Abstract
This article analyzes how unions contest the harmful impacts of automating algorithms and AI on job quality in frontline service workplaces. Integrating two theoretical approaches (labor process and power resources), we examine worker and union experiences with self-service job replacing technologies (self-service applications, chatbots, and interactive voice response systems) in Canadian and US call centers. Building on Smith's concept of double indeterminacy, we suggest that issues with automation lead to two types of bargaining over job quality: work effort bargaining and mobility bargaining. Work effort bargaining concerns contestation over work intensification arising from automation-driven disruptions to service delivery. Mobility bargaining concerns contestation over job insecurity arising from automation-driven labor replacement. We argue that these types of bargaining are not novel, but rather extensions of longstanding conflicts across different indeterminacies in the labor process. Unions can promote better outcomes across both indeterminacies. However, their effectiveness hinges on their ability to build, sustain, and mobilize institutional power resources over time. In doing so, unions face strategic challenges in contesting both work intensification and labor replacement across temporal scales. We conclude by suggesting that labor process and power resources scholarship examine broader scales of resistance to better understand union struggles over technology.
Scholarly reseach on algorithms and other automating technologies shows that their impacts on job quality are far reaching and depend on the extent that they replace, displace, control, and augment tasks and work processes (Berg et al., 2023; Cameron, 2024; Doellgast, 2023; Dunn, 2020; Lei & Kim, 2024; Liu, 2023; Wood, 2021; Wood et al., 2019). However, there is growing recognition that how these technologies are designed and adopted is “socially shaped” (MacKenzie & Wajcman, 1999) and can be resisted (Thompson & Laaser, 2021). Labor unions are important actors in this resistance. Thus, we extend knowledge on how labor unions resist the negative impacts of automating technologies on job quality (Berg et al., 2023; Cornfield, 1987; Doellgast & Wagner, 2022; Dupuis, 2025; Pulignano et al., 2025).
This study draws on labor process theory (Braverman, 1974), complemented by power resources theory (Arnholtz & Refslund, 2024), to explain how union contestation at the point of production mediates the impacts of automating technologies on job quality. Building on Smith's (2006) conceptualization of the “double indeterminacy” in the labor process, we identify two indeterminacies over which unions contest and negotiate the impacts of algorithmic technologies on job quality. The first indeterminacy is work effort, which we connect to worker struggles over the effort—both physical and mental—required to complete work tasks. The second indeterminacy is work mobility, which we connect to worker struggles over maintaining job security with their employer. This contrasts with traditional understandings of work mobility which stress the importance of worker exit threats as a source of bargaining leverage for improving job quality.
We then show that unions play a role in mobilizing power resources to protect workers from automation-driven job quality threats linked to both indeterminacies. By power resources, we focus on institutional power resources derived from language in collective agreements, as well as associational power resources leveraged to negotiate and maintain these agreements. We also recognize that these power resources are often mobilized alongside shifting constellations of other power resources (e.g., structural, ideational).
Our argument is that struggles over the impacts of technology use on job quality must be understood as historical struggles over work intensification and job insecurity in the labor process. These struggles involve the mobilization of power resources, which in itself is a historical process (Korpi, 1983). Given the temporal dimension of these struggles, activist unions that have negotiated stronger bargaining language in times preceding the intensive use of new automating technologies have greater institutional power resources to lean on after its onset. To be effective, unions thereby articulate their actions across intertemporal scales, continuously maintaining associational power resources and then identifying junctures of opportunity where they can be mobilized to build or maintain institutional power resources (Lévesque & Murray, 2010). This is complicated by the politics of technology use related to both indeterminacies, each requiring distinctive interventions to distinctive job quality threats.
The above argument is advanced with evidence on worker experiences with self-service automating technologies in call centers, including self-order applications on cellphones and websites, chatbots and interactive voice response systems. The current study draws on interviews with union staff, workers, and managers (N = 106) in telecommunications call centers in the US and Canada, complemented by matched descriptive survey findings in both countries (US: N = 2,891; Canada: N = 385) and findings from secondary literature.
The following section reviews relevant labor process and power resources literature. Drawing from them, it formulates a framework to explain how unions mediate the effects of automating algorithms and AI on job quality. Next, we provide an overview of our research design. These are followed by a presentation of the findings. Finally, we conclude the paper with a discussion of these findings and their implications for further research.
Literature Review
Linking Double Indeterminacy in the Labor Process to Worker Contestation Over Job Quality
This article draws on labor process theory, complemented by power resources theory, to provide insights on how unions mediate the impacts of automating technologies on job quality. We begin by unpacking how automating technologies impact job quality across different aspects of the labor process. Inspired by Marx (1867/1976), a central premise of Braverman's (1974) work was that capitalism affords management the capacity to purchase workers’ labor power and not labor itself. This labor power is a commodity unlike any other as its productive value depends on management's ability to extract value by exerting control over the labor process. The legacy of Braverman's work is that it placed the labor process front and center as an “object of class politics” (Thompson & Laaser, 2021, p. 143), wrapped in his original theorization of how managerial control and technology degrade work over time. Although his work left gaps in understanding the dynamics of labor control and resistance, it prompted deeper analysis of these aspects of the labor process in later waves of theory and debate (Thompson & Newsome, 2004).
We build on these later waves of labor process theory which interpret technological change through the lens of a “duality” between control and resistance at the point of production (Noble, 1978; Wilkinson, 1983). Labor process scholars suggest that technologies are essential components of modern “workplace regimes”, as they embed stringent mechanisms of worker control which are then contested by labor actors at the point of production (Burawoy, 1979; Dupuis, 2025; Steinhoff, 2024; Wood, 2025). In the context of digitalization, new configurations of “technical control” (Edwards, 1979)–enabled by algorithms and AI technologies–are reshaping structured antagonisms in the workplace (Edwards, 1986; Edwards & Hodder, 2022; Kellogg et al., 2020). In response, workers and unions can and do resist management's use of these technologies to intensify or control work effort (Cameron & Rahman, 2022; Kellogg et al., 2020; Krzywdzinski et al., 2024). In addition, these actors also resist the displacing effects of automating algorithms and AI on work (Cornfield, 1987; De Stefano, 2019; Doellgast et al., 2026b). We contend that both types of resistance must be examined together, and that doing so facilitates a better understanding of union struggles for better working conditions under digitalization.
