Agentic AI is taking root across health systems faster than the governance structures built to contain it. The technology questions, for the most part, are settled. What remains unresolved is whether the people responsible for operations understand they are now managing a digital workforce.
Crystal Broj, enterprise chief digital transformation officer at the Medical University of South Carolina in Charleston, S.C., sees the AI agent workforce recognition as pivotal to health system transformation. She leads digital transformation across the health system, university and research enterprise, work explicitly tied to MUSC’s stated ambition to rank among the top 20 health systems in the country and help South Carolina reach the top 20 states in health outcomes.
Over the past several years, MUSC has built an expanding AI agent infrastructure that automates prior authorizations, handles patient check-ins and manages inbound calls at scale. That operational foundation has made the next challenge more visible. Health systems don’t need more agents; they need a more coherent way to deploy, connect and govern the ones already being built.
“With the rise of AI agents, we often talk about how eventually, the agents in the department are going to be listed on the org chart,” Ms. Broj said during an episode of the “Becker’s Healthcare Podcast.” “They’re doing so much of the work for us, but it has to be really seamless.”
The shift is architectural as much as operational. Today’s AI agents in most health systems operate in silos. They are purpose-built for specific functions, unable to hand off work cleanly from one to the next. The friction surfaces in everyday workflows, such as a physician placing a referral, and the patient, instead of being immediately routed to an appointment, lands in a queue and plays phone tag trying to schedule. That’s not a single agent’s fault. The problem is structural because existing systems were never designed to communicate.
“I want [AI agents] to be able to all talk together so I don’t have to repeat myself, much like I’d like my staff to work together so I don’t have to repeat myself,” Ms. Broj said. “There’s a lot of tools that are coming a long way. Personally, I even use voice agents that convert things to text then actually help me plan for the next day. Having that same thing in a provider’s hand or back office hand where we connect voice agents and documentation, and intelligent workflow routing, is something that we’re going to see a lot more of with agents talking to each other. It is going to be about how humans work well with AI.”
MUSC is working to address this directly, consolidating platforms and moving toward a model where agents are embedded in workflows and route work without requiring human mediation at every handoff. The underlying logic is that agents requiring people to connect them haven’t reduced friction; they’ve just relocated it.
The governance challenge sharpens considerably as the ability to build agents reaches more of the organization. MUSC is deploying Microsoft Copilot across its entire workforce and giving staff the ability to create their own tools. The move accelerates capability. It also surfaces a problem that most health systems haven’t answered.
“How do we govern that so that you and I don’t write the same agent?” Ms. Broj said. “That’s not necessary. I could borrow your agent if it does the same or adapt your agent. How are we going to make these global agent libraries?”
The concept of a “global agent library,” a shared repository where purpose-built agents can be discovered, reused and adapted, addresses a problem that compounds as more employees gain the ability to build. Without one, organizations risk duplicating effort, fragmenting governance and accumulating agent infrastructure that scales in number without scaling in coherence.
“It’s really the human centered leadership and how we show people how to use AI in a safe, effective way and really be thoughtful about AI adoption,” said Ms. Broj. “The organizations that do this well are really going to define what the future of healthcare delivery is going to be like.”
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.