The Rise of Agentic Artificial Intelligence in Medicine: A Systematic Review of Autonomous Systems, Multi-Agent Collaboration, and Ethical Governance

Abstract

Artificial Intelligence (AI) in medicine has traditionally functioned as a passive decision-support tool that provides recommendations without autonomous action. The emergence of Agentic AI introduces a new paradigm in which AI systems can plan, execute, and self-correct tasks, creating a shift toward more collaborative human–machine interaction. However, existing literature remains fragmented, with technical development, multi-agent coordination, and ethical governance often discussed separately. This systematic review aims to synthesize current evidence on Agentic AI in medicine by examining autonomous architectures, multi-agent collaboration, clinical validation, ethical challenges, and future implementation barriers. The study followed the PRISMA 2020 framework and employed thematic narrative synthesis of 34 primary studies identified from major academic databases. The findings indicate that orchestrator-based multi-agent architectures integrated with large language models and retrieval-augmented generation represent the dominant approach in medical Agentic AI development. However, real-world clinical validation remains limited, with most studies still focusing on simulations and proof-of-concept evaluations. Furthermore, ethical and governance studies represented the largest research category, highlighting concerns regarding accountability, transparency, and human oversight. This review concludes that the future adoption of Agentic AI requires simultaneous advancement of technical capabilities and governance frameworks, supported by evaluation approaches that consider accuracy, computational efficiency, clinical safety, and human control.