In the field of social services, practitioners work with homeless clients to help them exit homelessness. Due to the uniqueness of each client and the dynamic nature of their environment, practitioners must choose from a variety of theories and techniques to perform the required tasks. A major obstacle in automating these tasks is the lack of high-fidelity models that emulate client behaviour. The purpose of this paper is to identify the requirements for emulating a homeless client and how artificial intelligence (AI) can fulfill these requirements. This paper also highlights how such an emulation can contribute to the areas of social service policy evaluation, understanding client behaviour, and education.