Thus, our analysis focuses on how structured antagonisms in automation play out across the two indeterminacies (Smith, 2006) in the labor process. The term indeterminacy reflects that capitalists–as purchasers of labor power–continuously engage in efforts to minimize uncertainties in how much work can be extracted from workers. The first indeterminacy relates to work effort. It reflects the reality that workers do not necessarily exert their full capacity to work. Rather, capitalists employ mechanisms of control to ensure that workers scale up their efforts in production. Work-effort bargaining thereby involves labor-management negotiations over the amount of work effort that will be realized, as well as any necessary exchanges to achieve this realization (e.g., higher wages for greater effort).
The second indeterminacy pertains to work mobility. According to Smith (2006), workers have mobility power to move between jobs and use it to threaten employers with exit from their position and to bargain for improvements in working conditions. In this sense, exit is not merely opportunistic behavior but a form of resistance and expression of conflict between workers and management (Alberti & Sacchetto, 2024). The notion of mobility power hinges on a similar logic to that of research on “structural power” (Greer, 2024; Wright, 2000), which is that some workers can leverage power from market demand for their skills and experience while others cannot.
While covering neglected aspects of the labor process, Smith's conceptualization of mobility power merits an extension. First, it provides a very narrow and one-sided understanding of this form of power. In contrast to the power of workers’ exit threats, employers possess a symmetrical ability to push workers out of the organization through terminations, layoffs, and by making the work climate undesirable (Gallie et al., 2017). As a result, workers’ mobility power is not merely a function of their ability to exit the organization, but also employers’ ability to discharge them from it.
Second, his conceptualization facilitates a narrow understanding of worker agency in shaping outcomes. Prior studies treat the contestation of mobility power between workers and employers as individual and informal (Alberti & Sacchetto, 2024; Gerber & Krzywdzinkski, 2019; Heiland, 2022); broadening our conceptualization of worker mobility allows us to capture how unions influence this indeterminacy. Workers can collectively resist harmful staffing practices within their organizations (Hodson, 1995). For example, a study by Sexton (1982) captured the role of “feather-bedding” provisions within collective agreements in restricting hospitals from understaffing nurses. In settings where automating technologies simplify, routinize, or eliminate work tasks and systems, hence rendering workers more expendable (Herr, 2024), the issue is not merely how technologies impact workers’ threat of exit. On the contrary, it also concerns creating rules to reduce uncertainty over whether workers remain employed in the organization. Theorizing mobility power in this way conceives it as a contested indeterminacy subject to the influence of workers’ individual and collective actions.
The above theorization has significant implications for understanding how automation impacts job quality. First, tensions surrounding the impacts of automating algorithms and AI on work effort are effectively tensions over the intensification of work (i.e., pace, volume, and complexity of tasks) (Green, 2006). Automating technologies may eliminate some jobs. However, they often leave higher volumes of intensive work tasks for those whose jobs have not been eliminated (Antón et al., 2023; Berg et al., 2023). This is especially problematic in many modern workplaces where stringent algorithmic control over work processes is common and workers must cope with performance metrics that are both draconian and unrealistic (O’Brady & Doellgast, 2021; Woodcock, 2016). Work intensification is often the subject of work effort bargaining, as workers often resist machine-driven increases in workflow and pressures from task complexification (Krzywdzinski et al., 2024; Lei & Kim, 2024; Wood, 2021). Second, tensions over the impacts of automating technologies on work mobility are effectively tensions over job security. The displacing effects of automating algorithms and AI are well documented (Acemoglu & Restrepo, 2019; Baldwin, 2019; Ford, 2015; Rifkin, 1995; Wajcman, 2017). These threats are particularly acute for vulnerable workers, such as those in outsourced or precarious positions (Vertesi et al., 2020). Although automation-driven job insecurity significantly erodes job quality (Liu, 2023; Yam et al., 2023), it is not uncontestable. Recent research suggests that unions can mitigate the negative effects of job insecurity by strengthening employment protections (Haapanala et al., 2023; Nissim & Simon, 2021).
These tensions have broader implications for job quality, since work intensification and job insecurity are often interrelated with other job quality problems like poor job satisfaction or stress and emotional exhaustion (e.g., Dengler & Gundert, 2021; Gallie et al., 2017; Glavin et al., 2011; Heaney et al., 1994; Liu, 2023; McClure, 2018; Nazareno & Schiff, 2021). Thus, the above insights tie labor process debates on technology to current discussions on the quality of work, by showing how labor politics across two indeterminacies in production matter to different job quality outcomes.
Power Resources and the Temporal Dimension
The previous section established that automating technologies can disrupt job quality across two indeterminacies in the labor process. The current section highlights the role of union power in contesting bad outcomes within either indeterminacy. Power resources are those that “social actors” can mobilize to “make, receive, or resist change in accordance with their own interests” (Arnholtz & Refslund, 2024). A key contention of this theoretical approach is that unions have multiple power resources at their disposal and that their influence over working life is scaled up when they can strategically mobilize them in combination with one another (Doellgast et al., 2018; Meardi, 2024; Schmalz et al., 2018). Rarely are single power resources (e.g., associational, institutional) alone sufficient for promoting better social outcomes (Doellgast et al., 2018), including in instances of technological change (Doellgast et al., 2026a). Here we highlight the importance of how power resources are mobilized to shape outcomes across different indeterminacies in the labor process, which create challenges for the effective regulation of digital technology use and job quality.
Research corroborates the relevance of union power to outcomes from technological change across each indeterminacy. First, emerging research suggests that unions or works councils can mobilize power resources to protect workers from work intensification driven by technological change (Doellgast et al., 2023; Doellgast et al., 2026a; Lloyd & Payne, 2023; Pulignano et al., 2025; Rego, 2022). Institutional power resources often play a particularly important role. For example, Doellgast et al. (2023) show that when workers possess strong codetermination rights, works councils can proactively intervene in AI-driven technological change and constrain the use of digital tools for intensified worker monitoring. They often leverage these institutions alongside discursive power resources in which the benefits of ethical AI use for worker trust, company reputations, and organizational performance are strategically communicated to employers (Doellgast et al., 2026a). Dupuis' (2025) study of aluminum factories in Canada demonstrates that collective bargaining provisions negotiated in earlier rounds can be mobilized to preserve frontline workers’ autonomy in response to algorithmic decision-making systems. Where institutional power is weaker, such as in contexts with weak collective bargaining language, research points to the importance of associational and ideational resources. In such a context, Riordan et al.'s (2025) study of the hotel industry highlights how union leaders draw on associational and ideational power resources to mobilize member support and informally negotiate work rules with managers to address shortfalls in previously codified rights and agreements.
