leading to public demonstrations and the formation of solidarity networks despite the company's attempts to curb unionisation efforts. Their actions highlight the potential for workers to counteract the depersonalised oversight of algorithmic management and assert their collective agency (Walker 2024). Tassinari and Maccarrone (2020) also conclude that workers can “exploit” the contradictions of the labour process, build their own networks of solidarity and engage in a variety of actions ranging from individual abstention (log-out) to collective-public demonstrations and strikes. Moreover, in the absence of guidance, platform workers create communities to address everyday challenges; while this is part of the cost-cutting strategy of algorithmic management, these online spaces also create opportunities for organisation and foster the interest for trade unions and workers’ associations (Wood et al. 2019, Wood & Lehdonvirta 2021).
Overall, AI-ruled workplaces have been examined as part of a dialectic of continuity-discontinuity with the organisational practices of the Fordist-Taylorist factory, focusing on continuity as the dominant aspect in this relationship. It needs to be stressed that the subordination of labour to capital remains partial and contradictory; the labour process is a contested field that provides opportunities for a political recomposition of the working class, antagonistic to capital’s effort (Briziarelli 2019, Woodcock 2021). The next section will examine the potential of workers’ struggle to act as a regulatory force that shapes the development and deployment of AI in the workplace.
The potential of workers’ activity to impact the integration of new technologies is often overlooked, even among critical scholars. Scholz (2017, 106) highlights the importance of maintaining the “association with the history of organized labor” and recounts digital workers’ struggles. In that sense, we concur with Dyer-Witheford’s, Kjøsen’s and Steinhoff’s (2019, 101) assertion that the emergent struggles “are not outright anti-AI struggles but each rejects or contests specific aspects of AI”.
We propose to ground our understanding of AI and labour within these struggles that have refusal as their starting point and proceed to develop the embryonic stages of a radical alternative. Does this entail a retreat to small scale-”folk politics” criticised by Srnicek & Williams (2016)? No, provided that workers’ resistance is conceptualised as a force that potentially transforms and regulates the development and deployment of AI systems.
This approach is partly rooted in the workerist tradition that treats labour as an active agent that shapes the production process through its personal and collective acts of resistance so that “platforms and the different technologies involved in their use can therefore be understood as a response to working-class activity rather than the usual narrative about innovative start-ups as the agent of change” (Woodcock 2021, 20). It is also influenced by the theory of social construction of technology which posits that technologies acquire significance through social dynamics and the development of a specific technology can be shaped by the expectations of different social groups (Pinch & Bijker 1984).
Between late 2023 and early 2024, a relevant case study unfolded in Las Vegas, where robotics and AI startups endeavoured to enter the hospitality sector. Recently, Consumer Electronics Shows (CES) displayed a series of innovative robotic baristas, chefs, and waitresses marketed to businesses seeking to infuse a futuristic aura into their operations (Yamat 2024). While existing narratives mainly focus on upskilling, trade unions have developed a more nuanced understanding. The Culinary Workers Union, after a months-long negotiation under threat of strike (Kullgren 2023) introduced a series of protections in the new collective agreement with employers’ associations in Las Vegas in 2023, including provisions such as “6 months’ notice before implementing new technology […] negotiations on the implementation of new technologies such as AI […] access and mandatory free job training if there are new jobs created…and a bonus package based on years of service [...] if a union worker is laid off due to technology” (Jiménez 2024). The union’s representative defies the notion of AI as an inevitability, stating that “this idea that technology, robotics and artificial intelligence is just running wild with no control at all can do incredible damage […] So what we have to do is get ahead of the curve” (Yamat 2024).
