With a significant amount of software now being developed for use in patient-facing studies, there is a pressing need to consider how to design this software effectively in order to support the needs of both researchers and patients. We posit that a microservice architecture—which offers a large amount of flexibility for development and deployment, while at the same time ensuring certain quality attributes, such as scalability, are present—provides an effective mechanism for designing such software. To explore this proposition, in this work we show how the paradigm has been applied to the design of CONSULT, a decision support system that provides autonomous support to stroke patients and is characterised by its use of a data-backed AI reasoner. We discuss the impact that the use of this software architecture has had on the teams developing CONSULT and measure the performance of the system produced. We show that the use of microservices can deliver software that is able to facilitate both research and effective patient interactions. However, we also conclude that the impact of the approach only goes so far, with additional techniques needed to address its limitations.