Second, researchers also suggest that unions can help protect workers from technology-driven job loss, although doing so involves negotiating over distinct sets of challenges in the workplace (Doellgast et al., 2026a; Pulignano et al., 2025). In their analysis of Hollywood Writers’ strikes, Grohmann et al. (2025) showed how discursive power involving the use of slogans, media outreach, and appealing to fans helped writers combat the threats of generative AI on jobs in the industry, ultimately contributing to the establishment of new bargaining provisions (institutional power) that restrict AI use. Thomas and Turnbull's (2025) research on the attempted creation of a digitalized remote tower center by Highlands and Islands Airports Limited (HIAL) in Scottland demonstrated how unions can mobilize various power resources (structural, associational, institutional, coalitional and ideational) to challenge the adoption of remote digital job killing infrastructure. Garneau et al. (2023) show that unions often find opportunities to “frame”, and hence mobilize discursive power, in hopes of building power in resisting automation. However, they recognize that unions in decentralized bargaining contexts face constraints compared to those with access to stronger centralized bargaining institutions. Thus, a dominant finding from literature on the impacts of union power within either indeterminacy stresses the importance of mobilizing combinations of power resources and tailoring mobilization efforts to the unique issues and contexts of the site of production.
This study departs from previous research in two ways. First, we provide insights on how the details of the labor process impact worker struggles for better work under digitalization. Subtle differences in the labor process often have important implications for how workers resist the impacts of technological change. Emerging scholarship has begun to capture how labor control regimes, which are shaped by power relations and industry contexts, matter to job quality (Cornfield, 1987; Heiland, 2022; Pulignano et al., 2025). Pulignano et al. (2025) show that the distinctiveness of digital “production regimes”, alongside union capacities to leverage power resources to confront issues within those regimes, can lead to variegated outcomes for worker control and the quality of work. Certain production regimes and combinations of power resources can impact whether unions pursue narrow agendas to enhance job security versus broader goals of worker wellbeing.
We build on these insights but suggest that greater analytical attention be paid to the different indeterminacies within a given production regime. As shown earlier, digital technologies reshape control in distinctive ways across these indeterminacies, generating different risks to job quality and provoking different forms of union intervention. A closer examination of how power resources are mobilized across both indeterminacies is needed to understand the full breadth of union impacts on technology use and their consequences for workers.
Second, we build on insights from those who call for a “temporal perspective” in analyzing struggles over technology. Power resources can become crystallized into institutional forms—such as laws, technologies, or collective bargaining arrangements—through past mobilization (Doellgast, 2018; Korpi, 1983; Schmalz et al., 2018). A small number of studies suggest that the temporal dimension is critical to understanding how unions respond to digitalization. They emphasize the historical nature of these struggles (Grohmann et al., 2025), and how the strategic mobilization of power resources at one point in time can impact efforts to shape technology adoption in another (Thomas & Turnbull, 2025). We build on these insights by highlighting the importance of historical struggles across different facets of the labor process, showing how struggles that build worker power and control over more facets at one point in time can empower workers in another.
Theoretical Framework
Drawing from the theorization above, we contend that: 1) unions must engage in both work effort and mobility bargaining to effectively resist the harms of automating technologies on job quality; and 2) this resistance entails the strategic mobilization of power resources across different indeterminacies and time periods.
This leads us to our framework in Figure 1. The starting point of our framework is automating technologies. A technology automates by using “machines and computers to substitute for human labor in a widening range of tasks and industrial processes” (Acemoglu & Restrepo, 2018). Many technologies have automating properties, ranging from algorithms which contain “step-by-step, distributed and nominally objective procedures” (Fourcade & Healy, 2017) to AI used to simulate human intelligence. We then suggest that automating technologies can impact both work effort and work mobility, which in turn have consequences for work intensification and job insecurity respectively. When technologies automate certain tasks, this can increase work effort and intensification for those whose jobs have not been eliminated, as they must complete complex tasks that could not be automated (Nazareno & Schiff, 2021). Furthermore, newly implemented technologies are often imperfect and create errors that must be fixed by workers.
Framework.
Automating technologies also alter work mobility and hence increase job insecurity through the elimination of work. Whereas traditional research on worker mobility focused on worker capacities to exit the organization (Li, 2022; Smith, 2006), we focus on their ability to remain within the organization. That is, automating technologies can lead to job loss when they serve as effective substitutes for the work done by existing workers. However, not all managers have full discretion to discharge workers that have been displaced by technology, due to restrictions imposed by collective bargaining and legislation.
We propose that each indeterminacy is subject to negotiation and contestation by unions. Union actions serve to curtail the negative effects of automation on work effort when they constrain management's ability to make work faster and more difficult. In this study, we focus on collective agreement provisions that constrain management's use of discipline to enforce performance targets and expectations. Union actions serve to curtail the negative effects of automation on work mobility when they constrain management's ability to discharge workers. Thus, we focus on collective agreements that do this. For our purposes, union actions to negotiate and leverage collective agreement provisions that provide workers with greater control over work effort represent work effort bargaining. Union actions to negotiate and leverage collective agreement provisions that provide workers with greater control over employment within their organization represent work mobility bargaining.
Union actions in both work effort and mobility bargaining draw on power resources. Here we suggest that there is a temporal dimension to these struggles. We examine institutional power resources in the form of codified collective agreements that reflect compromises from the past which set rules that still have bearing in the present (O’Brady, 2024; Schmalz et al., 2018). We also examine associational power, which reflects the “organisational properties that enable and express unity of action among workers to make, receive or resist change in accordance with their interests” (Ibsen, 2024). These properties include internal and mobilizing infrastructures, as well as information sharing practices. Associational power both reinforces and is reinforced by institutional power (Doellgast et al., 2018), as it serves to build institutional power resources and ensure they are effectively mobilized when needed.
Thus, we argue that work effort and mobility bargaining reflect the intertemporal use of complementary power resources. When unions use associational power resources to establish rules governing the labor process, these rules can later function as institutional power resources when new technologies disrupt work. When unions use associational power to enforce rules, they strengthen them by showing that violations carry sanctions. This dynamic points to the importance of union “articulation” over time (Levesque and Murray, 2010). Institutional power is not static; it is shaped by strategic decisions over time and ongoing efforts in building and mobilizing associational power resources. Thus, it is the accumulated strategic mobilization of power resources that helps to build institutions that govern the labor process. These decisions respond to opportunities to both build and defend institutional power and are often impacted by shifting configurations in other types of power.