In a similar vein, the 118-day-long strike organised by The Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) places AI at the forefront, stressing the actors’ need for protection against AI. In September a deal was reached that included terms curtailing the use of AI: “Studios will not be able to use generative AI to write or rewrite literary material, and AI-generated content cannot be used as source material. Companies must also inform writers if any materials they give the writer were created with the help of AI” (Jones & Fu 2023). Following this agreement, SAG-AFTRA members authorised a strike as part of the union’s negotiations with video game companies, arriving at a deal with voiceover company Replica Studios, agreeing that game developers that employ Replica’s AI platform, will only access licensed voices who have explicitly allowed their talent to be used (Gain 2024).
At Efood, a leading food delivery platform in Greece, the introduction of AI-driven changes led to substantial deterioration in the working conditions of its riders. Analysing these changes reveals the intricate dynamics between technology, labour, and capital, and underscores the critical need for regulatory frameworks and collective action to mitigate negative impacts. By 2021, Efood adopted and employed sophisticated algorithms to manage rider assignments, optimise delivery routes, and monitor performance. These algorithms even determine which rider got assigned to which delivery based on various factors such as proximity, past performance, and real-time traffic data.
Such tactics of algorithmic management placed immense pressure on workers. Riders were expected to meet stringent delivery times and performance metrics, often set without considering real-world challenges such as traffic jams, weather conditions, or road hazards. Thus a high-stress environment where creating the burden of maintaining service standards falls disproportionately on the riders, exacerbating job strain and fatigue. The lack of transparency in how these algorithms operate further alienated workers, who find themselves at the mercy of inscrutable, data-driven decisions (Tsardanidis 2024).
Efood initially offered riders three-month or six-month contracts with a combination of hourly payment and bonus per delivery. However, in 2021, shifted from stable monthly payments with newly hired riders subjected to a dynamic pay rate system, where earnings were directly tied to the number of deliveries completed and customer ratings (PAME Hellas 2021). The algorithms also introduced "surge pricing" during peak hours, offering higher pay to incentivize work during these periods.
Such a system made income highly unpredictable, as riders' earnings fluctuates based on factors beyond their control, such as customer demand and algorithmic assessments. While surge pricing could potentially increase earnings during peak times, it also encouraged longer and irregular working hours, contributing to physical and mental exhaustion. Moreover, customer ratings, which were subjective and can be influenced by factors unrelated to the rider's performance, added another layer of uncertainty and stress. The dynamic pay rate system exemplifies the precarious nature of gig work in the AI-driven economy, shifting financial risk from the employer to the employee, destabilising workers' livelihoods (SVEOD 2021).
In September 2021, 150 riders received messages stating that they would be terminated effective immediately – this was later on overturned by the ensuing strike (PAME Hellas 2021). The riders threatened with termination had a relatively larger job experience, having logged in the early days of the platform and had signed a contract that guaranteed a stable portion of a monthly payment – instead of the later modification where new riders were remunerated solely on their deliveries. Within this strike, an alliance between riders and customers unfolded, combining offline and online forms of action that led to the reversal of dismissals, concessions by the platform regarding algorithmic ranking and to the creation of a union of Efood workers (Tsardanidis 2024). Through union meetings and struggles, workers have realised that the desire deemed unacceptable by the platform were the relatively stable contracts and, at this point, they are taking collective action to the next level, demanding access to the parameters that define algorithmic management within the platform.
Τhese examples indicate the tendency for organised labour to resist, refute and regulate the introduction of AI in the workplace, by integrating technology issues in collective bargaining; a need that has been discussed by labour law scholars in the past years (Aloisi & Gramano 2019, Cazes 2023, de Stefano & Taes 2023) but the potential for mass workers’ activity to impose such agreements is rarely acknowledged (de Stefano 2019,2020). This is a missing link in current discussions on AI regulations. Law scholars have indicated a range of recommendations to address issues of workers’ privacy, data collection, data privacy, human accountability and overall restraint in the algorithmically-empowered managerial prerogative (de Stefano 2019). Such issues are yet to become a part of labour activism on AI. On the other hand, labour as an actor of regulation (rather than an object that AI is imposed upon) is rarely mentioned in scholarly debates on AI regulation – part of the problem is that research reports are deployed by organisations like OECD that treat employers and employees as “social partners” (Cazes 2023). However, as the importance of collective agreements is reiterated both by union militants and scholars, emerges the possibility of connections that can transform the ways scholars and policymakers approach labour in relation to AI.
