Contesting Algorithmic Workplace Regimes in an Era of Flexible Despotism

Abstract

This article investigates recent claims that the growing use of algorithms is giving rise to a novel workplace regime. The article makes two conceptual contributions: first, it identifies the generic characteristics of this supposed algorithmic workplace regime. Second, it puts into question the exceptionalism of algorithmic workplace regimes. This is achieved by bringing the centrality of non-algorithmic management techniques in co-constituting algorithmic regimes into focus and by historically situating the regime’s emergence within the wider workplace regime literature. In doing so the article questions the novelty and distinctiveness of algorithmic workplace regimes, arguing that such regimes are better understood as a subtype of flexible despotism that can be traced back to the 1980s.

Introduction

The use of algorithms to automate the monitoring, evaluation and direction of workers is now widespread (Kellogg et al., 2020). For instance, a recent representative survey undertaken by Hertel-Fernandez (2024) found that two-thirds of workers in the USA experience electronic monitoring of their work and nearly half have their schedules or tasks assigned by automated systems. Moreover, ‘algorithmic management’ (Lee et al., 2015) is associated with a range of negative outcomes, including increased work intensity, elevated anxiety, greater health and safety risks, and an increased likelihood of injuries (Hertel-Fernandez, 2024). In Europe, high levels of algorithmic management have also been found. For example, Fernández-Macías et al. (2023) found that 20% of German and 35% of Spanish workers experience at least one form of automated allocation or evaluation of their work. Increasingly researchers argue that this growing use of algorithms at work is leading to a new more despotic regime of workplace control (Delfanti, 2021; Dörflinger et al., 2021; Griesbach et al., 2019; Miszczyński and Zanoni, 2025; Schaupp, 2022, Vallas et al., 2022; Wiggin, 2025). But does the increased use of algorithms at work truly result in a new regime that can be considered distinct from conventional management techniques and extant workplace regimes, and if so in which ways?

Workplace control is key to the functioning of capitalism. This is because employers must continually transform workers’ potential to produce value, what Marx (1976 [1867]) refers to as labour power, into actual valuable activity (Thompson, 1989; Thompson and Smith, 2000). The resultant ‘transformation problem’ forms the motivating problematic of Labour Process Theory. However, in his influential study of a Chicago machine shop, Burawoy (1979) demonstrated that workplace control rests not only on securing exploitation via force, but also on obscuring exploitation so that resistance and conflict are minimised (Burawoy, 1979, 2012).

According to Burawoy (1979, 1985) the process by which control is secured and obscured is dependent upon the ‘factory regime’ (or workplace regime) under which paid work takes place. Importantly Burawoy (1985) contends that it is possible, despite national and industry variations, to identify common generic regime characteristics. These regimes are conceptualised as consisting of both the labour process and internal institutions that regulate and shape struggles and conflict within the workplace (Burawoy, 1985). Burawoy (2009) identifies several contrasting regimes (market despotism; colonial despotism; bureaucratic despotism; hegemonic despotism), which he argues occupy unique positions upon a continuum between the use of force and consent.

Workplace regime theory has proved influential in the sociology of work with Burawoy’s (1979, 1985) two foundational texts being cited more than 11,000 times (Google Scholar, August 2025). Moreover, this framework has recently been applied by several researchers seeking to understand the impact of algorithms on work. For example, Schaupp (2022) argues that the use of algorithms by delivery platforms, warehouses and manufacturers has given rise to a new ‘algorithmic workplace regime’. Griesbach et al. (2019) identify delivery platform work as a regime of ‘algorithmic despotism’. Delfanti (2021) argues that Amazon warehouses constitute a regime of ‘augmented despotism’ and Vallas et al. (2022) and Miszczyński and Zanoni (2025) refer to them as ‘techno-economic despotism’. Meanwhile Dörflinger et al. (2021) contend that similar practices in Dutch and German warehouses constitute a regime of ‘coercive flexibility’. The present article, therefore, seeks to identify the generic features of this supposed new algorithmic regime by critically reviewing case studies of algorithmic workplace regimes undertaken across diverse industry and country settings (i.e. Delfanti, 2021; Dörflinger et al., 2021; Griesbach et al., 2019; Schaupp, 2022). Identifying the generic features of algorithmic regimes brings the distinctiveness of this regime into question in two ways. ¹ First, it makes clear the centrality of non-algorithmic traditional management techniques to accounts of algorithmic workplace regimes. Second, it enables the emergence of this supposed regime to be historically situated within the wider workplace regime literature so that the uniqueness of the regime’s reliance on force and consent can be evaluated. As the existence of a workplace regime is dependent upon possessing a discrete location upon the force–consent continuum, this article goes on to argue against understanding algorithmic workplace regimes as exceptional and contends that they are better understood as an intensification of the previously identified regime of flexible despotism, which can be traced back to the 1980s (Chun, 2001; Wood, 2020).

