Agentic AI is moving beyond standalone models in healthcare. These models are starting to function as “digital co-workers” that orchestrate tools, data and workflows to take on tasks once handled exclusively by clinicians and researchers, according to one health tech expert.
Kimberly Powell, general manager of healthcare at Nvidia, believes this shift could rapidly reduce burnout and expand clinical capacity, but only once health system leaders stop viewing AI as software and start treating it as hireable labor.
“Thinking of these AI agents as hireable is the concept — I think that hasn’t completely clicked in the C-suite yet. They don’t see it as that — they see it as technology at the moment. They don’t see it as, you know, employees,” she said during an interview last month.
Powell pointed out that healthcare is fundamentally workflow-driven, which makes it well suited to agentic AI systems.
Agents are designed to mirror the step-by-step workflows clinicians already follow, coordinating models and tools to get real work done instead of simply serving as passive software, she explained.
Unlike older software designed as tools for humans, agents are designed to do the work themselves. Modern software architecture has made this possible, Powell noted — she said that advancements like APIs and tool calling mean that agents can interact with legacy systems, access only the data they need and operate within strict guardrails.
In her eyes, guardrails turn risky models into production-ready systems — she pointed out that AI agents only succeed by surrounding foundation models with domain expertise, safety constraints and up-to-date medical knowledge.
This approach reduces hallucinations and ensures regulatory compliance, which gives hospital leaders peace of mind that the AI tools can safely operate in clinical environments, Powell explained. With this assurance, adoption is accelerating quickly.
While “agents” may still sound like a buzzword, Powell said the technology has crossed into real production use, with some healthcare AI companies already scaling rapidly and offloading administrative tasks and clinical documentation burdens — she highlighted Abridge and Multiply Labs as examples.
She expects this trend to develop even more quickly over the next two years.
Powell argued that agentic AI holds great potential to augment understaffed teams. In the next couple of years, she thinks agentic tools will increasingly be treated like employees that elevate clinicians’ practice and improve access to care.
Photo: wildpixel, Getty Images
