If you are a care provider struggling to offer personalized health support to growing numbers of elderly residents without proportionally increasing staff — this project developed a multi-agent virtual coaching platform with behavior change detection that delivers integrated physical, social, cognitive and mental support. The system was tested through 3 iterative prototypes with real users across 6 countries and 8 partner organizations.
Virtual Health Coaching Team for Elderly Care Using AI-Powered Conversational Agents
Imagine having a team of personal health advisors on your tablet — one for exercise, one for diet, one for mental wellbeing — and they actually talk to each other about what's best for you. That's what COUCH built: a group of virtual coaches, each with their own personality and expertise, who work together to help older adults stay healthy. They listen, suggest goals, and even debate among themselves about the best advice, just like a real team of specialists would. The system learns your habits over time and adapts its coaching to what actually works for you.
What needed solving
Elderly care providers face a growing gap between the number of seniors needing personalized health coaching and the staff available to deliver it. Current digital health tools typically address only one aspect of wellbeing — exercise OR diet OR mental health — forcing users to juggle multiple disconnected apps. There is no easy way to provide older adults with integrated, multi-domain coaching that adapts to their individual needs and behavior patterns over time.
What was built
The team built a complete virtual coaching platform featuring multiple AI-powered conversational agents, each with distinct expertise and personality, that collaborate to support elderly users. Concrete deliverables include: a final integrated technical prototype connecting all subsystems, a shared knowledge base for coordinating coaching actions, 3 behavior analysis prototypes (short-term, long-term, and change detection), a spoken dialogue interaction system with embodied conversational agents, and an open agent platform enabling third-party developers to add new coaching modules.
Who needs this
Who can put this to work
If you are a health insurer looking to reduce costly hospitalizations among elderly policyholders through preventive coaching — this project built a shared knowledge base and long-term behavior analysis system that tracks lifestyle changes over time. The platform supports multiple coaching domains simultaneously, backed by EUR 3,704,000 in EU research funding and 52 deliverables covering everything from dialogue systems to sensing capabilities.
If you are a digital health company wanting to add intelligent coaching to your platform — this project created an open agent platform where new virtual coaches can be plugged in by third-party developers. The embodied conversational agents use spoken dialogue interaction and the architecture was designed to connect with FIWARE and universAAL communities for broader deployment.
Quick answers
What would it cost to license or integrate this coaching platform?
The project received EUR 3,704,000 in EU funding under an RIA (Research and Innovation Action) scheme, meaning results are typically available under favorable licensing terms. Specific licensing costs would need to be negotiated with the coordinator, Universiteit Twente. As publicly funded research, the platform architecture and open agent design suggest accessibility for commercial partners.
Can this scale to thousands of elderly users simultaneously?
The final technical prototype integrated all components — shared knowledge base, sensing, dialogue, and embodied agents — into a fully working system. However, the project focused on validating the concept through iterative prototypes rather than large-scale deployment. Scaling to thousands of concurrent users would require additional infrastructure work.
Who owns the intellectual property and can we license it?
As an EU-funded RIA project, IP typically stays with the consortium partners who generated it. The consortium includes 8 partners across 6 countries (BE, DK, ES, FR, NL, UK), with Universiteit Twente as coordinator. Any licensing arrangement would need to involve the relevant IP holders within the consortium.
Has this been tested with real elderly users?
Yes. The project ran 3 iterative functional prototypes specifically designed for gathering user input. The initial prototype used Wizard-of-Oz methods, the second focused on user experience, and the third tested acceptance and usability of advanced features including agent animations and behaviors.
Can we add our own coaching modules to the platform?
The platform was explicitly designed as an open agent architecture — not confined to a fixed number of coaches. It enables developers and innovators to introduce new coaches that provide additional expertise and services. The system was built to connect with FIWARE and universAAL communities for ecosystem expansion.
What health domains does the coaching cover?
The virtual coaches cover physical, social, cognitive and mental wellbeing support. The system includes short-term and long-term behavior analysis, behavior change detection, and a shared knowledge base that coordinates advice across all coaches. Each coach has distinct expertise, personality and coaching style.
Is this compliant with health data regulations?
Based on available project data, the system includes a shared knowledge base for collecting and retrieving user health information. The consortium includes 4 universities and 2 research organizations, suggesting academic rigor in design. However, specific GDPR or medical device regulation compliance details would need to be confirmed with the coordinator.
Who built it
The COUCH consortium brings together 8 partners from 6 European countries (BE, DK, ES, FR, NL, UK), led by Universiteit Twente in the Netherlands. The mix is research-heavy — 4 universities and 2 research organizations — with 2 industry partners including 2 SMEs, giving a 25% industry ratio. This composition is typical for a technology validation project: strong academic depth for building the AI coaching agents and dialogue systems, but limited commercial muscle for taking it to market. A business partner looking to commercialize this would be joining a consortium rich in technical know-how but likely needing additional commercial and distribution partnerships to reach elderly care markets at scale.
- UNIVERSITEIT TWENTECoordinator · NL
- UNIVERSITY OF DUNDEEparticipant · UK
- ROESSINGH RESEARCH AND DEVELOPMENT BVparticipant · NL
- FONDEN DEMOCRACY Xparticipant · DK
- UNIVERSITAT POLITECNICA DE VALENCIAparticipant · ES
- INNOVATION SPRINTparticipant · BE
- CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE CNRSthirdparty · FR
- SORBONNE UNIVERSITEparticipant · FR
Universiteit Twente (Netherlands) — reach out to their technology transfer office or the project's principal investigator for licensing and collaboration inquiries.
Talk to the team behind this work.
Want to explore how COUCH's virtual coaching technology could fit your elderly care or digital health product? SciTransfer can connect you directly with the research team and help structure a technology transfer or licensing deal.