Workplace Regimes: From Market Despotism to Flexible Despotism

Burawoy’s (1979) initial formulation of workplace regimes entailed the identification of two sequential ideal-type regimes: ‘market despotic’ and ‘hegemonic’. He argued that these regimes were constituted by contrasting ‘internal states’ ² and ‘internal labour markets’ – combinations of workplace institutions that regulate the labour process and shape struggles within the workplace (Burawoy, 1979, 1985). As mentioned above, Burawoy (1979, 1985, 2009) argued that the resulting regimes occupied unique positions upon a continuum between the use of force and consent: ‘market despotic regimes’ being identified as representing the typical form of labour organisation in the late-18th-century capitalist core and ‘hegemonic regimes’ becoming dominant in the post-Second World War period (Burawoy, 1979, 1985).

Market despotism is characterised by tightly controlled labour processes regulated by the overseers’ threats of fines, dismissals and blacklists (and, in some cases, even physical violence), whereas hegemonic regimes rely on a stable compromise equilibrium between capital and labour; maintained by the interplay of the ‘internal state’ and ‘internal labour market’ (Burawoy, 1979). Hegemonic internal state institutions include employment contracts, grievance and disciplinary procedures, independent trade unions and collective bargaining. Collective bargaining plays an especially important role in this formulation for it not only provides workers with a voice in the creation and policing of rules but also concretely coordinates the interests of workers and capital by providing institutional means for profit sharing. Hegemonic internal states are complemented by internal labour markets constituted by administrative rules (particularly seniority rules and job ladders) that shield workers from the vagaries of the external labour market. Burawoy (1979) famously argued that the security provided by the combination of internal states and labour markets laid the basis for the generation of consent by enabling workers to develop absorbing work games. These work games focus workers’ attention upon increasing output rather than challenging their exploitation.

Burawoy (1985) would further elaborate his theory to account for the demise of hegemonic regimes in the early 1980s and their replacement by ‘hegemonic despotism’ in advanced capitalist countries. However, I have previously argued that as this third regime relied upon the very hegemonic apparatuses (i.e. collective bargaining and trade unions) that its operation undermined through concession bargaining, it was invariably short-lived (Wood, 2020). In this view hegemonic despotism is more constitutive of the breakdown of hegemonic regimes during a transition phase than a long-lived new regime in its own right. As a consequence, Chun (2001) and Wood (2020) suggest hegemonic regimes were ultimately replaced by flexible despotism in which it is the flexible allocation of labour that simultaneously secures and obscures exploitation.

Flexible despotism was first identified by Chun (2001) who highlighted the harsh discipline faced by temporary agency workers. These workers, usually drawn from economically marginalised populations, were threatened with dismissal if they did not meet exacting quality standards. However, Chun (2001) also emphasised how the promise of being gifted permanent employment – and the associated benefits – simultaneously created consent to this otherwise despotic regime (see Mears (2015) for an account of the importance of gifts to control in a very different setting). My retail research (Wood, 2018, 2020) builds on Chun’s (2001) insights regarding flexible despotism to demonstrate how temporal, as opposed to numerical, firm flexibility also enables this regime to secure and obscure exploitation through what I term ‘flexible discipline’ and ‘schedule gifts’. Flexible discipline refers to the use of on-demand scheduling by managers to punish workers without recourse to official disciplinary procedures (and in a more subtle, ambiguous and adjustable manner than traditional punishments) (see also Ikeler, 2016; Price, 2016). Meanwhile, building on Bourdieu’s (1977) discussion of gifts and symbolic violence, I argue that schedule gifts enable the obscuring of exploitation as workers have to beg managers for more and/or better hours in order to make ends meet, care for their loved ones or partake in valued social activities (Wood, 2018, 2020). That managers can use their discretion to grant workers’ desired schedule changes shrouds this practice as a gesture of generosity and kindness. As workers are unable to reciprocate in kind, schedule gifts bind them to their manager through emotional debt and a sense of moral obligation to repay the manager in the only way that they can: hard work. Under flexible despotism, the obscuring of exploitation has thus transitioned from ‘work games’ to ‘work gifts’. In fact, I found managers disincentivised and disrupted games that workers initiated (Wood, 2020).

Algorithmic Workplace Regimes: Empirical Evidence

Whether the increased use of algorithms to automate management functions is producing a new regime of control is key for understanding both the conditions under which contemporary work is undertaken and how capitalism functions in the information age. This next section seeks to identify the generic characteristics of algorithmic workplace regimes by reviewing key empirical findings of such regimes operating across diverse industry and country settings (i.e. Delfanti, 2021; Dörflinger et al., 2021; Griesbach et al., 2019; Schaupp, 2022).