Research Design
Our study draws on sequential waves of mixed-methods data and compares worker and union experiences across call centers in Canada and the United States. The focus was on call centers within telecommunications, due to high levels of union presence and the heavy adoption of digital technologies in the industry. This sequential mixed-methods approach enabled us to analyze union strategies and worker outcomes across diverse time periods and data sources. Through it, we were able to use later rounds of data collection to validate findings from earlier rounds, and to triangulate by incorporating different types of data into our analysis. The comparative component of the project enabled us to compare outcomes across and within unions in two North American countries that share similar socio-cultural realities and Wagner Act models of industrial relations. Controlling for national and industry contexts helps us to identify and explain key trends in the digitalization of the labor process as well as union responses to those trends.
We examined worker experiences with three sets of technologies. First, we examined online self-service applications. These applications are algorithmically powered digital systems (e.g., mobile or web apps) which enable customers to obtain services or fulfill transactions without the assistance of a worker. Second, we examined chatbots. Chatbots rely on software that uses natural language processing and AI to simulate real time live messaging over some form of text interface. Third, and finally, we examined automated phone systems which use voice commands or keypad entries to fulfill customer queries and orders. Developments in speech analytics are rapidly increasing the sophistication of these technologies and enabling voice bots to tailor their responses to the nuances of the conversation.
This data was gathered in waves. A first wave of qualitative semi-structured interviews and workshops were conducted with 106 participants (managers, workers, worker representatives, and technology developers) in 2021 and 2022 (Canada: 46; US 60). Subsequently, a survey was disseminated to call center workers in December 2022 and March 2023. The survey received 3,277 responses (US: 2,892; Canada: 385). Our response rates were 14% and 29% in Canada and the US respectively. To ensure generalizability, we compared the demographic characteristics of our sample with those of similar professions using data from the Bureau of Labor Statistics and Statistics Canada and found them generally similar.
This survey was launched in collaboration with union partners in both countries and hence aimed to capture the experiences of their members with various automating technologies, as well as their perceptions of their union's effectiveness in addressing issues relating to the potential abusive use of these technologies. The survey design was informed by the first wave of qualitative interviews.
After administering the survey, we proceeded with a second wave of qualitative data collection in 2023. This wave involved seven workshops with worker representatives in call centers at major US employers. Across both waves of qualitative data collection, interviews and focus groups were conducted remotely (lasting an average of 64 min) and conformed to ethical protocols for anonymity and secure data storage. Interviewees spoke on the experiences of nine companies (Canada:3; US:6). We also analyzed collective agreements from thirty-four bargaining units affiliated with these companies (Canada:9; US:25).
Our union partners aided in the interview and workshop recruitment process by soliciting participation from union staff and workers through direct email communications with union locals and by introducing the researchers to prospective participants in union workshops and meetings. Thus, our sampling approach for the interviews was purposive. The goal was to recruit participants with deep knowledge of how automating technologies were impacting working conditions and the strategies unions adopted to regulate their use. In applying this approach, we also worked with union staff to recruit workers across a range of positions, geographical locations, demographic characteristics, and backgrounds with the union.
We adopted a convenience sampling process for the surveys. Union staff circulated the surveys to workers in their email listservs or union newsletters. These were followed by 2–3 reminders over the next four weeks. Finally, these unions provided support in collecting collective agreements and related documentation.
The strategy for analysing the qualitative data was iterative and involved various methods of validation. Initial steps involved reading transcripts and notetaking. Next, we pursued first-level (informed by case summaries) and second-level coding of the data using NVIVO. A half-dozen researchers took turns coding the data, to ensure validity and further refinement of the codes. We also consulted with union practitioners to ensure consistency between our interpretation of the interviews and the experienced reality of union staff and workers on the frontline. Finally, clustered summary tables were prepared to inform our analysis.
Findings
The Contested Impacts of Automation on Work Effort and Intensification
This section examines the contested ways that automating self-service technologies impacted work effort and hence intensification. Interviewees described how the implementation of self-service forms on websites, IVR systems, and chatbots intensified work tasks. This intensification was driven by errors made by AI and the use of protocols to encourage customer use of self-service (script use and deliberately prolonged wait times), both of which created barriers to efficient call resolution and heightened fears that workers may be disciplined for not meeting performance goals. Responding to these problems, worker representatives mobilized collective agreements as institutional power resources, enforced with associational power resources, to protect workers from being disciplined for not meeting performance goals.
Interviewees commonly reported increased work effort and intensity from automation-related service disruptions. They provide convenience for customers by providing them with potentially efficient service interactions but led to a rise in complaints about technology-related errors. One set of complaints were that AI chatbots and IVR systems made errors in interpreting customer queries. One worker criticized their employers’ chatbots: “the chatbot is very limited in the questions it can answer. Often after two or three trials it will allow the customer to connect to agent” (Worker, C Telecom 1, March 2021). These interpretation problems were a major source of customer frustration. One manager described how these delays and frustrations escalated worker-customer conflict before the call even began, leaving the customer “ticked off” and creating further work for agents to de-escalate (Manager, US Telecom 1, December 2021).
