1.1 The Limits of Industrialism

Scholarly perspectives on gig work owe a great debt to ethnographic studies of labor and automation in the industrial workplace (Braverman, 1974; Burawoy, 1979 & 1985; Stark, 1980; Vallas, 1993). Industrialism’s hierarchies valorize managerial forms of labor over shopfloor work, while spatial distinctions further ingrain the domain of the workers from upper management’s offices. In the Taylorist system and in Henry Ford’s factories, profits increase through task differentiation, managerial oversight, and labor automation (Burawoy, 1985; Marx, 2008). Managerial activities and their associated technologies resolve worker compliance problems like lateness, turnover, inefficiency, or unionization (Braverman, 1974). Workers can resist this oppressive regime outright, or they may contribute to firm goals through alliances with managers or playing games on the shop floor to ensure standards are met and outputs remain high (Burawoy, 1985; Stark, 1980). Managers, for their part, maintain distinct positions under the guise that workers need constant surveillance and instruction to control production, protecting their own role as “high skill” and irreplaceable (Braverman, 1974). Job restructuring and increased managerial control are the order of the day, as is the decline of worker voice, valorization, and representation (Vallas, 1993).

Scholars of the gig economy and automation have predominantly looked to models of the factory as analytical comparatives, especially as key scholars of industrialism have transitioned to studying contemporary conditions of digitally-managed labor (Vallas & Schor, 2020). Much as early studies of labor on the shop floor describe worker’s processes for improvising, or resisting on the job (Burawoy, 1979), algocratic workplaces are also subject to “making out” (Cameron, 2022), resistance based on judgment (Brayne & Christin, 2021; Kellogg et al., 2020), or sabotage (Levy, 2015; Merchant, 2023). Despite being a terrain of contested control (Kellogg et al., 2020), in a presumed algocracy one cannot argue with the algorithmic manager — despite the fact that its instructions for completing a task are often simple, incomplete, and devoid of everyday complexities (Aneesh, 2009). Workers find themselves in an “invisible cage”, without voice, subject to opaque workplace changes to which they must uncompromisingly adapt (Rahman, 2024). Hence workers describe their algorithmic bosses as cruel (Gray & Suri, 2019), despotic (Griesbach et al., 2019; Vallas et al., 2022), restrictive (Cameron & Rahman, 2022) or biased (Raghavan et al., 2020). With little in the way of a collocated workplace, crowdsourcing information on ancillary platforms online is seen as an important locus for collaboration, information exchange, training, identity work, and organizing among workers in analyses that align with prior industrialist forms (Orr, 1996; Petriglieri et al., 2019; Rosenblat, 2018; Schwartz, 2018).

This arrangement accounts for many elements of gig-work, especially visible in warehouses or automated assembly lines (Delfanti & Frey, 2020; Vallas et al., 2022; Vallas & Kronberg, 2023). Yet challenges to our theorizing about organizational structure remain. How should scholars interpret the extraordinary degree of worker discretion alongside app-based surveillance, the presumed sense of autonomy at the same time as being a “slave” to the algorithm (Alvarez de la Vega et al., 2023; Aneesh, 2009; Ravenelle, 2019; Rosenblat, 2018; Schor et al., 2020; Shestakofsky & Kelkar, 2020; Ticona, 2015)? What of the individuated work yet persistent and necessary lateral collaboration between various workers in order to get the work done (Dalal et al., 2023; Griesbach et al., 2019; Kellogg et al., 2006; Shestakofsky, 2017)? What of the continued space for resistance and challenges to controlling technologies (Brayne & Christin, 2021; Cameron & Rahman, 2022; Kellogg et al., 2020; K. Levy, 2024)? Scholars have grappled with these paradoxical conditions of gig work to elaborate the cultural reasons why this might be the case, including company rhetoric, compliance with management, solidarity among workers, or embracing forms of “making out” (Burawoy, 1979; Cameron, 2022).