Griesbach et al. (2019) identify a regime that they term ‘algorithmic despotism’ operating at US platform delivery firms, such as Instacart, DoorDash and Postmates. This regime is found to be constituted of algorithmic ‘black boxes’ that result in an internal state centred upon the automated evaluation and profiling of workers’ performance. For instance, metrics on task acceptance, hours and GPS data are used to evaluate workers with satisfactory performance being rewarded via bonuses and priority in the allocation of tasks and working time. Moreover, these algorithmic black boxes also result in an internal state that appears to workers arbitrary, opaque and subject to frequent unilateral alterations. Discipline is reinforced via algorithmic ranking systems based on customer ratings that both provide bonuses and ‘deactivate’ (or in other words fire) or filter work away from poor-performing workers. According to Griesbach et al. (2019) then the internal state in this regime is constituted by opaque rules that are unilaterally altered by management, exposing workers to a regime of arbitrary authority.

This sense of arbitrariness and opaqueness is found to extend to labour allocation with workers being directed in their tasks by algorithms so that these self-employed workers experience ‘little control over either their time or the activities that they perform’ (Griesbach et al., 2019: 9). Moreover, pay rates are set by dynamic pricing algorithms that match labour supply in real-time with customer demand. The use of opaque algorithms to dynamically price and/or allocate work is argued to enable the ‘gamification’ of work. Gamification refers to ‘introducing elements from games into the work environment with the purpose of improving employees’ affective experiences’ (Mollick and Rothbard, 2014: 2). Griesbach et al. (2019) argue that as a result of dynamic pricing workers develop strategies aimed at earning higher rates and influencing the jobs they are offered. Griesbach et al. (2019: 6) claim this is a ‘game of food delivery not dissimilar from the games described by Burawoy’. However, the degree to which such games generate consent by mystifying the labour process as opposed to simply incentivising increased productivity remains unclear. Indeed, Miszczyński and Zanoni (2025) argue that algorithmic management produces consent not via gamification but by reducing the need for traditional managerial relations and thus increasing workers’ sense of agency and control (see also Veen et al., 2020; Wood et al., 2018).

Schaupp (2022) likewise documents an algorithmic workplace regime in German platform delivery work in which smartphone apps direct delivery riders to exactly where they need to go and what they need to do. This algorithmic direction is argued to enable training and language requirements to be diminished so that large numbers of migrants can be engaged via temporary contracts. Moreover, GPS monitoring via smartphones enables workers’ performance to be automatically profiled to determine contract renewals. When workers’ productivity drops the app automatically notifies them via a noise or phone call.

Interestingly Schaupp (2022) finds similar algorithmic workplace regimes in the very different settings of German warehouse and manufacturing work. Handheld digital scanners are used to direct workers around the warehouse. The devices additionally dictate the times and speed that workers should work at. As a result, Schaupp (2022) finds that training could be reduced to one-and-a-half days, local language proficiency was no longer necessary and a high labour turnover could be sustained. Moreover, migrants were found to make up 70% of the workforce in some warehouses. The handheld device also monitored every movement that workers took and evaluated their productivity, which was automatically fed back to workers, with the effect of intensifying work. Schaupp (2022) also investigates two manufacturing companies (one in chemicals and one in electronics) and here too digital systems monitored and evaluated workers’ productivity either automatically assigning them additional tasks or tracking in seconds the time between the completion of tasks and displaying their performance in comparison with the average. Again, the aim of algorithmic management is a greater utilisation of migrant and lower-skilled workers. As a manager of an electronics company put it, ‘the aim of the system is “either to make things go faster or to enable people with less qualifications to do it”’ (Schaupp, 2022: 316). This is achieved by having workers enter descriptions and pictures of their work into the digital system so that untrained and migrant workers can be directed by the system without needing to speak German.

Delfanti (2021, 2022) provides an especially detailed account of a regime that he terms ‘augmented despotism’ at Amazon in Italy. As in Schaupp’s (2022) German case studies, handheld scanners record and evaluate workers’ performance. Performance metrics, both historic and real-time, are thus available to management at the press of a button. Moreover, poor-performing workers are automatically flagged to managers to be disciplined. If workers work too slowly, or spend too long off task, a message pops up on their device’s screen: ‘Meet team leads for a feedback session’ (Delfanti, 2022: 55). Temporary workers are especially at risk of being disciplined (Delfanti, 2022: 70). Vallas et al. (2022) researched Amazon warehouses in the USA and concluded that a ‘sophisticated algorithmic system . . . places workers at risk of termination if they fail to “make rate”’. Vallas et al. (2022: 435) contend that this system differs from the technical control embodied by the assembly line because workers ‘can be held individually accountable to a greater degree than was true of assembly line workers in the past’.

