Hello! I’m Svea: An AI innovation project in the Swedish public sector - A qualitative study of how public organizations collaborate to manage knowledge boundaries in the age of AI

Sammanfattning

The adoption of generative artificial intelligence (AI) in the public sector is increasingly framed as a response to demands for efficiency and improved service quality. However, existing research has primarily focused on AI implementation within single organizations, offering limited insight into how public organizations collaborate around shared AI initiatives. This study examines the Swedish innovation project Svea, a shared digital assistant implemented across more than 120 public-sector organizations. Based on 20 semi-structured interviews and 26 observations, the study explores how knowledge is shared, interpreted, and adapted across organizational contexts. Drawing on boundary object theory, knowledge boundary perspectives, and research on knowledge brokering, the findings show how Svea functions as a shared reference point for collaboration, but that its coordinating role depends on continuous brokerage and boundary-spanning work. The study further demonstrates that transfer, translation, and transformation of knowledge occur simultaneously rather than sequentially, shaped by differences in organizational conditions such as AI maturity and governance structures. By shifting the focus from single organizations to interorganizational collaboration, the study contributes to public-sector AI research by showing how shared AI initiatives are coordinated and enacted differently across diverse organizational settings.