Call routing technologies also provoked more negative customer relations when chat-bots or the IVR connect customers to the wrong departments. Often, “folks are getting sales calls that shouldn't be getting them, or the sales folks that are getting calls that are technical assistance, that should be getting sales calls” (Worker, US Telecom 2, April 2021). Thus, workers are made to take on a more challenging role in correcting these errors and managing resulting customer frustrations. Under circumstances where service quality is frequently linked to workers’ compensation and disciplinary measures, it is ultimately the workers who bear the consequences of these disruptions. One worker described how customers would complain when the automated system did not understand their reason for calling. Often, these systems would begin “resetting all their equipment”, causing service disruptions for customers who had no issues with the equipment in the first place (Worker and Union Representative, US Telecom 5, May 2021). Similarly: A lot of our members that have been trained with the job know the support group that they need to go to, but they're stopped and blocked by a chat bot before they can even get there. And then depending on the words that they use, the chat bot determines if they should continue to a live representative or not. So that's going to cause delays on the call, delays on the work. It's going to slow productivity down in a lot of scenarios because in some of these work groups, if you're needing to chat with back office for fiber, if you're needing to chat with repair for copper, or if you're needing to chat on the mobility side, if you don't say the word that the chat bot [has] built an algorithm for, you're going to be sent to the wrong department, you're going to have to start over again and your customer is now waiting. (Worker, US Telecom 1, March 2023)
While errors from automating technologies extended wait times, workers also accused their employers of deliberately extending customer wait times to encourage self-service. One worker representative compared it to having “10 lanes of self-help” and “one person checking groceries out” with a live cashier in a grocery store (Union Representative, US Telecom 3, June 2022). Customers often received automated messages encouraging them to use self-service systems while on hold for agents. Additionally, employers have been reducing the contact points for customers to reach agents. One worker representative complained that customers “are going to be forced to stop writing to me because they are abolishing our email addresses” (Union Representative, C Telecom 1, March 2021). Others indicated how contact information for customer service is decreasingly available on company websites. Customers were often searching for agents, but “you just keep getting the [chatbot] and you're losing your mind” (Union Representative, US Telecom 5, July 2022). Thus, workers and worker representatives often reported complaints of customer frustrations over not reaching live agents.
Alongside customer driven frustrations over wait times, workers complained of new scripting protocols that require them to encourage customers’ use of self-service technologies. Workers that did not encourage self-service could be penalized. As one worker illustrated: “When they listened to the calls they got, they got scored, that they explained to the customer that self-service was available. I remember fighting it because I thought that's ridiculous. But anyway. And if you didn't mention it, then you got you got, you know, docked on your matrix.” (Worker, C Telecom 1, March 2021)
A large number of qualitative comments in the survey corroborated the above findings. Of the 414 comments that were collected, there 91 negative comments on automating technologies (out of a total of 130 comments on automating technologies in general). 62 (68%) of them referred to increased work intensity caused by errors in self-service technologies. One worker stated: “call trees and chat bots have made my job harder & more stressful—often misroutes piss customers off, so it is harder to talk to them and sell services. Internal chat bots waste my time and often don’t help, leaving me without assistance or info [sic].” Moreover, 14 (15%) comments emphasized that these errors were particularly demanding because rigid management practices restricted workers’ autonomy to address them and shifted the risks of errors onto employes. One worker noted that “either misrouted or routed calls make it harder to make numbers or goals.” Another mentioned the high chance of receiving low feedback from angry customers: “because you cannot give them what they want, and they are angry with [the company], but take it out on you, the agent.” Taken together, the central issue for workers was that these technologies disrupted service delivery, and that these disruptions affected how they performed on metrics and could even lead to discipline.
Union locals varied in the strength of their contract language and the strength of associational power resources available to enforce that language. First, some union locals have bargaining language which required that they be consulted on the introduction of new digital technologies. Half of the studied bargaining units have collective agreement language requiring that managers provide them with notice (ranging from 4 weeks to six months) that major technological changes are forthcoming. Three of the agreements go further by requiring employers to consult with unions on how these technologies will be implemented. In C Telecom 1: “The company agrees to provide as long as possible notice of at least four weeks to the unions, before the introduction and implementation of this technological change. They are obliged to notify me a month in advance to tell me [the worker representative], it is coming, such automation. Or this tool will come and then, consequences, the impacts that it will have in the team.” (Union Representative, C Telecom 1, March 2021) “You know, they go on hunts. Sometimes, if they get a bad customer survey that says you did something wrong, they cannot individually pick that call to go on a witch hunt. … So, the language in there allows it to be random, individually selected. And then, you know, also on it, if there's certain things they discover, it's automatically a coaching not a discipline.” (Union Representative, US Telecom 1, May 2022)
Many union representatives acknowledged that service quality and employee outcomes could be significantly improved if automating technologies were made more effective. Thus, some union representatives saw their role as advocating for the effective implementation of such technologies. They would routinely gather information on technology errors and communicate them to management. Speaking of a technology that automates problem resolution by following algorithmic decisions but reduces service quality by restricting worker discretion, one worker representative stated: “Ok, we understand you have this tool to make things more efficient, but you need to have a way for employees and for the union to appeal when it's not right. So, if it's telling us to do something wrong, you need to have a mechanism in place to say, ‘a process that you're telling us to do is wrong’. So, we fought to get that done, but it took a lot of fights, I think, before they backed down for sure.” (Union Representative, C Telecom 1, March 2021)
In sum, the implementation of automating technologies intensified work by creating errors that required fixing and through new work protocols that encourage self-service. They intensified work effort because: 1) agents now had to de-escalate conflicts with customers frustrated by delays in service delivery; 2) they automated simple service interactions, leaving agents with higher shares of complex service interactions; and 3) agents had new obligations to follow scripts on reminding customers of self-service tools. As a result, workers were burdened by additional tasks associated with these new processes, as well as greater work demands involved in handling customers frustrated by service delivery disruptions. None of the unions had the power to stop the implementation of these technologies. However, they varied in their ability to leverage associational power resources to enforce institutional power resources (collective agreements) that protect workers from harmful discipline. This included provisions that protect workers from discipline based on monitored calls, lessening fears of punishment for mistakes made during periods of heightened customer demands. The mere presence of these provisions was insufficient; they needed to be enforced with associational power resources.
The Contested Impacts of Automation on Work Mobility and Job Insecurity
This section examines the contested ways that automating self-service technologies impacted worker mobility and hence job security. Interviewees described how automating technologies have threatened the number of jobs within call centers by facilitating customer self-service, routinizing work, eliminating tasks, and making it easier to outsource work. However, this was subject to contestation, particularly by leveraging collective bargaining language as institutional power resources, and leveraging associational power resources to enforce the language effectively. Worker representatives in locals who successfully negotiated strong outsourcing restrictions, layoff provisions, and bumping rights in the past could leverage this language to protect workers’ job security in the present.