There are also similarities with the US platform regime uncovered by Griesbach et al. (2019), with Delfanti (2022: 75) highlighting how the use of algorithms to manage warehouses is likewise a highly opaque process, as only managers are able to view the aggregated performance data that workers are benchmarked against. Moreover, workers are only provided with vague details by managers (such as the percentage of the target achieved) without disclosing what the target is or how it is calculated. Again, handheld scanners not only enable the monitoring and evaluation of workers’ performance but also algorithmic direction and labour process simplification (Delfanti, 2022). Specifically, Delfanti (2021, 2022) explains how handheld scanners enable tasks to be broken down and assigned to workers who are directed via instructions displayed on the device. This algorithmic direction of workers reduces training time to a few hours and enables workers to be quickly put to work (Delfanti, 2022: 104). As with platform work, algorithms enhance the ability of capital to control labour outside of the standard employment relationship without high labour turnover having a deleterious impact on company operations (Delfanti, 2021). For example, Delfanti (2021, 2022) finds that major temporary work agencies Adecco and Manpower have offices on Amazon’s premises, providing workers who are sometimes contracted for as little as a week and whose hours can be altered with just 24 hours’ notice. Algorithmic direction reducing training requirements further opens up the possibility of employing workers with precarious migration statuses (Delfanti, 2021). Likewise, Delfanti (2022: 79) reports that Amazon makes use of ‘masses of women, migrants and racialised workers’ to fill low-level positions. Doing so ensures a reliable on-demand workforce comprised of precarious workers who cannot refuse the sudden imposition of unplanned shifts or overtime (which Amazon makes clear is mandatory) (Delfanti, 2022: 86).

These accounts highlight the importance of factors unrelated, or only indirectly related, to technology in co-constituting these regimes. Indeed, Vallas et al. (2022) argue that at Amazon in the USA, the power of algorithmic work systems ultimately rests upon the economic dependency of workers with few alternatives available to them (see also Barnes and Ali, 2022). As Miszczyński and Zanoni (2025) put it, these regimes combine technology in the form of sophisticated algorithmic systems with economic relations in the form of casualised employment and labour market vulnerability. Likewise, algorithmic workplace regimes are only possible because the collective regulation of the internal state is minimal. These regimes are openly hostile to unions, including deploying anti-union propaganda and analysts and managers tasked with monitoring and intimidating labour activists (Delfanti, 2022). Indeed, Wiggin (2025) highlights how algorithmic management can be effectively weaponised to counter unionisation attempts and keep firms, such as Amazon, union free. Even when workers have some success in collectively organising, they find the extensive use of temporary agency workers a difficult barrier to overcome (Delfanti, 2022). It is this combination of limited collective organisation and high levels of individual dependency that renders workers unable to influence the algorithmic system and determines their work experience.

Again, as in the US platform work studied by Griesbach et al. (2019) gamification is also identified by Delfanti (2021) as a key method of labour allocation in warehouses. For example, Amazon makes use of ‘power hours’, during which increased productivity levels are rewarded with ‘prizes, such as a company t-shirt, as well as public recognition in front of the team’ (Delfanti, 2021: 50). Delfanti (2022: 65) describes this gamification in detail:

‘maybe they tell you, tomorrow both the morning shift and the night shift will do a power hour and then they decide what the best team is. And if you win, they give Amazon swag, a water bottle, a T-shirt.’ Prizes don’t go to the entire team but, as Tina put it, ‘to the person who did more pieces at the assembly line’.

Delfanti (2022) also reports the use of video games viewed on tablets positioned throughout the warehouse. Workers play these games by completing their work tasks. For example, in a car racing game, picking items from the shelves increases the speed of the worker’s virtual car on a racetrack, with workers winning company merchandise. Vallas et al. (2022) report this gamification to be at least somewhat successful, with some workers keeping track of their relative productivity score and seeing their work as a race. Likewise, Delfanti (2022) reports that power hours successfully squeeze high levels of productivity from workers who find it ‘cool’ that they receive individual prizes.

But here too we find that the algorithmic regime is intertwined with non-technological methods of control with Delfanti (2021, 2022; see also Vallas et al., 2022) detailing Amazon’s attempts to legitimise exploitation (see Burawoy and Lukács, 1992). These include the Orwellian use of language and organisational culture. For example, Amazon refers to warehouse workers as ‘associates’ and to disciplinaries as ‘coaching’, as well as creating symbolic positions such as ‘learning ambassador’, ‘problem solver’ and so on, and deploying slogans, such as ‘Work hard. Have fun. Make history’. The company additionally provides table football in canteens and punctuates the daily grind of the work with celebratory days, such as ‘Hawaiian Day’, ‘Chocolate Day’ and ‘1990s Day’ (Delfanti, 2022: 60). Amazon also requires workers to engage in daily rituals as Delfanti (2021: 49) reports:

management . . . impose worker participation in Amazon’s workplace culture. For example, [during briefings] workers are asked to raise their hand and suggest a ‘success story’ in front of the rest of the team . . . managers may also say something about the team’s performance, which workers are implicitly required to celebrate. As reported by a worker, managers at times say things like ‘Yesterday we had an insane productivity rate!’ followed by applause.