Following Michael Burawoy’s (1985) injunction to use worker experiences as a structural lens upon firm organization, we take these contradictions of gig work seriously as an indication of the gig firm’s entrenched dual structure. After all, factories with industrial hierarchies are only one way of organizing a company, and worker control and surveillance only one way of producing profit or commitment. Further, the preponderance of gig economy systems and platform firms of the 21^st century were born of the managerial philosophies of the 1990s, not those of the 1920s. In the intervening era, technology firms restructured from industrial to flat hierarchies, reconfiguring key elements of worker control and commitment, firm communication structures, and managerial discretion away from top-down control and toward shared organizational culture and, especially, worker self-discipline and self-starterdom (Heckscher & Donnellon, 1994; Kunda, 1992; Stark, 2009; Turco, 2016; Vallas, 2006; Vertesi, 2020).

Histories of these new firm structures do not begin with Ford or Taylor, as do studies of automation. Instead, they examine the consolidation and capture of markets and infrastructures in the late 19^th century, like small firms or railroads (Chandler, 1977; Langlois, 2023; Osterman, 1996). These firms experimented with distributing managerial tasks across lower tiers of the company, outsourcing core competencies, folding in competitors, and squeezing out middle management to redistribute and peripheralize tasks instead of centralizing them. Recent firms may even orient toward venture capital to buffer dramatic profits instead of through the implementation of a fully functional technical system (Shestakofsky, 2024). These visions and versions of firm management, control, and worker identity are arguably as ideologically sacred to contemporary Silicon Valley technology companies as Scientific Taylorism was to early industrial capitalists (Heckscher & Adler, 2006; O’Mara, 2019; Turner, 2009).

Most importantly for the gig firm, post-bureaucracies do not generate profit through increased efficiencies gained via increasing worker automation or control. Instead, they continually seek reinvention toward profitability, control, and commitment through creative restructuring (Kunda, 1992; Vallas, 2006). This makes for downstream changes to the scope of work, available roles within an organization, and the development of technologies that pick up the slack. In order to understand contemporary workplace pressures among the algorithmically-managed workforce, we must trace these structural changes and the pressures exerted upon managerial roles as firms flattened.

1.2 The Demise of the Managers

Post-bureaucracy as an organizational form did not achieve ascendancy out of nowhere. Its characteristic pressures must be understood in the historical trajectory of the American firm’s evolution¹, which lends context to ensuing changes in managerial work within such firms. In the height of bureaucratic hierarchies in the 1970s, accounts of the ratio of managers to workers at the American firm make much of its perceived dis-proportionality: for instance, from 18 administrators per 100 workers in 1929, to 30 administrators per 100 workers in 1970 (Osterman, 1996). This is because by the 1970s, a significant bureaucratic layer of middle management was expected to coordinate many simultaneous, scaled-up production processes (Chandler, 1977). With multiple product lines and complex systems of coordination across geographies and time zones, middle managers were tasked with maintaining consistent communication between workers and teams, coordinating production and output, managing production cycles and operational decision-making. Middle managers also had to ensure workers had the necessary equipment they need to complete a task, established clear work schedules, and gave workers enough information, feedback and support to complete tasks and improve upon their work.

This pattern ended in the recession-fraught 1990s, when Wall Street investors pressured firms to adopt cost-cutting strategies such as the “High Performance Work Organization” or “Total Quality Management” (Ashton et al., 2002). This was marketed as an opportunity for firms to improve production output and redistribute higher gains to fewer remaining employees not by restricting worker wages, but by reducing the size of the internal bureaucracy, i.e. cutting managerial labor costs. Osterman argues that such companies became top-heavy, requiring ever greater numbers of upper managers in order to maintain relationships with stakeholders like executive board members, venture capitalists, and shareholders (Osterman, 2006). Certainly, Goldstein shows that pushes to reduce middle managers in this period paradoxically resulted in an even greater managerial class under firm restructuring (Goldstein, 2012). Regardless, productivity gains between the 1970s and 2000s were largely captured by remaining senior managers rather than lower-tier workers. Under this paradigm, any large layer of middle management was subject to suspicion by external consultants and corporate boards alike, despite their core function of coordinating production and managing workers.

