AI-Care: A Conversational Agentic System for Task Coordination in Alzheimer's Disease Care
AI-Care is a conversational AI system designed to help individuals with Alzheimer's disease and related dementia manage daily tasks through natural language interaction, reducing cognitive barriers to using digital tools. The system prioritizes safety through caregiver-verified records and controlled clarification flows, with preliminary pilot testing showing positive user trust and task completion outcomes.
AI-Care represents a practical application of agentic AI systems to healthcare accessibility, specifically addressing the gap between digital tool functionality and cognitive impairment. The system acknowledges a critical user need: individuals with AD/ADRD face significant friction when navigating multi-step digital interfaces, creating barriers to independent task management. Rather than forcing users to adapt to existing software, AI-Care simplifies interaction through conversational interfaces that map natural language requests to backend actions.
The architectural approach demonstrates thoughtful safety-first design principles. By grounding medical responses in caregiver-verified data rather than model generation, the system avoids hallucination risks that could have serious health consequences. The stateful orchestration pipeline—moving requests through sanitization, intent classification, and explicit slot collection—prioritizes reliability over cutting-edge model sophistication. This reflects maturity in healthcare AI implementation, where deterministic pathways often outperform uncontrolled generative approaches.
The pilot study with four mild-to-moderate AD/ADRD patients showing high trustworthiness and task completion rates suggests viable market demand. Healthcare providers, family caregivers, and institutional care settings represent potential adoption vectors. The voice-first design with controlled text-to-speech chunking directly addresses accessibility needs for users with cognitive decline.
Future scaling depends on validation across larger patient cohorts and clinical evidence of sustained benefit. Integration with existing electronic health record systems and remote caregiving platforms will determine real-world deployment success. The dual input/output support (text and voice) increases usability across varying user capabilities. This work positions conversational AI as a tool for reducing healthcare friction rather than replacing human judgment.
- →AI-Care uses stateful orchestration and caregiver-verified records to ensure safety in healthcare AI without relying solely on model generation.
- →The system addresses real accessibility barriers by simplifying multi-step digital tasks through natural language conversation for cognitively impaired users.
- →Pilot results show high user trust and task completion, indicating potential adoption in remote caregiving and institutional settings.
- →The design prioritizes controlled clarification flows over silent failures, reducing ambiguity in safety-critical medical contexts.
- →Voice-first interaction with chunked text-to-speech output directly supports accessibility for users with Alzheimer's disease.