Enacting AI in everyday work: affordances, constraints, and emerging practices during early-stage implementation in a public sector organization

Abstract

Organizations are under growing pressure to integrate artificial intelligence into their operations, yet the reality of what happens when AI is introduced into everyday work remains poorly understood. While existing research has tended to focus on outcomes, success factors, and leadership perspectives, the experience of employees navigating AI in practice has received comparatively little attention, particularly in public sector contexts where demands for accountability, transparency, and limited resources intensify the stakes of implementation. This study examines how employees engage with AI tools during an early stage of implementation in a Swedish public sector organization, and what organizational conditions shape, afford, and constrain these practices. Using a qualitative approach, 20 semi-structured interviews were conducted with employees across multiple roles within a financial unit where Microsoft Copilot had recently been introduced. Applying the theoretical framework of sociomateriality, the findings show that employees engage with AI in cautious and situated ways, treating it as a supplementary support tool while retaining full accountability for its outputs. Rather than replacing professional expertise, AI redistributes it toward evaluation, contextualization, and judgment. The absence of formal governance affords bottom-up experimentation while simultaneously reproducing existing differences in digital confidence, and organizational sustainability values emerge as an unexpected constraint on adoption. This study contributes to the literature on AI adoption and digital transformation by showing how early-stage implementation in a public sector organization unfolds as a fragmented and bottom-up process, where individual competence, informal peer learning, and the absence of formal governance collectively shape how AI is enacted in everyday work.