As middle management layers thinned, managerial styles evolved from roles with clear responsibilities and boundaries exerting direct oversight over a scaled workforce, to forms that encouraged more improvised managerial labor ever lower in the firm’s hierarchy. Organizational fads during this period such as the Matrix Organization required workers to “manage up” and “manage across” (Burns, 1989; Davis & Lawrence, 1979; Morrill, 1995). Those managers that remained were supposed to be constantly “learning” and “innovating” by problem-solving on tasks outside their specific job descriptions: framed as an opportunity for gaining new and important skills required for expanding roles. Local improvisation and responsiveness to emergent issues were expected to be resolved by workers lower in the chain of command in the best interest of the firm. Culture did the work of control instead of direct authority (Kunda, 1992; Turco, 2016; Vallas, 2006). The discretion given to remaining middle managers in this model is variously described as “situated” or “entrepreneurial” management.² Managers were now expected to complete explicitly defined tasks of control and additional strategic “situated” work, to manage vertically as well as laterally.

By the close of the 20^th century, centralized and hierarchical models of firms lost favor to these flatter organizations. The model of coordinating workforce in this era of bureaucracy-cutting and loss of local primary industry focused on methods of centralizing strategy and decentralizing operations in a balance to “allow for control and creativity” (MacDuffie, 1996). Courses on management at business schools pivoted to courses on leadership instead, reflecting both the rise of the charismatic CEO and new responsibilities distributed across all layers of the workforce. Stories of empowering laborers to “pull the chain” like Japanese automobile manufacturers abound, voice-rights and decision-making responsibilities are distributed across the firm as forms of “taking ownership”, and power is all the while exerted but far less directly manifested in hierarchy (Dowding, 2015; Freeland & Zuckerman Sivan, 2018; McCaffrey, 2023; Turco, 2016; Vertesi, 2020; Zuckerman, 2010). Workers were expected to manage themselves and small adjacent teams, including projects that require lateral coordination among specialists (Girard & Stark, 2003; Kellogg, 2014; Kellogg et al., 2006; Stark & Broeck, 2024). Outsourcing increasingly excised core functions to outsiders (Kunda & Barley, 2006), lessening the importance of a manager’s direct oversight of laborers as necessary or valued. Boundaries were effaced between “inside” and “outside” the corporation, workers and management, work and home. Companies embraced outsourcing not to tighten control over their workforce (as in the industrialist form) but to eject managerial responsibility entirely from their purview. Through these structural reinventions, the firm flattened its hierarchical pyramid. Competencies were extracted for outsourcing, middle management tasks were redistributed, and externalized workforces could be partially and laterally administered outside the boundaries of the firm.

The story of how middle-management is continually being expunged from the 21^st century firm has been told elsewhere in more detail (and debated, see Goldstein, 2012). Our goal in this brief retelling is to demonstrate how we might productively situate the story of automated management on gig firm platforms within this lineage rather than strictly within the industrial form. Such a perspective challenges our typical assumptions of control, resistance, and automated technologies.

For instance, driven by flattening structures and search for profitability, the role of middle management changed from one of worker control to one of situated responsivity. As “entrepreneurs” within their organizations, workers were expected to exhibit “creativity”, to diverge from centrally-determined scripts to include contextually-important information or expert interpretation. They were even encouraged to be “rule-breakers” as their responsiveness to context instead of deference to hierarchy contributed to firm goals (Canales, 2011). Middle managerial work was now about improvising and anticipating organizational needs as much as it was about completing explicitly defined and routine coordination tasks.³ Of course, there has always been an element of negotiation on the job, even in industrial organizational forms, as classic studies of the shopfloor describe (Burawoy, 1979; Stark, 1980). “Situated” management, however, escalates and extends this discretion. All workers are now their own managers. Further, all workers (even lowest-level ones) are acting as good “sociological citizens” (Silbey, 2011) on behalf of the firm’s goals when they “exercise good judgment” (Turco, 2016) and respond flexibly as “situational” managers to local circumstance. This expands hegemonic activity beyond the managerial class, even as it simultaneously limits opportunities for worker empowerment (Vallas, 2006).

