From Hierarchies to Loops: Rethinking Public Sector Structures for Responsible AI Integration

<span>Public institutions worldwide are increasingly deploying artificial intelligence to support high-stakes decisions, yet most bureaucratic structures remain designed for linear, hierarchical governance rather than continuous human–AI feedback.</span> This paper argues that existing public administration models are structurally misaligned with AI-mediated decision systems, producing risks to accountability, legitimacy, and performance. Drawing on public sector case evidence and decision science theory, the paper introduces a loop-based model of institutional design that reconceptualises bureaucracies as adaptive intelligence systems rather than rule-processing hierarchies.