Furthermore, the above-mentioned cases signify the potential for cross-sectoral solidarity between different sections and strata of the working class. In deindustrialized Western countries, baristas and service workers are often considered part of the most vulnerable and exploited parts of the working class, while artists may be considered a privileged section; however, AI is threatening to displace both of them. The missing link in these cross-sectoral alliances, with software developers being the ones that form a new type of labour aristocracy in digital capitalism and a crucial part of such alliances (Fuchs 2014b). Cant (2020, 64) grasps this contradiction: as algorithms separate office workers and street workers, the second ones often “couldn’t see any difference at all between software engineers and the actual bosses”. However proletarianisation of intellectual labour progresses through deskilling, standardisation and intensive division of labour (Woodcock 2019,78) and is now intensified by AI, then the “possibility of solidarity” (Cant 2020, 102) exists and is at stake in present and future struggles. These different strata of the working class have a common interest in collective agreements that regulate AI deployment and provide protection against further degradation of work.
Α different perspective to regulation is emergent at this point, a “workerist” one, based on the collective action of workers and unions. This perspective builds on the historical role of trade unions in influencing the introduction of new technologies in the workplace, a role that was progressively undermined by neoliberal policies and anti-union attacks from governments and employers (Marjoribanks 2000).A workerist perspective to AI regulation is still weak and not focused around a single document or a “manifesto” of sorts. However, in examining the aforementioned struggles, certain common points arise that can be summed up as a first effort to provide an outline of what would be key demands of the workerist perspective on AI regulation.
● Full transparency regarding the introduction of
AI in the workplace, the parameters that underpin its operations and the usage
of workers’ data for AI training
● Introduction of workers’ representatives into
decision making processes, with veto powers
● Human oversight in AI involved into
hiring-firing decisions as well as decisions regarding payment system and work
rate
● Addition of clauses into collective agreements
governing AI in the workplace and ensuring the enforcement of the role of
unions
On these issues, aspects of mainstream approaches to AI regulation can be tactically utilised, in spite of their overall focus on the needs of AI capital. Acknowledging such opportunities into existing regulation can be fruitful for current and future struggles and for a systematisation of the workerist perspective. For example, in EU’s finalised AI Act (European Parliament 2024) the following articles can be found:
a) Article 2, par. 11: “This Regulation does not preclude the Union or Member States from maintaining or introducing laws, regulations or administrative provisions which are more favourable to workers in terms of protecting their rights in respect of the use of AI systems by employers, or from encouraging or allowing the application of collective agreements which are more favourable to workers”
b) Article 26, par. 7: “Before putting into service or using a high-risk AI system at the workplace, deployers who are employers shall inform workers’ representatives and the affected workers that they will be subject to the use of the high-risk AI system.”
c) Article 86, par. 1: “Any affected person subject to a decision which is taken by the deployer on the basis of the output from a high-risk AI system listed in Annex III,… and which produces legal effects or similarly significantly affects that person in a way that they consider to have an adverse impact on their health, safety or fundamental rights shall have the right to obtain from the deployer clear and meaningful explanations of the role of the AI system in the decision-making procedure and the main elements of the decision taken”
d) Annex III, par. 4 specifies
that under the categorisation of high-risk befall “AI systems intended to be
used to make decisions affecting terms of work-related relationships, the
promotion or termination of work-related contractual relationships, to allocate
tasks based on individual behaviour or personal traits or characteristics or to
monitor and evaluate the performance and behaviour of persons in such
relationships”.