Tensions associated with automation that are common to post-bureaucratic organizations also echo through the contemporary gig firm. The cultural emphasis on “high performance organizations” in the 1990s suggested that automating industrial managerial functions — such as coordination, machinery logistics, and worker surveillance — could relieve the firm of perceived bureaucratic bloat and streamline operations while preserving shareholder value. The managerial jobs that remained no longer focused on routine administrative tasks but evolved to look more like technicians who needed to identify organizational needs within their production processes on their own and problem-solve until they addressed the issues they identified (Osterman, 1996). In response to concerns of an overabundance of managerial discretion, operations research attempted to make this management more predictable and routine (Cohen et al., 1996), feeding the production of workplace software to assist in standardizing managerial tasks. As a result, ever fewer middle managers were asked to oversee more teams and tasks than they had previously through the use of extensive monitoring technologies (Meyer, 1968; Osterman, 2006; Prechel, 1994). Workers on the ground, meanwhile, faced increasing layers of technical surveillance. They also faced increased workloads as they took up the leftover coordination tasks formerly handled by this shrunken managerial layer (Lambert, 2015).

Today’s gig economy firms and their workers are the inheritors of this historical context, its structural shifts in the workplace, and associated characteristics of work. As the latest iteration in a long line of corporate restructuring, today’s gig companies do not necessarily implement algorithmic tools to improve managerial oversight and control over their workers. Instead, they implement algorithmic management to excise managerial functions from their purview, removing worker oversight from the responsibility of the firm. Rather than the apotheosis of Taylorist control, then, we see the gig economy firm — with its absent, automated, managerial authority and outsourced, distributed labor-force — as the apotheosis of a multi-decadal story of the relentless purge of perceived managerial bloat from the 21^st century corporation.⁴ This perspective suggests new ways in which capitalism reinvents itself in the digital era (Srnicek, 2017). It also raises new interpretations of otherwise paradoxical experiences gig workers report on the ground.

2.1 The Importance of Being Absent

It is difficult to study an absence. This may be why prior studies of gig work have taken the algorithmic manager at face value, focusing on its controlling characteristics as experienced by workers and its industrialist metrics of productivity. Unlike studies of automation where we might observe job loss or “shadow work” that remains (Lambert, 2015), on a gig platform there never were any middle managers to begin with whose jobs were taken away. They are founded both without workers (only contracted outside the firm) and without managers (now abstracted into computational tools). These absences from the outset make the outcomes of automation and labor displacement difficult to observe without explicitly invoking longstanding historical shifts in the organizational of work and their naturalization in this corporate form. Amid this trajectory, we consider these managing algorithms as a partial, vestigial trace of a missing managerial layer in the organization: a shadow of what was once an internal layer in a vertically integrated firm. The ghost of middle management.

In our view, algorithmic management relies on partial replacement, a term with tripartite implications. First, automated functions replace only one part an integrated production pipeline: in this case, logistical tasks of production control and worker coordination, not situated or entrepreneurial decision-making. After all, there are only a few managerial “problems” for which “solutions” are computationally available (Pinch & Bijker, 1984), and enough managerial functions can be enacted algorithmically to support the appearance that human managers are unnecessary. Second, with managers omitted from the firm and workers externalized, these situated management tasks necessary to get the job done are lost, unaccounted for in the sociotechnical system. This omitted labor is diffused invisibly among workers’ ranks. Perhaps even more starkly than in the taxi industry or among contracting consultants (Camerer et al., 1997; Kunda & Barley, 2006; Occhiuto, 2017), under the gig firm’s combination of flat hierarchies, labor externalization, and automation, gig workers are left alone to pick up the slack by becoming their own, uncompensated middle managers. This is form of productivity capture, as core responsibilities are excised from the firm’s sphere and transferred to the external gig worker’s own unsupervised self-management. Third, partial replacement is a flexible strategy, allowing for both industrial oversight and post-industrial situated entrepreneurialism. Partial replacement thus represents a triple win for the gig firm: the company profits by eliminating costly managers, displacing responsibility for many managerial tasks onto uncompensated workers, while inviting workers to take “ownership” of getting these tasks done.

