The development and successful application of analytical solutions relies on data, analytical models and their configurations. In today’s smart service systems, these are typically distributed across various organizational entities. However, an exchange of their data, analytical models or configurations within the system often is not feasible due to inter-organizational barriers (e.g., data confidentiality preservation). Thus, the potential of many smart service systems still remains unexploited. We argue that the application of base, meta and transfer machine learning could overcome these barriers: It allows to find patterns and make predictions across these distributed data sources as well as exchange analytical models and configurations across entities. In this work, we describe how a cognitive entity in a smart service system can be constructed to support comprehensive analyses as well as the exchange of analytical knowledge across entities.