These clauses, in and of themselves, constitute little guarantee for workers’ rights; afterall, high-risk AI systems are mostly regulated through self-assessment by providers and deployers (articles 27 & 43). However, these clauses (articles 2, 26, 86 & Annex III) specify a field of opportunity for EU workers’ struggles to strive and implement regulations that significantly reduce the autonomy of capital.
Evidently, these are defensive measures that mostly seek to implement accountability, strengthen the role of unions and curb AI’s consequences in the workplace, as outlined in the previous sections. This defensive stance reflects the present balance of forces where “futurism” is mostly shaped by capital and its expectations. However, critical discourse has no shortage of visions of the future (Srnicek & Williams 2016, Bastani 2020) that try to conceptualise AI and automations as agents of radical change, without being able to interact with actually existing struggles and concerns. Starting instead from the early seeds of an alternative approach grounded in collective resistance that shifts the balance of force, perhaps new visions of the future can emerge.
This paper has traversed through the historical and contemporary landscapes of AI's integration into the workplace, drawing parallels and distinctions with the era of Fordism-Taylorism. The critique presented underscores the diminishing agency and increasing marginalisation of workers in AI-driven environments, where their contributions are often reduced to mere data points or considered deviations from the norm.
Therefore, identifying continuities between AI-driven workplaces and Fordism allows us to re-examine the role of workers’ actions that forge links between “traditional” work issues (payment, health plan, work rate) and new technologies.
Moreover, as discussions around the adverse effects of AI gain traction, there's a pressing call to transition from theoretical debates on AI ethics to the implementation of concrete regulatory frameworks. However, the prevailing discourse often positions regulation as the exclusive purview of state or supranational entities, such as the EU, thereby sidelining grassroots movements and the collective agency of workers. This paper argues for a reorientation, shifting attention to actually existing workers’ struggle who either refute the introduction of AI in the workplace or aspire to set certain limits and rules on its deployment. Furthermore, union action aims to formalise the achievements of these struggles into collective agreements that will govern AI in the workplace and will elevate unions into regulatory agents with decision making powers. While a workerist perspective is still in its infancy (reflecting the early stages of class struggle on Ai), it carries potential in challenging the top-down, technocratic regulation models, while benefiting from some of their aspects, as outlined above.
Overall, regulation is neither technical nor neutral. It can empower workers and citizens provided that it learns from their struggles and sheds any notion of inevitability regarding AI.
Aloisi, Antonio, and Elena Gramano. 2019. Artificial Intelligence Is Watching You at
Work: Digital Surveillance, Employee Monitoring, and Regulatory Issues in the
EU Context. Ιn Comparative
Labor Law & Policy Journal 41(1): 95-121.
Bastani, Aaron. 2020. Fully Automated Luxury Communism: A
Manifesto. London: Verso.
Braverman, Harry. 1998. Labor and Monopoly Capital: The Degradation of
Work in the Twentieth Century. New York: Monthly Review Press.
Böhm, Steffen, and Chris Land.
2012. The New ‘Hidden Abode’: Reflections on Value and Labour in the New
Economy. In Sociological Review 60(2):
217-240. https://doi.org/10.1111/j.1467-954X.2012.02071.x
Briziarelli, Marco. 2019. Spatial Politics
in the Digital Realm: The Logistics/Precarity Dialectics and Deliveroo’s
Tertiary Space Struggles. In Cultural
Studies 33 (5): 823-840. http://dx.doi.org/10.1080/09502386.2018.1519583
Brynjolfsson, E., & McAfee,
A. (2014). The Second Machine Age: Work,
Progress, and Prosperity in a Time of Brilliant Technologies. New York:
W.W. Norton & Company.
Cant, Callum. 2020. Riding for Deliveroo: Resistance in the
New Economy. Cambridge: Polity Press.
Cazes, Sandrine. 2023. Social
Dialogue and Collective Bargaining in the Age of Artificial Intelligence. In OECD Employment Outlook 2023: Artificial
Intelligence and the Labour Market. Paris: OECD Publishing. https://doi.org/10.1787/c35af387-en .