Self-service options were the most cited causes of automation-driven job losses in the call centers. In the initial phases, company websites and cellphone applications were used to perform simple functions such as fulfilling basic orders. From the early 2010s till today, these applications have expanded to provide a wide range of services to fill orders as well as answer customer queries. According to one worker, if a customer wants to add a “a programming change on their TV, if they want to add a long-distance plan, disconnect, connect anything like that, basically, they can do an entirely self-serve, order. And they also have like, you can submit complaints through web forms” (Worker, C Telecom 1, March 2021). In describing the impacts of these technologies on job loss, another worker described how there is: “No need to contact us directly, [customers] can go on a website, they can fill in a request for self-serve or depending on what they need. And this has really downsized a lot of our work… Most of it is webpage. So that's had a dramatic, you know, impact on our membership.” (Worker and Union Representative, C Telecom 1, February 2021) So previous to I don't know, six months ago, maybe when a customer called in, [they] actually asked you to ask for their ID over the phone. Now our IVR asks those questions before the customer comes through to the call center. So they're supposed to be fully verified before they get through. Um, and I know there is meetings and whatnot about other bots that will take over like, okay the kind of the way I was explained to it from a technical support side of it: If say a customer is calling in for troubleshooting of, like Mac's equipment, or TV service not working properly. The first step might be like auto, like rebooting or running, running tests through the system to see like line tests and stuff like that, that in the future may be performed by a bot. (Worker and Union Representative, C Telecom 2, May 2022) “When I started with the company in 2001, they had a department that literally was like the IVR. So, the automated system and if they could answer your question, within 30 seconds or less, they would; if they could not, they would transfer you wherever you needed to go. Now in 2003, maybe 2004, in Idaho Falls, we had, we were just shy of 800 employees and had three buildings. They have not back filled because of how much a computer can actually do. That has now gone down to, like I said about maybe about 100.” (Worker, US Telecom 5, May 2021)
Overall, the broad set of self-service technologies described above are reducing demand for call center workers. A key path for job loss was attrition by means of retirements and “voluntary severance packages”. This has been described as using “stealth” to shrink the unionized workforce without the need for obvious and dramatic mass layoffs (Worker and Union Representative, C Telecom 2, February 2021). In other instances, it has caused entire call centers to close, leaving workers without jobs. Workers in some of the studied call centers were offered to relocate to retain their jobs, but such a relocation comes with personal costs, especially those with families or other mobility restrictions. Automation's threats to job security occur alongside similar threats from outsourcing, as uses of outsourced vendors has increased over time in most call centers.
The negative impact of these technologies on job loss depended on union capacities to leverage institutional power to limit automation-related job losses. First, unions leveraged formal bargaining language to provide strong protective measures against potential layoffs due to automation. For example, C Telecom 3 had a strong agreement prohibiting layoffs resulting from technological change, stipulating that “no regular employees who attain two years of regular service will lose their employment as a result of technological change.” Although C Telecom 3 was the only workplace with explicit no-layoff language directly connected to technological change, most of the studied workplaces were protected by clauses shielding workers from layoffs caused by outsourcing.
Collective agreement provisions that prohibit employers from laying off workers while outsourcing also protected workers from automation-related layoffs. Every employer in the study engaged in some level of outsourcing. Thus, employers could not justify any layoffs so long as they continued to outsource work. This includes layoffs resulting from automation-related efficiencies. Any such layoff could be subjected to a grievance, as employers would need to stop outsourcing before they could lay off workers for other reasons. According to worker at US Telecom 2, “the lack of investment in infrastructure has, ironically, worked in favor of employees regarding automation, as the contract language protects them from contractors” (Worker and Union Representative, US Telecom 2, March 2021). Similarly, a union president at C Telecom 1 described how “you can't come in saying ‘I've automated, so I'll cut.’ You don't have the right…there have been no job cuts lately” (Union Representative, C Telecom 1, March 2021). However, at US Telecom 5, which lacked language protecting workers from layoffs, the situation was notably different: “We've had a huge decline. We've really been impacted by layoffs. It's been really interesting because our members, for some reason, have deflected the blame to the union rather than the company. Because we haven't been able to, you know, stop automation or, you know, save their jobs. Again. So that's been really difficult to deal with. We're still trying to figure out how to handle that.” (Union Representative, US Telecom 5, May 2021)
The presence of the above contract clauses was insufficient for maximizing their usefulness as power resources. They were best leveraged where local union staff were proactive in identifying and policing violations in the collective agreement. For example, several US locals identified how employers were implementing new technologies without providing union leaders with “notice”. They acknowledged the importance of retaining strong lines of communication with workers to identify violations, while ensuring that stewards were well trained to handle grievances. Overall, the goal of many locals was to demonstrate their organizational strength to management, leaving them fearful of union reprisals should they violate negotiated agreements.
Our survey results support these findings by showing that job protection provisions reduce workers’ perceptions of job insecurity associated with self-service automation. In particular, we examined watermark provisions, which require employers to maintain a minimum ratio or number of calls within bargaining units. For example, in US Telecom 3-Unit A, one provision specifies that call centers in the bargaining unit must “handle at least 83% of all calls originating in the [bargaining unit]'s footprint.” Workers without watermark provisions consistently reported higher perceived layoff risks from new technologies. For chat-bots, 68.2% of workers without provisions considered layoffs somewhat or very likely, compared to 58.5% among those with provisions. A similar pattern appeared for automation overall: 65.4% of workers without provisions considered layoffs somewhat or very likely, compared to 53.8% of those with provisions. For self-service applications, concerns were high in both groups, but again the non-provision group reported a higher level of perceived risk, with 72.3% rating layoffs somewhat or very likely compared to 68.2% in the provision group. These results suggest that even though this provision is designed to prevent outsourcing from leading to job cuts, it also helps to preserve call volumes in contexts where self-service automation takes over calls.
In sum, the implementation of self-service technologies was associated with greater fears of job loss as they reduced demand for live agents. However, unions diverged in the strength of job protection provisions that restrict layoffs and provide bumping rights. These provisions were a source of institutional power and helped union officials constrain employee layoffs from automation. Requirements to provide notice before the adoption of new technologies also provided unions with a window to negotiate their effects on job security with management. However, as illustrated above, these agreements were best enforced by union locals that developed associational power resources (i.e., membership communications and grievance training) needed to effectively enforce contract language.
The Intertemporal Dimension of Power Resource Mobilization
The two previous sections illustrated that unions were most effective when they could leverage institutional power resources (i.e., collective agreements) to constrain management discretion over the outcomes of automation. Bargaining language was often mobilized on two fronts: first, to protect workers from discipline in the context of automation-driven service disruptions and demands; and second, to protect workers from job security threats caused by automation. Unions that could rely on strong language to confront harmful practices intersecting with both indeterminacies in the labor process experienced the most success. However, the strongest collective agreement provisions were negotiated and implemented prior to the adoption of these technologies. This aspect of the findings suggests that there is a temporal dimension to these struggles, as unions drew on institutional power resources built in the past to confront technological change in the present.
