Virtual Assistants – Feasibility in real-life setting

26 January 2026

Following the initial analysis of rehabilitation practices, the project progressed to the development and testing of Virtual Assistant (VA) prototypes. These pilots are built on the OpenAI Assistants framework and employ a Retrieval-Augmented Generation (RAG) approach to ensure that responses are grounded in verified knowledge sources. In parallel with the RAG-based development, the VAs were further adapted through fine-tuning. For fine-tuning, a dataset of high-quality question–answer pairs was created to reflect the desired behavior and domain-specific use of the VA. Each participating country focuses on a specific rehabilitation context:

Finland: Supporting individuals with Multiple Sclerosis (MS), particularly in managing fatigue
Estonia: Addressing chronic pain
Poland: Assisting patients with Chronic Obstructive Pulmonary Disease (COPD)

The development process has been collaborative and evidence-driven. In addition to the literature review, rehabilitation professionals, physicians, and physiotherapists have contributed their expertise to define relevant content and ensure clinical accuracy. We have involved different voluntary groups in usability testing, providing both quantitative and qualitative feedback on early versions of the VA. Especially the ease of use received positive feedback. Experiences and initiative results from the first focus groups were discussed at the second partner meeting held in Eskilstuna. The onsite meetings provide a perfect opportunity for insightful discussion between partners.

The first demonstrative versions are now ready, accompanied by clear instructions for end users. The coming months will focus on refining these prototypes through real-life testing and gathering feedback from both professionals and patients. This iterative process will help ensure that the solutions are practical, ethically sound, and aligned with the needs of public healthcare.