Crawford, Kate. 2021. Atlas of AI. New Haven: Yale University
Press.
De Stefano, Valerio. 2019.
‘Negotiating the Algorithm’: Automation, Artificial Intelligence and Labour
Protection. In Comparative Labor Law & Policy Journal 41 (1): 15-46. DOI:
10.2139/ssrn.3178233
De Stefano, Valerio. 2020.
Master and Servers: Collective Labour Rights and Private Government in the
Contemporary World of Work. In International
Journal of Comparative Labour Law and Industrial Relations 36 (4): 425-444.
https://doi.org/10.54648/ijcl2020022
De Stefano, Valerio
and Simon Taes. 2023. Algorithmic management and
collective bargaining. In Transfer:
European Review of Labour and Research, 29(1), 21-36. https://doi.org/10.1177/10242589221141055
Dyer-Witheford,
Nick, Atle Mikkola Kjøsen, and James Steinhoff. 2019. Inhuman Power: Artificial Intelligence and the Future of Capitalism.
London: Pluto Press.
Elish, Madeleine Clare, and danah boyd. 2018. Situating
Methods in the Magic of Big Data and AI. In Communication
Monographs 85 (1): 57-80. https://doi.org/10.1080/03637751.2017.1375130
Edgerton, D. E. H. (2007).
Creole Technologies and Global Histories: Rethinking How Things Travel in Space
and Time. In History of Science and
Technology Journal, 1(1), 75-112. doi:10.13169/workorgalaboglob.15.1.0126
European Parliament. 2024. https://artificialintelligenceact.eu/the-act/ . Accessed 10/6/2024
Fagioli, Andrea. 2021. To
Exploit and Dispossess: The Twofold Logic of Platform Capitalism. In Work Organisation, Labour &
Globalisation 15 (1): 126-137. https://doi.org/10.2307/j50010512.
Fuchs, Christian. 2008. Internet and Society: Social Theory in the
Information Age. New York: Routledge.
Fuchs,
Christian. 2014a. Social
Media: A Critical Introduction.
London: Sage Publications.
Fuchs,
Christian. 2014b. Digital Labour and Karl
Marx. New York: Routledge.
Hardt, Michael and Negri,
Antonio. 2000. Empire. Cambridge, MA:
Harvard University Press.
Hirsch, Joachim and Roth, Roland. (1986). Das neue Gesicht
des Kapitalismus vom Fordismus zum Post-Fordismus. Hamburg: VSA-Verlag.
Gandini, Alessandro. 2019.
Labour Process Theory and the Gig Economy. In European Educational Research Journal 72 (6): 72-83. https://doi.org/10.1177/0018726718790002
Gain, Vish.
2024. SAG-AFTRA Strikes Again, This Time It’s a Deal with an AI Studio. In Silicon Republic. Accessed June 2, 2024.
https://www.siliconrepublic.com/business/sag-aftra-replica-studios-ai-voice-gaming-deal-agreement .
Jarrett, Kylie. 2022. Digital Labor.
Cambridge: Polity Press.
Jiménez, Juliana. 2024. Latino
Casino, Service Workers in Nevada Fear AI Could Replace Them. In Noticias Telemundo/NBC. Accessed June 2, 2024. https://www.nbcnews.com/news/latino/latino-casino-service-workers-nevada-fear-ai-threat-jobs-rcna136208 .
Jones, Phil. 2021. Work Without the Worker - Labour in the Age
of Platform Capitalism. London: Verso
Jones, Tom, and Angela Fu.
2024. Writers Guild Wins Protections against Artificial Intelligence. In Poynter. Accessed June 3, 2024. https://www.poynter.org/commentary/2023/writers-guild-wins-protections-against-artificial-intelligence-newsroom-unions.
