AI Agents in Healthcare: The Need for Governance

Abstract

Artificial intelligence (AI) agents are poised to redefine virtual care by streamlining routine tasks and enhancing clinical decision-making. Unlike static decision aids, these systems can autonomously manage complex, multi-step processes such as care coordination, longitudinal patient monitoring, and data integration across fragmented health systems. Early applications include voice-enabled assistants for documentation and order entry, as well as health wallets that advance the self-sovereign patient paradigm by actively managing data sharing, consent, and treatment planning. The next stage is the rise of multi-agent systems, where specialized agents collaborate with one another and with clinicians to deliver distributed, adaptive care. These advances offer solutions to workforce shortages, administrative burden, and patient engagement, yet also raise new challenges around trust, liability, bias, and emergent risk. This article argues for governance by design as a critical framework, embedding oversight throughout the agent lifecycle and extending it from individual tools to collective multi-agent ecosystems.