AI systems encode values and distribute benefits long before anyone interacts with them, through decisions about data, context of use, intended users, and interaction space. I am a postdoc at the University of Copenhagen. I use research-through-design and participatory methods to learn how these systems embody perspectives and shape outcomes, extending our understanding of human-centred AI. I aim to investigate how systems that benefit many users also redistribute expertise, reshaping what it means to know, decide, and act in AI-mediated environments.