Labor Power and the Role of Subcontracting in the AI Economy

Recent public discourse on artificial intelligence has begun to pull back the curtain on the systematic failures of global labor supply chains, highlighted by reporting regarding the sudden redundancy of more than 1,000 Kenyan data workers after Meta terminated its contract with Sama, a data labeling firm. While these investigations reveal a landscape of chronic precarity, they also point to a deeper structural reality: work in the global economy is deliberately organized through what South African journalist Marché Arends calls an “outsourcing maze,” enabling global capital to extract economic value from workers through layered webs of intermediaries.

The dynamics may evoke digital colonialism, yet the underlying logic is central to capitalism, wherein local intermediaries mediate international capital-labor relations, enabling the circulation and extraction of surplus value. While on-demand AI development through crowdworking platforms is rising, reliance on intermediaries remains central to a system of uneven development which favors large subcontractors over smaller local firms due to their reputational capital and perceived trustworthiness.

Workers are indispensable to production processes and to generating surplus value for capital. Yet labor power is also marked by “indeterminacies” or uncertainties that management continually seeks to control. This has particularly significant implications for context-dependent work, such as content moderation or data entry, and the business risks it entails. As a result, technology companies outsource such work to specific suppliers and employ distinct outsourcing mechanisms to manage this labor indeterminacy in outsourced service work.

These subcontractors exercise extensive normative and bureaucratic control, cultivating cultures of professionalism, ensuring productivity, and reducing labor resistance in the workplace. Yet, they are not simply instruments of global capital; their own positioning within global value chains (GVCs) varies depending on how they can negotiate for power by capturing higher process and product related activities in the value chain relationship.

In 2020, the technology company Cognizant—a key player in the export-oriented IT services sector—officially exited from its content moderation project with Meta. The project was transferred to comparable subcontractors of similar scale, reputation and project portfolio. Cognizant’s decision to exit was partially motivated by negative publicity after a public revelation of the working conditions at its moderation center in Arizona in the United States. But it was also informed by the asymmetrical and controlled relationship in the content moderation value chain that Cognizant found itself in with Meta. Despite substantial exit and project transfer costs, as visible in its SEC filings for that financial year, the company chose to exit because it could not upgrade to higher-value activities within the content moderation value chain, constrained by the extensive control Meta exercised through its priority work software and algorithmic management over workers.

In India, Cognizant transferred a dominant share of its workforce in the city of Hyderabad to another large subcontracting firm. For the worker, being transferred from one corporation to another rarely results in improved working conditions, opportunities for skill development and upward mobility. Instead, it often reinforces their position at the bottom of the value chain hierarchy. The new employer typically inherits the same control mechanisms—high surveillance, low pay and limited possibilities for career mobility—ensuring that the labor process remains stable even as the corporate entities change. The transfer of work reflects the historical pattern of labor process fragmentation, in which qualitative advances in information networking technology have played a central role in enabling the “transferability of the techno-economic system of the call center” to offshored and outsourced firms.

Having subcontractors as intermediaries to the capital-labor relation also mystifies how work is organized and how the international class relations function. Workers are kept in the dark about the identity of the global firms they work for, the scale of the supply chain they are embedded in, and the source and location of the control exerted over their labor. The use of AI and algorithmic technologies further hides the potential choke points in the supply chain that workers can organize around to stall the production process.

Garment supply chains offer examples of this dynamic: worker organizing, combined with sustained negotiations between trade unions and suppliers, has led to agreements securing workers’ reinstatement and measures to address gender-based sexual violence. Such outcomes become far harder to achieve in GVCs adopting and developing AI, both because class institutions such as trade unions are weakened and because of a broader industrial culture of secrecy that is enforced through non-disclosure agreements and the restriction of workplace and organisational information from workers.

The failure of AI value chains to serve workers is not an aberration but reflects the uneven development inherent to capitalism, where benefits accrue disproportionately to large subcontracting companies capable of negotiating higher-value activities. However, as global tech companies consolidate their grip over the production process, opportunities for subcontractors in these value chains have grown scarcer. Developmental gains, meanwhile, do not trickle down to workers and affected communities. The failure then must therefore be understood not as an exception but as a structural condition, one that is bound to reproduce so long as workers remain in an asymmetrical position vis-à-vis capital. This asymmetry is felt acutely by the global workforce of data labelers and content moderators whose labor sits at the peripheries of innovation and value generation.

How, then, does one respond to this political project of disempowerment? A critical governance gap is emerging around labor and environmental standards in the adoption and development of AI technologies within GVCs. Existing supply chain due diligence (SCDD) laws only partially apply to AI companies, leaving significant regulatory voids. Additionally, they have only addressed direct suppliers, leaving the rest of the supply chain uncovered. Meanwhile, most private initiatives filling this space remain downstream-focused, prioritizing consumer harms while upstream labor rights and environmental impacts embedded in the production process remain under-reported and under-examined.

Will these soft governance strategies—common in many GVCs and framed as they are around cooperative logic between corporations, states, and civil society—genuinely represent social interests? This is possible in moments of labor struggles and social movements, but given the increasing authoritarianism suppressing worker and marginalized voices, there is a real fear that class struggle will play out at the top, leaving behind the interests of workers and even subcontracting companies that are exposed to the brutal competition of the AI race. Ultimately, genuine accountability for workers necessitates a rethinking of the organizing logic of social and economic life that can prioritize collective human needs over the demands of capital.