This thesis explores how Large Language Models (LLMs) can be integrated into physiotherapy workflows, specifically how it may support journaling without compromising patient interaction. By conducting interviews, fieldwork and testing with physiotherapists, this work identifies opportunities using artificial intelligence (AI) in clinical documentation. Results indicate that these tools ought to augment writing while automating tedious tasks without diminishing professional reflection. Moreover trust is identified as emerging through repeated interactions and explainability. The final design is a wearable device, capturing audio and video during treatment to enable a comprehensive context window. This project contributes to understanding human-agent collaboration and AI as a design material, in interaction design.