Agilní rámec pro zavádění umělé inteligence ve veřejném sektoru

Artificial Intelligence (AI) adoption offers significant potential for public administration but encounters distinct challenges related to risk aversion, complex regulations, public values, and organizational constraints. A critical gap exists in practical, empirically validated frameworks tailored to guide iterative AI adoption specifically within this unique public sector context. This dissertation addresses this gap through the development and evaluation of the 'Adoption Framework for AI in the Public Sector' (AFAIPS). AFAIPS provides actionable guidance by integrating Rogers' Diffusion of Innovations theory with agile methodologies including Design Thinking, Lean Startup, and MLOps considerations. Employing an Elaborated Action Design Research (EADR) methodology, the framework was iteratively developed, initially informed by formative evaluation through semi-structured expert interviews, and then empirically evaluated via an in-depth, longitudinal case study. This primary case study involved applying AFAIPS to guide the implementation of an AI chatbot designed to enhance complex healthcare cost verification processes within the finance and controlling department of a Swiss cantonal corrections and reintegration office. The research evaluates the framework's effectiveness in structuring this real-world adoption process, highlighting the role of its integrated methodologies in managing uncertainty and ensuring user-centricity, and identifies key contextual factors and adaptations influencing successful implementation within public administration. This research contributes a rigorously developed and empirically evaluated framework offering practical guidance for public managers and teams undertaking AI initiatives, thereby bridging the gap between AI's potential and its successful, responsible implementation in the public sector.