This study investigates the interplay between human and AI systems in the public sector via a case study of Microsoft Copilot in a Swedish municipality. Examining the autonomous capabilities of Copilot reveals shifts in task delegation, workforce dynamics, and ethical responsibility. The findings indicate tensions between efficiency oriented towards an AI agent and the desired effects of human oversight as Copilot begins its steady pathway of transition from a passive tool into a collaborative partner. Nonetheless, as Copilot reduces the burden of routine administrative tasks, issues of inequitable access and opaque decision-making have implications that may result in further socio-technical stratification and role displacement. By situating human-AI delegation as a mutually constitutive process, the study contributes to the Information Systems literature and brings insights into harmonizing AI integration with bureaucratic accountability and governance in public institutions, where institutional trust is foundational.