Vallas et al. (2022) argue that these practices reinforce the perceived virtue of hard work and lead to the internalisation of a ‘diligent worker’ identity or as Delfanti (2022: 61) puts it, a ‘work hard/play hard’ culture modelled on Silicon Valley tech firms. However, these attempts at normative control are also found to have limited effectiveness due to contradicting the material realities of the work. For example, Delfanti (2022: 58) reports workers referring to Amazon’s ritualised briefings as ‘“dog and pony shows” or “Alcoholics Anonymous meetings”’. Delfanti (2022) highlights several hidden ways in which warehouse workers can misbehave and resist. These acts range from theft to sabotage, such as placing items in incorrect spaces to save time or picking up an item such as a book and reading it before putting it back on a random shelf – dooming such items to be lost to the algorithms for eternity. Moreover, workers have increasingly engaged in overt forms of resistance, such as joining unions, striking and participating in slowdowns (see also Schaupp, 2022).

The role of non-technological factors in co-constituting algorithmic workplace regimes is further evidenced by Dörflinger et al. (2021) who uncover a similar regime at Dutch and German warehouses, which they term ‘coercive flexibility’. Again, this research highlights the use of digital devices to track and monitor workers with managers being able to see at every moment the location of employees, what tasks they have completed that shift and what they are doing in real-time. High levels of scheduling and income uncertainty are found to be endemic. But as at Amazon the power of the algorithmic work systems ultimately rests upon the economic dependency, with these workplaces also making use of high levels of temporary agency workers (20–70% of the workforce). Importantly Dörflinger et al. (2021) demonstrate that the algorithmic regime not only rests on individual workers’ lack of power but also workers’ limited collective power. Dörflinger et al. (2021) find regimes of ‘hegemonic flexibility’ and ‘institutional mutuality’ operating in Germany and Belgium when workers have sufficient collective power to constrain the use of algorithmic technologies as mechanisms of control.

Algorithmic Workplace Regimes: Generic Characteristics

The above review of algorithmic workplace regime research highlights how such regimes rely upon opaque systems that automate the direction of the work and arbitrarily manage performance via ‘black boxes’. The use of these systems enables the fissuring of employment relationships via self-employment and temporary agency workers and likewise limits the need for investment in on-the-job training. As a result, few rights or responsibilities are enshrined in workers’ contracts and there are no seniority rules, limited real job ladders and limited possibilities for vertical mobility. Algorithms also enable the dynamic allocation of tasks and the extensive use of gamification. However, it is precarity that ultimately provides algorithmic work systems with their power (Vallas et al., 2022). Algorithmic workplace regimes are thus co-constituted by both technologies and non-algorithmic practices. Their operation requires both individually and collectively precarious workers while at the same time expanding the possibilities for utilising such labour by reducing skill, training and language barriers and heightening precarity by subjugating workers to opaque and arbitrary automated management systems. Likewise, the obscuring of exploitation within this regime rests on the combination of algorithms with more traditional management practices. Algorithmic gamification combines with normative controls aimed at legitimising the workplace. The use of normative control is frequently identified in platform work and this setting provides insights into how precarity, normative manipulation and algorithmic gamification combine. Vieira (2023), for instance, documents how precarity and gamification combine to reinforce entrepreneurial identities among platform delivery workers. Having outlined how the confluence of algorithmic and non-technological features combines to strengthen and shape the generic characteristics of algorithmic workplace regimes it is now possible to turn to the question of whether this regime represents a distinct identifiable position upon the continuum between force and consent, and thus whether algorithms are constituting a novel workplace regime; doing so requires locating the above research within its historical context.

Situating Algorithmic Despotism in Its Historical Context: The Transformation of Workplace Regimes 1980s–2010s

Post-Hegemonic Regimes: Lean Production

Workplace regime research during the post-hegemonic period initially focused on the spread of lean production in manufacturing. This research detailed an emerging regime of employer-controlled flexibility, surveillance (whether electronic-, customer- or team-based), strict discipline and limited labour protections (Turnbull, 1988). Delbridge et al. (1992: 98) argued that this regime severely limited workers’ task autonomy and control over their work, and thus ‘more completely subordinated labour to capital than previous production regimes’. This was due to lean production entailing enhanced labour process visibility, surveillance and monitoring even while reducing the number of supervisors and managers involved in these activities. This was achieved via transparent workplace layout and the use of information systems. As Turnbull (1988: 12) explains, workplaces in such regimes are laid out to ‘create an “invisible conveyor” between “multi-functional workers” operating several different machines’. Delbridge et al. (1992) describe team performances being displayed alongside production targets on television screens throughout the factory and updated every hour. These researchers also highlight the role of customers in performance management, with one company using customer satisfaction scoreboards placed next to work teams. Likewise, Sewell and Wilkinson’s (1992) study of ‘Kay Electronics’, a Japanese-owned manufacturer, documents the use of surveillance via the public display of quality and productivity information, which provides immediate feedback to both individuals and work teams.