Partial replacement forms a technocratic arrangement, too, in that it distributes tasks among humans and machines in such a way as to configure and inscribe these roles and responsibilities across the gig firm’s platform (Suchman, 2006). In the social studies of technology we frequently observe tasks that are incompletely automated and require creativity on behalf of workers to get the job done, whether through “ghost work” that picks up the pieces left behind by artificial intelligence, the “frontier of control” or “assembly practices” that require human-to-human interactive negotiation, reconciliation, improvisation, or iteration, (Bailey et al., 2010; Gray & Suri, 2019; Kellogg et al., 2006; Stark, 1980). As designers assign certain code-able and discrete tasks to digital tools and relegate others to humans (Seaver, 2019; Suchman, 2011; Ziewitz, 2016), this presents implications for both the politics and experience of labor on the one hand and the presumed power of machinery on the other (Haraway, 1991; compare to Winner, 1986). Taken together, the co-presence of the industrial elements built into gig work algorithms and the situated managerial functions that are left out of this sociotechnical assemblage present political and practical implications for the gig workplace.

Automation in Ford’s factory drew a boundary between middle management and laborers in a manner that both upheld the ideal industrial organizational form and supported middle management in increasing oversight and control over workers. The post-bureaucratic company, meanwhile, diffuses control among the ranks by encouraging all to become situated managers, automating replaced enough routine functions such that middle management could be further reduced and responsiveness more broadly shared among the lower-paid. Partial automation in the gig economy straddles these two worlds. It automates the barest traces of middle management using mechanisms of quantification and labor logistics reminiscent of industrial oversight. Its coordinated workers must pick up the situational slack, even as the app exerts industrialesque oversight and haphazard control. While the post-bureaucratic orientation effaces boundaries, the app reinforces a stark boundary between the gig firm and the workers who provide its services.

2.2 Gig Drivers as Situated Managers

Understanding the gig firm’s sociotechnical configuration as a heterarchical arrangement explains certain dynamics familiar to studies of gig work. Instead of paradoxes explainable through culture or necessity, they become examples of displaced situated management that are simultaneously subject to industrial values of managerial control. For instance, many characteristics of gig labor have been described as moments of guesswork in the face of algorithmic opacity, local skills and fluencies, forms of organizing or resistance, or techniques of information sharing (Jarrahi et al., 2021; Lee, 2018; Rosenblat & Stark, 2016). Yet if we observe these same activities from the post-bureaucratic perspective, they become evidence of the diffusion of situated management responsibilities across a flat and outsourced workforce. Rather than deviating from scripts, they are acting as “good sociological citizens” to the corporation’s benefit.

We illustrate these competing interpretations with empirical reference to interviews we conducted with forty drivers for Amazon Flex, Uber and Lyft, collected in 2019 and available to researchers as an open source dataset (Enriquez, 2021).⁵ Our goal is not to be exhaustive, especially amid so many detailed extant studies of gig labor. Rather we refer to these interviews to illuminate the dual interpretations that arise once we re-examine worker activities from the post-industrial perspective: that is, embedded within a sociotechnical system that both omits middle managers (a post-bureaucratic structure) and only partially replaces them by favoring industrial tasks and neglecting situated ones.⁶

For instance, in industrial firms where employees work in shifts, a manager is responsible for allocating workers where they are most needed, optimizing local resources to meet the demand of customers. If the Taylorist middle manager was responsible for scheduling blocks or shifts, the post-bureaucratic situated manager is expected to respond to local situational needs flexibly and using their discretion. Gig economy algorithms, in turn, adapt in near-real-time to worker demand and encourage workers to sign on to drive through pricing updates that adapt to consumer demand. Technologists and business strategists see this as a logistics breakthrough, especially as labor has been framed as an expensive problem to solve by reducing “idle time” wages as much as possible. Yet despite the use of computational systems for scheduling gig work, schedule coordination in the post-bureaucratic firm is not a purely automatable task but includes situational knowledge and responses outside the purview of calculation.

When gig workers try to guess how the market will behave to schedule their time accordingly, this is a type of situated management (compare to Camerer et al., 1997). Rosenblat (2018) first identified how drivers spend considerable time trying to guess when there will be enough work and, in the absence of information about potential trips’ profitability, evaluating whether the wear/tear on their cars is worth signing in. Drivers we interviewed also described “trying to watch… gas versus miles versus trip reviews… I like to watch that information so I can track to see where best to put myself, to put my time.” One driver we spoke to used his graduate school skills in Excel sheet management to develop elaborate tracking mechanisms to shape their situational scheduling decisions. Responding to an evolving and dynamic situation mediated through the app, drivers reported to us that they stayed online until they met their paycheck goals, guessing patterns associated with pricing on the weekends versus weeknights, or seeking “blocks” of work by refreshing their phones at random times throughout the week.

