Most health system AI conversations are still focused on the first wave: which tools to adopt, which workflows to automate, which vendors to evaluate. Nasim Afsar, chief AI and analytics officer at City of Hope, is already thinking about the second wave — and the mess it could make.
“One of the most important things that I keep thinking about day in day out is how do we really orchestrate all of these different AI tools and, these days, AI agents that every organization is adopting for point solutions. We want to get ahead of it so that in a year or so, we don’t have tens, hundreds, thousands of agents running around the organization, each doing their own thing, and they cannot talk to each other,” Ms. Afsar said.
The problem is structural. As health systems deploy AI tools across departments and functions, each deployment tends to optimize locally — solving a specific problem in a specific workflow. That system can devolve into chaos.
“How do we orchestrate them to work effectively with one another and optimize the whole system rather than just one part of — a local optimization of we solve this problem, but then we shift the bottleneck somewhere else, and then we have to solve that other problem,” Ms. Afsar said.
The bottleneck-shifting dynamic is familiar to anyone who has managed complex operational systems. Fixing one constraint without visibility into the whole tends to create a new constraint downstream. In a clinical environment, where workflows are interdependent and the stakes of unintended consequences are high, that pattern is particularly costly.
City of Hope’s response is to treat agent orchestration as a governance and technology problem simultaneously — building the infrastructure now, before the proliferation gets ahead of the organization’s ability to manage it. That requires both technical architecture capable of connecting disparate tools and a governance model that provides visibility across the entire agent landscape.
For Simon Nazarian, executive vice president and chief digital and technology officer at City of Hope, the organizational ambition driving this work is explicit: scaling agency so that researchers, clinicians, and teams can build and run their own AI tools, with decisions made closer to the work.
“A major priority is scaling agency across the organization, but responsibly. It’s really important for us to empower our researchers, our clinicians, and our teams to be able to create and run their own agents,” he said.
Therein lies the tension. Distributed agency means more agents, faster. It means more point solutions built by more people across more workflows and it means the orchestration challenge gets larger and more complex.
“Most organizations are built for centralized permission. AI specifically requires distributed responsibility,” Mr. Nazarian said. “So how do we shepherd the organization and work together in order to go through that kind of change management?”
Once the right workflows and structures are in place, and technology integrated, it depends on the clinicians and frontline staff to really optimize AI.
“The technology is there for a lot of things that we want to accomplish, but does it really get adopted and used and make the impact? That’s the big question,” Ms. Afsar said.
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.