Work teams were central to Kay’s lean production and were comprised of 12 to 40 members who assembled printed circuit boards (Sewell, 1998). Production and quality targets were set for each team and every team was collectively responsible for their achievement but had a ‘degree of discretion as to how they deploy their human resources within the team in order to achieve them’ (Sewell and Wilkinson, 1992: 281). Team members then acted to self-monitor each other and confronted those who were not ‘pulling their weight’ (Sewell and Wilkinson, 1992: 281). As a result, considerable peer pressure was created (Delbridge et al., 1992). Most importantly, however, Sewell and Wilkinson (1992) document the use of electronic testing at each stage of production to identify errors, with the results automatically relayed to a central inventory control database so that the individual responsible could be instantaneously identified (Sewell, 1998: 412). Sewell and Wilkinson (1992: 283) argue:

the development and continued refinement of electronic surveillance systems using computer-based technology . . . [enables the] open prison of JIT/TQC [Just in Time/Total Quality Control] work team, with discipline ensured through the means of electronic tagging . . . [It is in] expos[ing] an individual as the source of the failure within such a short time delay . . . [that] provides the disciplinary power.

Punishments for poor performance in this regime were found to involve not only threats, material penalties and, ultimately, the sack but also public humiliation in front of teammates. For instance, at the start of each shift, coloured cards corresponding with the previous day’s performance (e.g. red for exceeding acceptable quality standards, green for no errors and amber for some errors) were placed above workstations. Performance for the previous 20 days’ work was also displayed (Sewell, 1998). Delbridge et al. (1992: 103) thus argue lean production constitutes a regime of ‘responsibilities without rights’. Importantly, Turnbull (1988) argues that the weakness of trade unions during this period was central to enabling the development of lean regimes. Thus, we can see that as far back as the 1980s the deployment of electronic systems of control in the absence of collective regulation enabled employers to achieve the individual accountability that Vallas et al. (2022) argue to be a hallmark of algorithmic methods of control.

Post-Hegemonic Regimes: Call Centres

In the late 1990s, workplace regime research increasingly moved away from manufacturing and instead focused on service work; with the call centre frequently seen as archetypal of the emergent post-industrial work setting. Early research into this setting emphasised the increased possibilities for computers to enable the monitoring of workers in real-time. For instance, Fernie and Metcalf (1998: 2) argue, that ‘for call centres, Bentham’s Panopticon was truly the vision of the future and these organisations are the very epitome of what Foucault had in mind’. While the use of the panopticon metaphor, Foucault, and the attendant lack of attention to resistance have been severely critiqued (e.g. Bain and Taylor, 2002), the high level of electronic surveillance in such work has, nevertheless, been widely reported (e.g. Taylor and Bain, 1999; Woodcock, 2017). For instance, Taylor and Bain (1999) found that call centre workers were subjected to monitoring via ‘mystery shoppers’ as well as real-time covert remote electronic monitoring of customer interactions. Moreover, this surveillance fed into performance monitoring, with workers receiving either a red, amber or green mark on eight quantitative and qualitative elements of the interaction – with four reds constituting failing the observation. Workers were disciplined for poor performance via ‘coachings’ that could lead to being fired (Bain and Taylor, 2002: 11). Likewise, at other call centres, computer software enabled electronic surveillance by tracking workers’ activities, automatically comparing worker performance and issuing ‘alarm’ reports for poor-performing workers. This system was used by team leaders or managers to ‘coach, cajole or discipline “under-performing” workers’ (Taylor and Bain, 1999: 108).

Type-written prompts and on-screen templates were used in an attempt to script the work (Taylor and Bain, 1999). Moreover, what would nowadays be termed algorithmic direction was present, with computer software initiating calls and allocating them to workers, thereby determining the pace of work (Bain and Taylor, 2002; Callaghan and Thompson, 2001). While resistance and trade union activism are documented in call centres, so too is the limited institutional power of trade unions and the hostility of these firms to organised labour.

Callaghan and Thompson (2001: 22) make an important contribution by highlighting that digital technology not only enables direction, monitoring and evaluation but also an embryonic form of gamification:

the walls adjacent to different teams often have tables comparing CSRs [customer service representatives] within teams and between each team in the call centre. Such comparisons pressurise not just individual CSRs, but also team leaders who have to compete against each other.