The relevant collective agreement provisions were negotiated through collective bargaining dynamics that took place many years and often decades preceding our waves of data collection. Most of the union representatives that participated in this study had little historical knowledge of the origins and evolution of the provisions discussed above. This is further complicated by the fact that most of the bargaining units in this study were impacted by mergers and acquisitions that often came with significant disruptions to collective bargaining and hence rapid changes in demands and expectations for concessions from employers (Doellgast, 2017).
Thus, we draw on secondary literature to shed light on this aspect of the argument. Past studies suggest that telecom call center unions leveraged associational power to wage strikes or pressure employers to accept concessions in collective bargaining during the 1980s to early 2000s (Brophy, 2009; Doellgast, 2017; Goldman, 2024; Katz et al., 2003). To illustrate the historical origins of one job security provision, one bargaining unit under a telecom company that merged into C-Telecom 1 in Canada waged a five-month strike in 2004. The strike sought to pressure C-Telecom 1 into accepting “contracting-out language” that would restrict its employer from outsourcing and offshoring work. Through the strike, the union won watermark provisions that placed limits on the volume of outsourcing, as well as notice provisions and other related employment protections. One union representative expressed gratitude that this provision could “carry over” to C-Telecom 1 after the merger (Union Representative, C Telecom 1, March 2021
Our review of eight peer reviewed publications enabled us to account for the origins of these provisions in eight collective agreements. In each instance, the story suggests that more militant rounds of bargaining or strike activity were necessary to win and retain strong language. Union successes were attributed to successful membership engagement, organizing efforts, strategic public relations campaigns, and strike action. Each case involved mobilizing some component of “associational power”, as success relied on union density and member engagement. Structural power resources also mattered. Many provisions were negotiated before market liberalization, digitalization, and the deregulation of telecom made it easier to offshore work (Batt et al., 2009). Being more difficult to replace, workers then had greater power to disrupt customer service. However, making use of this structural power was contingent on union efforts to maximize its potency by leveraging associational power resources to pressure employers. Thus, bargaining units that did not build or mobilize associational power were not able to maximize potential gains in a period of higher relative structural power.
Bargaining units with the strongest provisions had to repeatedly mobilize associational power over time to resist concession-making pressure. One union representative with US Telecom 2 described how they are proud of their “no layoff” and “minimum headcount” provisions in the collective agreement, but insisted that they are “difficult to hold on to” and that they often have to counter pressures from management to abandon them in collective bargaining (Union Representative, US Telecom 2, March 2021). A New Jersey local with US Telecom 4 described how they are engaging and mobilizing their members to resist concession-making pressures, as their employer is trying to weaken contracts to weaken the confidence of membership in their union (Union Representative, US Telecom 4, March 2021). Other locals in Canada and the US have complained that there is too much ambiguity in their collective agreement language requiring employers to provide advanced notice before introducing new algorithm-based tools. A major concern is that this language imposes notification requirements but does not mandate consultation. Thus, unions can be informed but not necessarily consulted.
Notably, we did not identify any clear company or country effects. Of the 34 collective agreements that we analyzed (Canada: 9; US: 25), only four of them have very strong protections across both indeterminacies in the labor process. Each of these four agreements is with a different company and each of these companies have other bargaining units that have negotiated weaker terms. Moreover, only one of the very strong agreements is with a Canadian company, and the other three with US companies. Six out of nine of the Canadian agreements provide weak protections against both job insecurity and the intensification of work effort, whereas this was true for only seven out of twenty-five US agreements. The other agreements contained a mix of strong and weak provisions.
Considering the overall number of agreements analyzed within each country, no national regime was more strongly correlated with the strongest contract language. However, the US agreements were less likely to be among the weakest. One potential explanation is that the bargaining units affiliated with the US union were more successful at building institutional power resources through past mobilizations of associational power. This is only a hypothesis, as our data does not provide us with the complete history of bargaining dynamics across each local. However, the US central union in our study has historically concentrated its efforts at mobilizing local unions in telecom call centers; whereas the Canadian central unions are amalgamated unions in which telecoms must share priority with other industries in their portfolios. Given that the evidence we reviewed thus far suggests that unions had to win strong provisions through past mobilizations of associational power and continued to mobilize associational power resources to maintain those provisions, it is likely that this was the case for all the unions since they faced similar pressures.
Discussion and Conclusion
The introduction of automating technologies in service workplaces often threatens critical aspects of working life. However, these threats are rarely a fait accompli and can be resisted (Howcroft & Taylor, 2023). Drawing from labor process theory and power resources theory, this article advanced a framework to explain how unions can resist the harmful impacts of automating technologies on job quality. This framework was applied to analyze outcomes in call centers within Canadian and US telecom companies.
Our findings highlight the dual nature in which automating technologies impact the labor process and what challenges this poses for unions trying to mitigate their harmful effects on job quality. Building on current theorization on the two indeterminacies in the labor process (Smith, 2006), we showed how the implementation of automating algorithms and AI can become the object of both work effort and work mobility bargaining. The former represents workplace struggles over “work intensification” and the latter represents workplace struggles over “job insecurity”.
We identified two sets of outcomes. We first found that the implementation of automating technologies increased work effort (and hence work intensification) by producing errors in service delivery, increasing task complexity, and through mandated “protocols” (e.g., scripts and ease of access to agents) that encourage the use of self-service technologies. Workers whose jobs were not eliminated thus experienced heightened work intensification as they had to perform more steps and more complex tasks in servicing customers, fixing errors made by the algorithms, and coping with customers frustrated by algorithm-related service disruptions. A key issue for workers was that the increase in work effort and intensification made workers vulnerable to performance management-related discipline, as the extra workload made it difficult for them to meet performance standards set by managers. Second, we found that the implementation of automating algorithms and AI led to conflicts over worker mobility (i.e., job insecurity), as workers lacking job security protections in collective agreements were particularly vulnerable to job loss.
