Governing generative AI in public administration: Productivity, work redesign and reflective human–AI collaboration

This article examines the organizational adoption of generative artificial intelligence in public administration through a mixed-methods quasi-experimental case study of Microsoft Copilot in the Emilia-Romagna region, Italy. The authors investigated how generative AI reshapes productivity, work practices and professional capabilities, and the conditions under which it may generate public value. The findings show measurable time savings, especially in textual, communicative and information-retrieval tasks, with estimated productivity gains ranging from 2% to 7% depending on adoption intensity and quality. However, value for money does not depend on automation alone. It requires organizational governance, workflow integration, human validation, training and new reflective capabilities. The article contributes to public management by showing that generative AI should be understood not merely as a tool for task automation, but as a socio-technical intervention that redesigns work, affects professional roles and requires public organizations to govern the interaction between algorithmic output and human judgment.