As a consequence of this gamification, exploitation is partially obscured in this regime by camouflaging the nature of control through rendering decisions as ‘objective’ due to both the centrality of informational statistics and ‘objective’ bureaucratic standards (Callaghan and Thompson, 2001). Fleming and Sturdy (2011) provide further details of how exploitation can be obscured in such settings through normative controls. In particular, workers were found to be encouraged to ‘be themselves’ by practising freedom in terms of lifestyle, sexual and consumer displays while still experiencing extensive coercive controls over their work. However, the contradictions between these normative controls and the more coercive technical and bureaucratic work controls (such as those outlined above) led this ideological framing to be unstable, with many workers exhibiting scepticism towards the positive symbolic representation of their employment experience. Sallaz’s (2015) ethnographic research at a US call centre offers an alternative route by which exploitation can be mystified via a ‘learning game’. In this particular game, Sallaz (2015) argues that workers are faced with the challenge of mastering customer interactions. Moreover, they do so lacking training and facing immense social pressure from customers to handle the calls competently. However, the call centre researched by Sallaz (2015) was distinctly less despotic (seemingly due to labour shortages and difficulty standardising the type of interactions involved) than those investigated in most workplace regime research. Furthermore, the ability for this game to be absorbing and create commitment was clearly limited as most workers quit within a few weeks.

Post-Hegemonic Regimes: Retail

In the early 2010s, the focus of workplace regime research again shifted, this time to the retail sector (e.g. Ikeler, 2016; Price, 2016; Wood, 2018, 2020). Interest in retail was driven by a view that it exemplified ‘the new generic form of mass employment in the post-industrial social-economic landscape’ (Bozkurt and Grugulis, 2011: 2). For instance, Lichtenstein (2009) and Vidal (2012) argued that US retailer Walmart was paradigmatic of employment in the 21st century, much as General Motors and Ford were during the previous century. This retail-based research again uncovered a regime of digital surveillance via scan and pick rates, CCTV and task management software (Ikeler, 2016; Price, 2016; Wood, 2020). For instance, at a US retailer, referred to as ConflictCo, I document digital surveillance taking three forms: task direction software, which monitored workers’ speed at restocking shelves; checkout scan rate software, which monitored workers’ speed processing customer purchases; and store-wide CCTV cameras. These systems of surveillance provided overlapping means for monitoring performance. For example, ‘my guide’ task direction software flagged workers exceeding specified time allocations for completing restocking – the system being renamed ‘my-slave’ by resentful workers. Likewise, scan rate software monitored the speed with which workers scanned products – workers being rebuked by managers if they failed to reach a specific scan rate. These performance monitoring tools were reinforced by the extensive use of CCTV cameras. Ostensibly, these cameras were installed to protect against customer theft but left workers feeling they were being constantly watched by management. CCTV was particularly important in facilitating the effectiveness of task monitoring tools as it enabled managers to check the behaviour of workers failing to meet productivity rates and verify their excuses (Wood, 2020). This digital surveillance fed into harsh disciplinary practices facilitated by workers having limited protections provided by the law, trade unions, employment contracts or developed disciplinary and grievance procedures. As one worker in this study of ConflictCo put it: ‘You can get a coaching for anything man, ’cause it is so crazy. Yeah, there has been some cool people fired over some bullshit reasons, good people’ (Wood, 2020: 63).

Once again it was workforce precarity that gave surveillance its power. Ikeler (2016) thus terms this regime one of ‘contingent control’ as both employment and schedule insecurity provide a means to secure control via the threat of job loss or cuts to hours. Price (2016) similarly emphasises the role of flexible scheduling and the ability to reward and discipline workers via changes to their hours as a mechanism of control. However, as discussed above, I go further arguing that the prevalence of flexible ‘on-demand’ scheduling (often undertaken in real-time) to match labour supply with customer demand not only creates the possibility for managers to ‘flexibly discipline’ workers without recourse to official disciplinary procedures (making it more subtle, ambiguous and adjustable than traditional punishments), but also obscures exploitation via ‘schedule gifts’ (Wood, 2018, 2020). That flexible scheduling enables both the securing and obscuring of exploitation, led me to build on Chun (2001) in terming this regime temporal ‘flexible despotism’ as a counterpart to Chun’s numerical ‘flexible despotism’ (Wood, 2020).

I show that the obscuring of exploitation via schedule gifts takes place within the context of ineffective normative and hegemonic controls (Wood, 2020). For instance, at the ConflictCo case study, the firm attempted to frame itself and its founder as pro-worker through in-store posters and magazines. As one worker explained: ‘In the break room, they have all those kinds of magazines and all these stories and things about ConflictCo, and the founder and they never say anything bad about ConflictCo’ (Wood, 2020: 66).

