The Limits of Social Dialogue in Algorithmic Work: Institutional Mismatch in Labour Governance Across Employment Forms

As AI supply chains increasingly rely on platform-mediated forms of work, the governance of labour — how work is evaluated, controlled, and collectively regulated — is shifting from institutional frameworks to algorithmic systems. Traditionally, social dialogue has structured labour governance through collective agreements, qualifications systems, and employment protections. This paper examines how algorithmic management (AM) relocates labour governance beyond the reach of social dialogue across three structurally distinct employment forms: platform-mediated cloudwork, on-location platform work, and institutionally embedded service sector employment. It argues that fragmentation across these forms is not incidental but constitutes a central mechanism through which social dialogue fails to engage with AM. The paper asks what forms of institutional realignment would be required to bring AM within the scope of collective governance.

The analysis integrates three streams of original empirical evidence. First, it draws on platform-level analysis of 90 cloudworker profiles from Serbia active on Upwork, collected via Gigmetar, a proprietary monitoring tool developed by the authors. Profiles are analysed through coding of structured and narrative signals to examine how work is evaluated and made visible in platform-mediated knowledge work. Second, the study includes 20 qualitative interviews with food delivery couriers working for Glovo and Wolt in Serbia, focusing on how algorithmic systems structure task allocation, performance evaluation, and behavioural adaptation. Third, it incorporates 11 qualitative interviews with social partners, alongside institutional analysis of collective agreements, social dialogue texts, and position papers conducted for a Eurofound thematic review on collective bargaining and AM, covering national, sectoral, and company-level dynamics.

The findings reveal a fragmented governance vacuum across all three employment forms. In cloudwork, algorithmic signals — ratings, metrics, and platform-generated indicators — determine visibility and access to opportunities, displacing credential-based evaluation beyond institutional oversight. In on-location platform work, algorithmic governance is opaque, non-portable, and largely uncontestable, requiring continuous behavioural adaptation to maintain performance scores without meaningful avenues for appeal. At the institutional level, collective agreements are concentrated in sectors where AM is largely absent (e.g. construction, manufacturing, transport), whereas the service sector — particularly ICT, finance, professional services, and platform-mediated activities — where AM is most actively deployed, remains largely excluded from bargaining coverage. As a result, labour continues to be governed, but no longer through institutions capable of exercising collective oversight.

The paper develops the concept of institutional mismatch as a multi-sectoral analytical framework, arguing that the challenge facing social dialogue is not simply regulatory lag but a structural relocation of labour governance to spaces beyond the reach of existing institutions. Drawing on labour process theory, industrial relations theory, and conceptualisations of algorithms as epistemic infrastructures (Beer, 2019), and informed by platform labour governance scholarship (De Stefano & Aloisi, 2018), it shows that AM operates across employment forms in ways that disrupt sector-based regulatory logics. In response, the paper proposes three axes of realignment: technology-based rather than sector-based bargaining rights, graduated signal transparency rights calibrated to collective institutional capacity, and cross-border social dialogue mechanisms aligned with the transnational character of AI supply chains.