Small Language Models for Developing Agentic AI in Healthcare: A Comprehensive Systematic Review and Critical...

Agentic artificial intelligence (AI) systems are emerging as a transformative approach in healthcare, enabling autonomous task execution through integrated reasoning and tool use. While early implementations have largely relied on large language models (LLMs), growing evidence suggests that smaller language models may be better suited for many healthcare workflows due to their efficiency, scalability, and practicality in real-world clinical environments. This review examines the current landscape of small language models (SLMs) used in agentic healthcare applications, including clinical documentation, decision support, patient triage, and administrative automation. We synthesize available evidence on their performance, safety, and economic implications, and discuss key considerations for clinical deployment, including regulatory alignment and governance. Overall, small language models appear to offer sufficient capability for most agentic healthcare tasks while providing meaningful advantages in deployability, cost, and operational efficiency, supporting their role as a viable and often preferable alternative for clinical implementation.