As at Amazon this framing includes an Orwellian use of language where employees are termed ‘associates’, the grievance policy ‘the open-door policy’ and the disciplinary procedure ‘coaching for success’. This terminology supposedly reflects the founder’s official ‘rules of business’, which emphasise the sharing of profits, partnership with and appreciation of employees and ConflictCo’s ‘core business belief’ in listening to all its workers. Moreover, this regime makes heavy use of anti-union propaganda with all new recruits being required to complete a computer module that depicted trade unions as only interested in taking workers’ voice and money without providing anything in return (see also Ikeler, 2016). In addition, ConflictCo placed anti-union notice boards and slide shows next to the time clocks. As one worker put it:

ConflictCo had at the two time-clocks televisions on each side and it’s constantly streaming about the worker association and the union and how evil they are and how many jobs they have lost and how many stores have closed because of the union and how you shouldn’t have to pay anyone to speak for you and you shouldn’t have to be involved in anything like that when you have an ‘open door’ policy. (Wood, 2020: 67)

Finally, workers were forced to take part in ritualistic cheers at the start of each shift: workers had to enthusiastically spell out the name of the company along with phrases such as ‘We are number one!’ and ‘Whose ConflictCo? My ConflictCo!’ (Wood, 2020: 68). However, in emphasising values of shared profits, partnership and worker voice that contradicted workers’ material lived experience of the workplace, these normative controls acted only to strengthen feelings of injustice and legitimise resistance. The evidence suggests that while normative control may frequently be attempted, rarely is it entirely successful. This perhaps explains why conflict is common in platform work (Martindale et al., 2024; Schaupp, 2022; Umney et al., 2024; Wood and Lehdonvirta, 2021).

Algorithmic Workplace Regime Exceptionalism?

In reviewing evidence from platform, warehouse and manufacturing work, it is possible to distil the following common elements characteristic of algorithmic workplace regimes across diverse industry and country settings. The algorithmic despotic internal state was found to rely on the use of automated performance management systems, digital work direction, the fissuring of the standard employment relationship and limited opportunities for collective regulation. Internal labour allocation was meanwhile found to be based upon the use of algorithms to dynamically allocate tasks and to gamify work undertaken by a workforce lacking seniority protections, real job ladders and on-the-job-training. Moreover, this regime was identified as making extensive use of highly dependent (migrant) labour. Attempts to obscure exploitation mainly take place via normative controls that seek to replicate Silicon Valley influenced notions of the ‘diligent worker’ (Delfanti, 2022; Vallas et al., 2022), yet these framings are frequently rendered ineffective by the material conditions that contradict them.

The reliance on algorithmic performance management, digital direction, dynamic task allocation and gamification, and the use of normative controls to obscure exploitation, give this regime an appearance of novelty. However, this article has questioned the distinctiveness of this regime by highlighting the centrality of non-algorithmic traditional management techniques in co-constituting the regime (in particular, precarious labour, union avoidance and normative controls). Moreover, the question of whether it represents a distinct identifiable position upon the continuum between force and consent can only be answered by locating algorithmic despotism within its historical context. As the article has shown, post-hegemonic regimes in lean production, call centres and retail have, in fact, long made extensive use of digital technology to facilitate electronic surveillance to both enforce performance through metrics and to enable the work to be incentivised via gamification and obscured via ‘gifts’. Indeed, in one study of warehouse work, Vallas et al. (2022) find that a more successful means of obscuring exploitation at Amazon is via the ‘gifts’ of employment and voluntary time off that bound workers to managers via feelings of gratitude. The power of these work gifts rests upon the vulnerability of workers to threats of job loss, changes to schedules and desire for permanent contracts, as well as the weakness of trade unions in collectively regulating workplaces in the capitalist core during the post-1980s era.

Seen in this light, algorithmic regimes do not then occupy a distinct position on the continuum between force and consent but rather represent the amplification of trends towards the despotic use of digital surveillance and direction, gamification, and the sidelining of trade unions present in core capitalist economies for the past 45 years. Indeed, as far back as the 1980s algorithmic systems of control were identified as seeking to ‘reduce decision-making as much as possible to a set of self-contained rules . . . implemented by a computer’ (Appelbaum and Albin, 1989: 252). In developing this finding Vallas (1993: 12) outlined the emergence of an ‘algorithmic regime’ at AT&T in which ‘programmable machines acquire the ability to regulate the internal operations of the firm – to control the flow of work between different departments . . . and to assign tasks to human workers’. We should, therefore, be wary of viewing the use of algorithmic systems within the workplace as wholly novel or overly exceptional. Moreover, algorithmic technologies in the workplace seemingly strengthen attempts to obscure exploitation via normative controls and gifts. Thus, algorithmic workplace regimes can be understood as the embedding of extant post-hegemonic relations of domination within workplace technologies that solidify existing despotic trends. Moreover, the increasing deployment of these technologies may herald the social closure of more democratic workplace regimes if efforts are not made to imprint more egalitarian ‘scripts’ (Wajcman, 2006: 775) within these technologies.