Abstract
In an increasingly fast-paced and dynamic world with exponentially more data, more roles are required in agile software development. At the same time, the development team needs to maintain speed and autonomy. The challenges surrounding the organization of cross-functional teams are thus exacerbated. Through a multi-case study with interviews from five organizations, we show how agile companies use different models of organizing data scientists. We find that there are specific challenges related to each of these organizational models and that some challenges are shared among all the organizational models. Challenges include difficulty coordinating development strategies and a lack of resources. In addition, we identify strategies used to overcome the challenges, including coordinating mechanisms for platform teams, communities of practice for data scientists, and the development of shared playbooks.