HIMSS26: The Risks and Rewards of Agentic Artificial Intelligence

Why It Matters

Agentic AI can free clinicians for higher‑value care, but without proper oversight it may compromise patient safety and regulatory compliance, reshaping how health systems manage automation.

Key Takeaways

• Agentic AI automates prior authorizations, scheduling, outreach
• Independent AI actions raise workflow risk without governance
• Human verification essential in AI decision loops
• Governance requires multidisciplinary stakeholders from start
• Target low‑value, high‑friction tasks for AI pilots

Pulse Analysis

Agentic artificial intelligence—systems that can act autonomously without constant human prompts—has moved from experimental labs into mainstream health‑tech roadmaps. Vendors now tout capabilities ranging from real‑time prior‑authorization processing to AI‑driven patient outreach bots that can triage symptoms and schedule follow‑ups. This shift aligns with broader industry pressures to curb administrative overhead, improve revenue cycle efficiency, and meet escalating patient expectations for digital interactions. As hospitals invest in these tools, the market is seeing a surge in pilot projects aimed at streamlining repetitive, low‑complexity workflows.

The autonomy that makes agentic AI attractive also introduces new layers of risk. Unsupervised decision pathways can propagate errors, breach privacy regulations, or generate unintended clinical recommendations. Consequently, health systems must embed robust governance structures that define clear guardrails, audit trails, and escalation protocols. Human‑in‑the‑loop verification becomes a non‑negotiable safety net, ensuring that AI‑generated actions are reviewed before they affect patient care or billing. Multidisciplinary oversight committees—comprising clinicians, informaticists, compliance officers, and IT leaders—are essential to balance innovation with accountability.

Strategically, organizations should prioritize AI deployments in areas where friction is high and clinical value is low, such as appointment reminders or claim denials. Starting with tightly scoped pilots allows teams to refine models, measure ROI, and build confidence across stakeholder groups. As governance matures, the scope can expand to more complex decision‑support functions, ultimately reshaping workforce dynamics and enabling clinicians to focus on direct patient interaction. Successful adoption will hinge on transparent policies, continuous performance monitoring, and a culture that views AI as an augmentative partner rather than a replacement.