While the self-learning nature of AI systems that use machine learning calls for sustained co-creation between developers and users during development, implementation and use, information systems and management scholars still largely build on a long-established tradition of separating technology development from use. Instead, the self-learning nature of AI calls for letting go of this tradition to separate between development and use, which is starting to happen in practice but has not yet found appropriate theoretical and methodological tools among researchers. In this paper we show some real-life attempts to develop sustained collaboration among developers and users, based on empirical cases of five organizations. In particular, we propose how blurring boundaries makes data production, explainable AI and AI deployment fields of practice where development and use intertwine. We suggest embracing the blurred boundaries of AI implementation in our theorizing, understanding the different parts of AI as fields of practice where development and use come together in the co-creation of AI and work.