Situated managers respond to immediate changes in equipment or on-the-job flexibility using a local calculus of productivity, and we observed these activities among our interviewees as well. Gig companies may develop these sensibilities among workers who innovate their own ways of managing equipment, repairs, and the fiscal and cognitive expenses of their shift work (on cognitive labor see Daminger, 2019). Licensing, gas, insurance, and repairs factor into each workers’ own calculus for optimizing productivity, such that today’s “working out” (Cameron, 2022) now entails managing multiple factors and currencies. More than one driver we interviewed talked through their complex calculus for balancing rental fees versus income, considered over a week of fluctuating rides. Even in tasks considered easily routinized for automation like package delivery, workers implement cognitive and material managerial techniques to complete assigned tasks, such as where to place packages in their cars for optimal grab-and-go. Drivers described to us their creative workarounds like screenshots to record deliveries in places with “no Wifi, no 3G, no 5G, nothing.” We observe these practices as examples of innovative workarounds both in the face of changing circumstance and in the absence of managerial direction. Far from of unjustified or untrustworthy discretion, we argue, this situated decision-making is encouraged and even celebrated as a form of productive creativity and entrepreneurial management in post-bureaucratic firms.

Between ambiguity in scheduling and desire to make ends meet, many drivers we spoke to went to great lengths to improve their access to scheduling blocks or to make good decisions about a pickup. This improvisation also takes place in the gap between industrialized forms of automatable scheduling and situated management necessary in the manager-less firm. Gig workers thus develop just-in-time responses to just-in-time scheduling information to manage their own availability. Responding to “random” release, workers also must factor in considerable amounts of unpaid time waiting for shifts, as one driver explained: “I was spending four or five hours every morning because they drop them at random times, for me to get that block.” Some engaged “Script Cartels” or “Block Extractor” companies to sign up on their behalf, essentially employing their own middle layer of management to handle the costs of coordination. Even in the presence of algorithmic tools that target logistics and scheduling, responsive practices associated with situated management are a routine part of the gig worker’s unpaid workday.

It is well-known that gig workers turn to digital infrastructures to establish peer networks, offer support and learning opportunities, and to organize (Schwartz, 2018; Ticona, 2022; Watkins, 2022; Wood et al., 2019). This connective tissue is also a form of distributed situated managerial task assumed by the post-bureaucratic worker. Under post-bureaucratic companies’ concept of “ownership”, workers are expected to take personal responsibility for their own feedback, mobilize toward opportunities for improvement and collective problem solving. Instead of trusting the app’s generic feedback, “tips”, or and pre-recorded training videos (described in the dataset as “vague… not especially useful”, “obvious”, “a kindergartner could do this job”, or “forgettable”), we observe workers crowdsource problems associated with their own professional development in the absence of managerial interactions or evaluation. Shadow professional development seminars, theories about scheduling scripts, and peer-to-peer tips on how to maximize income were often visible on online driver groups led by individuals who revealed secret strategies for payment. This need was especially keenly felt as drivers noted that emails or “tips” from the company were rarely aimed at performance improvement but were counted as tallies against them instead.

Active online groups also provide a form of mentorship and real-time assistance in problem-solving when something falls apart, again in the absence of management. This includes assisting someone on a delivery route when their map and location tools breaks, offering help responding to termination emails, or providing a direct phone number for a customer service agent known to answer questions beyond a proscribed template. Delivery workers strike up peer-to-peer relationships with in-store workers associated with other components of the platform’s promised system (Maffie, 2024). Brokerage ties like these are familiar to the post-bureaucratic firm, enabling work to get done across lateral groups without direct oversight (Kellogg, 2014; Kellogg et al., 2006). Taken together, peer networks and lateral connections filled in for management or top-down directives, by taking on situated managerial tasks as workers’ own problems to resolve.