Overview

Medicine is undergoing a fundamental shift from using artificial intelligence (AI) as passive, single-step tools to deploying autonomous agents capable of reasoning, planning, and acting across complex clinical workflows. Unlike conventional medical AI models that depend on human users to provide input and interpret output, agentic systems can proactively monitor patient data, orchestrate multi-step diagnostic or therapeutic decisions, and collaborate with clinicians as genuine healthcare teammates. Recent studies spanning electronic health record (EHR) analysis, medical imaging, treatment planning, and differential diagnosis have demonstrated that multimodal large language model (MLLM)-based agents consistently outperform standard MLLMs on clinical tasks, signaling enormous promise for transforming healthcare delivery.

Despite this rapid progress, significant gaps remain between current research and real-world clinical deployment. On the technical side, agent architectures optimized for general-purpose tasks often lack the domain-specific reliability, multimodal integration, and long-horizon reasoning required in high-stakes medical settings. On the evaluation side, existing benchmarks predominantly assess single-step question answering (QA) and fail to capture the multi-step reasoning, tool use, and adaptive decision-making that characterize truly agentic behavior in clinical workflows. More critically, the pace of technical innovation has substantially outpaced the development of safety frameworks, governance infrastructure, and regulatory guidance, leaving clinicians, health systems, and policymakers without adequate tools to assess or manage the risks of autonomous clinical agents. At the same time, questions around human-AI teaming, clinician trust, workflow integration, and health equity remain insufficiently addressed as deployment accelerates in practice.

To this end, we introduce the First Workshop on Agentic Medical AI (AMAI), an online workshop co-located with IEEE-BIBM 2026. This workshop aims to directly address these gaps by bringing together researchers and practitioners from agntic AI, MLLM, clinical informatics, and medicine. We seek to advance the field across four interconnected dimensions: (1) the design of robust agent architectures and their clinical applications; (2) human-AI interaction and collaborative care models; (3) rigorous evaluation, benchmarking, and safety; and (4) governance, ethics, and real-world deployment. By fostering dialogue across academic research and clinical practice, this workshop aims to accelerate the translation of agentic medical AI from laboratory prototypes into trustworthy, equitable, and impactful healthcare solutions.

Topics of Interest

• Agent Architectures and Clinical Applications

  • Design and evaluation of single-agent and multi-agent architectures for clinical tasks

  • Tool-augmented and retrieval-augmented generation (RAG) in medical agents

  • Memory, planning, and reasoning mechanisms for long-horizon clinical tasks

  • Multimodal medical agents integrating text, imaging, genomics, and sensor data

  • Agentic AI for clinical decision support, differential diagnosis, and treatment planning

  • AI agents for patient monitoring, triage, early warning, and personalized medicine

  • Agentic workflows in radiology, pathology, drug discovery, and EHR analysis

• Human-AI Interaction and Collaborative Care

  • Human-in-the-loop design and clinician-AI teaming in agentic systems

  • Shared decision-making, trust, transparency, and explainability

  • Conversational and patient-facing AI agent interfaces

  • Cognitive load, workflow integration, and clinician adoption

• Evaluation, Benchmarking, and Safety

  • Benchmark design for multi-step clinical reasoning beyond multiple-choice QA

  • Robustness, generalizability, and failure mode analysis of medical agents

  • Hallucination detection and mitigation in high-stakes clinical contexts

  • Safety frameworks, risk management, and runtime monitoring for autonomous agents

• Governance, Ethics, and Real-World Deployment

  • Regulatory pathways and compliance (FDA, EU AI Act, etc.) for agentic systems

  • Bias, fairness, and health equity in agentic AI deployment

  • Privacy-preserving approaches and federated learning for medical agents

  • Integration with EHR systems, FHIR standards, and clinical workflows

  • Scalable deployment, post-deployment monitoring, and real-world evidence

Important Dates

• Due Date for Full Paper Submission: September 27th 2026

• Notification of Paper Acceptance: October 18th, 2026 

• Camera-ready Paper Deadline: November 8th, 2026

Organizers