Abstract
European public agencies increasingly adopt Artificial Intelligence (AI)-based chatbots. However, research on this topic is limited to an understanding of chatbots as interfaces between citizens and caseworkers. We performed a qualitative study of the design and deployment of the chatbot Selma by a public Labor and Welfare agency (LWA) in a Scandinavian country. We illustrate how Selma has gradually transformed the relationship between citizens and caseworkers. Selma has evolved from a communication channel between citizens and the public agency to an agent that coordinates with citizens and co-workers at the agency. We show how caseworkers’ decisions about Selma’s features impact the (re)organization of the caseworkers’ tasks to also include chatbot training. We suggest that making decisions about how AI is designed and deployed is intertwined with how decisions are made with AI. We conclude by discussing the implications of our study for human-AI collaboration in the public sector.