Knowledge sharing, responsibility, and coordination as GenAI use expands : The case of Svea usage in Swedish municipalities

Sammanfattning

The thesis examines how knowledge sharing, responsibility, and risk are organized around Svea, a shared generative AI assistant used across Swedish municipalities. As GenAI use expands in the public sector, organizations need to coordinate not only technical access but also everyday practices for guidance, accountability, and responsible use. However, it is still unclear how this coordination happens in everyday work across municipalities. Based on eight semi-structured interviews with coordinators from seven municipalities, together with non-sensitive municipal documents and artifacts, the study examines how shared GenAI use is organized in everyday municipal work. The findings show that knowledge sharing around Svea takes a hybrid form. Municipalities use codified guidance such as guidelines, onboarding material, and shared channels, while much practical support remains personalized through coordinators, peer networks, and informal discussions. Responsibility and risk are often handled through individual judgment, especially when output reliability, legal uncertainty, and data restrictions remain unresolved. The study also shows that artifacts such as prompts, assistants, and routines carry governance knowledge, but their maintenance is fragile when ownership depends on individuals rather than formal structures. As participation expands, cross-municipal knowledge sharing creates more available material, but also increases the need for filtering, local adaptation, and trusted peer networks. The thesis contributes to research on GenAI governance, interorganizational coordination, and knowledge sharing by showing that knowledge sharing becomes a form of governance work in shared GenAI initiatives.