Agentic AI systems could carry specialist dementia expertise into primary care clinics, where most Alzheimer’s diagnoses are missed or delayed, researchers argue in a perspective published July 30 in Nature Aging.
Teams from Emory University and the University of California, San Francisco envision AI agents that train on data from practicing specialists, then help generalist clinicians collect and interpret complex patient information, streamline workflows and apply the latest medical knowledge. The system would feed outcomes back to specialists in a continuously learning loop, so care improves as the model is used.
The paper’s authors say US neurological care demand is outpacing the workforce. Demand for dementia specialists far exceeds supply, and diagnosis in primary care is often delayed or missed, a gap agentic AI could narrow by scaling specialist-level support to nonspecialist settings.
Their roadmap runs through six phases: standardized multimodal data collection, decision support, clinical workflow integration, rigorous validation and monitoring, continuous learning through clinical feedback, and robust ethics and risk management frameworks.
The approach is deliberately human-centered, the authors write. AI would optimize clinician capabilities in data collection, interpretation and timely application of medical knowledge, while clinicians remain responsible for care decisions. The framework prioritizes patient safety, healthcare equity and transparency.
Generative AI built on large language models now makes such systems technically feasible, the authors note, but responsible design and phased integration into care are essential. They call for validation protocols and monitoring before deployment at scale.
The perspective is authored by Andrew G. Breithaupt, Michael Weiner and colleagues, with corresponding authors at Emory and UCSF.