Kullgren, Ian. 2023. Las Vegas Hotels
Reach Deals for New Union Contracts, Averting a Strike. In Los Angeles Times. Accessed November 18, 2024. https://www.latimes.com/business/story/2023-11-10/las-vegas-hotels-reach-deals-for-new-union-contracts-averting-a-strike
Lindgren, Simon, and Virginia
Dignum. 2023. Beyond AI Solutionism: Toward a Multi-Disciplinary Approach to
Artificial Intelligence in Society. In
Handbook of Critical Studies of Artificial Intelligence, edited by Simon
Lindgren, 163-172. Cheltenham: Edward Elgar Publishing.
Maffie, Michael David. 2020. The Role
of Digital Communities in Organizing Gig Workers. In Industrial Relations: A Journal of Economy and Society 59 (1): 123-149. https://doi.org/10.1111/irel.12251
Marjoribanks, Timothy. 2000.
The `Anti-Wapping’? Technological Innovation and Workplace Reorganization at
the Financial Times. In Media, Culture
& Society 22(5): 575-593. https://doi.org/10.1177/016344300022005003 .
Marx, Karl. (1976). Capital: A Critique of Political Economy.
Volume 1. London: Penguin Books.
Mosco, Vincent. 2004. Digital Sublime: Myth, Power, and Cyberspace.
Cambridge, MA: MIT Press.
Mueller, Gavin. 2021. Breaking Things at Work: The Luddites Are
Right About Why You Hate Your Job. London: Verso.
Nolan, Hamilton. 2024. Workers
Need to Fight to Make Sure AI Works for Them. In Jacobin. Accessed June 2, 2024. https://jacobin.com/2024/01/workers-labor-artificial-intelligence-technology .
PAME Hellas. 2021. EFood Delivery Workers Have Won! Accessed June 1, 2024. https://pamehellas.gr/efood-delivery-workers-have-won .
Pasquinelli, Matteo. 2023. The Eye of the Master: A Social History of
Artificial Intelligence. London: Verso.
Pinch, Trevor J., and Wiebe E. Bijker. 1984. The Social Construction of Facts and
Artefacts: Or How the Sociology of Science and the Sociology of Technology
Might Benefit Each Other. In Social
Studies of Science 14 (3): 399-441.
Pleios, George. 2017. Communication
and Symbolic Capitalism: Rethinking Marxist Communication Theory in the Light
of the Information Society. In Marx and
the Political Economy of the Media, edited by Christian Fuchs and Vincent
Mosco, 98-137. Chicago: Haymarket.
Rosenblat, Alex, and Luke Stark. 2016.
Algorithmic Labor and Information Asymmetries: A Case
Study of Uber’s Drivers. In International
Journal of Communication 10: 3758-3784
Scholz, Trebor. 2017. Uberworked and Unverpaid:
How Workers Are Disrupting the Digital Economy. Cambridge: Polity Press.
Sohn-Rethel,
Alfred. 2021. Intellectual and Manual
Labour: A Critique of Epistemology. Chicago, IL: Haymarket.
Simos, Manolis,
Konstantinos Konstantis, Konstantinos Sakalis, and Aristotle Tympas.
2022. AI Can Be Analogous to Steam Power or from the Post-Industrial Society to
the Fourth Industrial Revolution: An Intellectual History of Artificial
Intelligence. In ICON 27 (1): 97-116.
SVEOD. 2021. Notes on the Efood Labour Regime#3. Efood
SVEOD Employees' Committee. Accessed June 10, 2024. https://sveod.gr/?p=3262.
Rifkin, Jeremy. 1995. The End of Work: The Decline of the Global
Labor Force and the Dawn of the Post-Market Era.
New York: G.P. Putnam’s Sons.
Srnicek, Nick, and Alex
Williams. 2016. Inventing the Future: Postcapitalism and a World Without Work. London: Verso.
