A systematic mapping of generative AI-powered virtual assistant in European public administration

Abstract

The rapid development of LLMs has accelerated the adoption of GenAI across multiple sectors. Among current applications, GenAI-powered virtual assistants represent the most prominent and widely implemented use case in public sector. However, there is currently little research that systematically maps the functions, anticipated effects and trends of these emerging tools in public sector. Addressing this, this study maps and analyses 39 GenAI assistant deployments across European public administrations, examining their functional characteristics and anticipated public value effects. Based on publicly available documentation, the findings show that citizen-facing assistants are primarily designed for information provision and service guidance, with expected effects focused on improved availability, accessibility, convenience, responsiveness, efficiency, cost and time savings, and fairness, etc. Internal assistants mainly support knowledge retrieval and document workflows, aiming to enhance administrative efficiency, productivity, resource management, and institutional capacity and processes, etc. However, expected effects related to trust and institutional legitimacy are not consistently supported in reported outcomes, highlighting the need for deeper empirical and qualitative investigation on public value effects and for governance frameworks that enable responsible GenAI integration in public sector.