Unions sought to address these problems via bargaining over both work effort and work mobility. Union successes on either front were largely impacted by their access to institutional power resources. In terms of work effort and intensification, some unions succeeded at leveraging collective agreement provisions to restrict performance-related discipline. This proved crucial as these workers often had to navigate through automation-driven disruptions in service delivery which made it tougher for them to meet performance standards. Under the presence of strong contract provisions, workers had more control over work tasks and work effort, and hence, less intensification. Relatedly, they experienced less fear of management reprisals for shortcomings in performance. In terms of work mobility and job security, unions that could leverage stronger job protections from collective agreements experienced greater successes in protecting workers from automation-related job loss. Only a minority of the bargaining units could rely on strong collective agreement language to confront threats across both indeterminacies in the labor process. Most often, language was only strong within a single indeterminacy or weak altogether.
This brings us to the second set of findings which relates to power resources and the intertemporal dimension of resisting automation. Union success was usually contingent on the availability of collective agreement language negotiated in the past. Often, language dated from the distant past (even decades earlier), and these provisions were acquired through the mobilization of associational power resources (leading to strikes or public campaigns). The technologies may be new, but the issues are old. Thus, it is a question of scale, but not substance. Call center unions have long sought to limit managerial efforts to intensify the pace of work, including its enforcement with punitive forms of performance management, alongside efforts to employ technological advancements and outsourcing to downsize workforces (Doellgast, 2017). Successes in mobilizing and building power resources to expand the scope of worker control over all aspects of the labor process in the past can help support struggles in the future, even when the technologies are different.
This research makes three contributions to extant literature. First, we contribute to efforts at integrating labor process theory and power resources theory to explain union successes in regulating harmful digital technologies. For years, labor process theorists have emphasized that negative work outcomes from digitalization reflect power imbalances in the workplace and can be resisted (Braverman, 1974; Briken et al., 2017; Gandini, 2019; Thompson & Laaser, 2021). More attention has been given to resistance in gig work (e.g., Cant, 2019) or other work in non-union settings (e.g., Woodcock, 2016). Our research provides insights on how union power resources are shaping labor-management struggles over the digitalization of the labor process, including its impacts on the quality of work (Dupuis & Massicotte, 2025; Pulignano et al., 2025; Thomas & Turnbull, 2025).
Our findings are distinctive in that they shed light on historical struggles over how technology use impacts both work intensification and job security in the labor process. Emphasizing the intertemporal dimension, we show how current struggles are an extension of (and not a break from) prior struggles for control over the labor process. Union successes are traced to the continued maintenance of associational power resources on intertemporal scales. These power resources are not only mobilized to build institutional power resources in isolated periods of time. They are strategically re-mobilized as these institutional power resources come under threat. Thus, articulation becomes a critical union capability as power must be mobilized in “constant arbitration between actions as regards time and space” (Levesque and Murray, 2010). The presence of varied work issues across different indeterminacies in the labor process adds further complexity to this arbitrage, as union actions must be articulated both over time and across these indeterminacies.
Second, we provide important empirical insights on how labor contestation contributes to the social shaping (MacKenzie & Wajcman, 1999) of job quality outcomes from technology. A growing number of studies are showing that unions often pave the way to better outcomes (e.g., Berg et al., 2023), with rich insights on the impacts of “union presence” in the workplace (Gallie et al., 2017; Haapanala et al., 2023; ten Berge & Dekker, 2025), or the role of competing national models for industrial relations (Doellgast et al., 2023; Lloyd & Payne, 2019). Our study builds on those that explore the actions undertaken by unions on the shopfloor (e.g., Dupuis, 2025; Riordan et al., 2025), including an analysis of the role and contents of collective agreements (Montreuil & Foucher, 2023). In this sense, we also build on recent efforts to identify “best practices” in collective agreement language to address the threats digitalization poses to the quality of working life. 1
Third, the theorization underpinning our arguments entailed a reshaping of Smith's (2006) conceptualization of the double indeterminacy in the labor process. Traditional uses of this concept hinge on a narrow understanding of mobility power. Under this interpretation, mobility power reflects the extent that workers can exit (or threaten exit from) current employment arrangements. It is a function of worker skills and self-promotion alongside employers’ efforts at retention. We stress that workers also possess a symmetrical power to remain within the organization. Job security shaped by macro-structural factors such as trade union power and legislation represents an important aspect of workers’ mobility power. Contestation over this aspect of the labor process is a common reality for workers and their union representatives. We need to ensure that it does not remain a blind spot in contemporary studies of labor control in the sociology of work.
Future research can build on our contributions. First, an important part of our analysis focuses on the impacts of workplace-level bargaining on job security. However, a growing body of research suggests that the effects of new digital technologies are also conditioned by workers’ structural position in the labor market, such as migrant status and access to alternative employment opportunities, as well as by government policies shaping labor market conditions (Heiland, 2022; Schaupp, 2022; Vallas et al., 2022; Winton et al., 2025). Protecting workers from job loss is an important goal in the short term. However, unions and governments also play a broader role in providing workers with employment security. Labor market activation policies and union agreements that facilitate job reassignment and retraining are important resources for workers displaced by digital technologies. Examining interactions between different sources of regulation would help us better understand how to mitigate labor market insecurity.
Second, our findings suggest that unions with entrenched histories of resistance and hence control over different aspects of the labor process are better able to navigate threats from technological change. However, we also recognize that resistance is not a panacea for the malaise of digital labor. An alternative model involves unions co-governing firms and societies towards more sustainable futures. Future studies can consider the trade-offs between pursuing more conflictual or collaborative approaches to these transitions. The viability of either depends on what power resources are available to organized labor. Third, and finally, while our study focuses on formal aspects of collective bargaining, it pays less attention to the informal processes through which unions and managers negotiate the implementation of technological change. Prior research shows that such informal interactions can play a critical role in shaping workplace outcomes (O’Brady & Doellgast, 2021). Building on this, future research should examine how work effort and mobility bargaining unfolds through these informal processes and what types of power resources are mobilized in such interactions. Doing so would provide a more comprehensive understanding of how technological impacts are negotiated in practice, beyond formally codified agreements.
Footnotes
Acknowledgements
The authors are immensely grateful to Valeria Pulignano and Chris Tilly for their excellent editorial work. Furthermore, we wish to acknowledge Virginia Doellgast for her guidance and support, as well as her role in data collection. We also wish to acknowledge Simon Gim, Shannon Kribbs, Anna Sloots, Jelena Starcevic, and Della Walters for their indispensable research assistance.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this research was provided by the Social Sciences and Humanities Research Council of Canada (SSHRC) [grant numbers 430-2020-00045 and 1008-2020-0007] and the AFL-CIO Tech Institute.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.