Tassinari, Arianna, and Vincenzo Maccarrone. 2020. Riders on the Storm: Workplace Solidarity
among Gig Economy Couriers in Italy and the UK. In Work, Employment and Society 34 (1): 35-54. DOI: https://doi.org/10.1177/0950017019862954
Taylor, Frederick Winslow.
1913. The Principles of Scientific
Management. New York and London: Harper and Brothers.
Tsardanidis, Gregory. 2024. #cancel_Efood:
Online Solidarity to Platform Workers. In Futures
of Work 27. Accessed June 10, 2024. https://futuresofwork.co.uk/2024/04/29/cancel_efood-online-solidarity-to-platform-workers/
Vercellone, Carlo. 2007. From Formal
Subsumption to General Intellect: Elements for a Marxist Reading of the Thesis
of Cognitive Capitalism. In Historical
Materialism 15 (1): 13-36. DOI: https://doi.org/10.1163/156920607X171681
Walker, Michael. 2024. The
Impact of AI on Contracts and Unionisation. In Handbook of Artificial Intelligence at Work, edited by Martha
Garcia-Murillo, Ian MacInnes, and Andrea Renda, 356-370. Cheltenham: Edward
Edgar Publishing.
Wood, Alex J., Mark Graham, Vili Lehdonvirta, and Isis
Hjorth. 2019. Networked but Commodified: The (Dis)Embeddedness of Digital
Labour in the Gig Economy. In Sociology
53 (5): 931-950. DOI: https://doi.org/10.1177/0038038519828906
Wood, Alex J.
and Vili Lehdonvirta. 2021.
Antagonism beyond Employment: How the ‘Subordinated Agency’ of Labour Platforms
Generates Conflict in the Remote Gig Economy. In Socio-Economic Review 19 (4): 1369-1396. DOI: https://doi.org/10.1093/ser/mwab016
Woodcock, Jamie. 2019. Marx at the Arcade: Consoles, Controllers
and Class Struggle. Chicago, IL: Haymarket.
Woodcock, Jamie. 2021. The Fight Against Platform Capitalism: An
Inquiry into the Global Struggles of the Gig Economy. London: University of
Westminster Press.
Woodcock, Jamie. 2022. Artificial Intelligence at Work: The Problem
of Managerial Control from Call Centers to Transport
Platforms. Frontiers in Artificial Intelligence 5: 1-9. DOI: https://doi.org/10.3389/frai.2022.888817
Yamat, Rio. 2024. As Robot Baristas
and AI Chefs Debut at Vegas Tech Show, Casino Union Workers Fear for Their
Jobs. In Los Angeles Times. Accessed
June 3, 2024. https://www.latimes.com/business/story/2024-01-13/robot-baristas-and-ai-chefs-caused-a-stir-at-ces-2024-as-casino-union-workers-fear-for-their-jobs.
Alexandros Minotakis
Alexandros
Minotakis holds a PhD in Media Studies from the
National and Kapodistrian University of Athens. He is
currently a postdoctoral researcher at University College Dublin, Centre for
Digital Policy. His research focuses on AI regulation through the engagement of
workers and trade unions.
Antonis Faras
Antonis
Faras is a PhD candidate at the National and Kapodistrian University of Athens. His research interests
include maintenance and repair studies, the history and philosophy of
technology, and political economy. His research focuses on the socio-political
implications of technological systems and infrastructures.
[1] The following should be also
taken into consideration: “from one standpoint the distinction between absolute
and relative surplus-value appears to be illusory. Relative surplus-value is
absolute, because it requires the absolute prolongation of the working day
beyond the labour-time necessary to the existence of the worker himself.
Absolute surplus-value is relative, because it requires a development of the
productivity of labour which will allow the necessary labour-time to be
restricted to a portion of the working day […] Once the capitalist mode of
production has become the established and universal mode of production, the
difference between absolute and relative surplus-value makes itself felt
whenever there is a question of raising the rate of surplus-value” (Marx 1976,
646).