Introduction: The Next Wave of AI in Healthcare
In a recent conversation, Notre Dame’s New AI Project hosts Graham Wolfe and Aiden Gilroy sat down with healthcare investor and Notre Dame graduate Kevin O’Brien to explore the next frontier of artificial intelligence in medicine. The discussion moved beyond familiar generative AI tools like chatbots to focus on the rise of a more powerful and autonomous technology: agentic AI. This emerging category of AI is uniquely architected to execute complex, goal-oriented tasks, positioning it as a potentially revolutionary force capable of tackling some of the most persistent and costly challenges in the healthcare industry.
Distinguishing Agentic AI from Generative AI
Kevin O’Brien began by drawing a crucial distinction between the two types of AI. He explained that while generative AI is built to create content—such as text or images—based on specific, discrete user prompts, agentic AI is designed to act. It functions as an autonomous, goal-driven system that can independently perform complex, multi-step workflows without requiring human intervention at every stage.
To make this technical difference more intuitive, Aiden Gilroy offered a simple analogy: ordering a pizza. With a generative AI, a user might have to prompt it for each step: find a pizza place, list the toppings, provide payment information, and so on. In contrast, an agentic AI could handle the entire workflow from a single command like “order me a pizza,” autonomously making calls, placing the order, and providing payment and delivery details. This ability to manage an entire process from start to finish is what makes agentic AI a game-changer for process-heavy industries like healthcare.
The Immense Opportunity: Tackling Healthcare’s Administrative Burden
The most significant and immediate application for agentic AI, as detailed by O’Brien, is automating the immense administrative burden plaguing the healthcare system. He provided a concrete example: an AI agent handling a patient’s complex insurance verification in real-time. When a patient arrives at a doctor’s office, the agentic system could instantly query the correct insurance provider, drill down into the specifics of the patient’s plan, cross-reference it with the proposed medical services, and deliver a clear summary of coverage and out-of-pocket costs on the spot.
The scale of this opportunity is staggering. O’Brien cited a key statistic: administrative workflows account for an estimated 20-30% of the United States’ nearly $5 trillion annual healthcare expenditure. By automating these tasks, agentic AI can unlock massive efficiencies and reduce waste, freeing up capital not just for efficiency’s sake, but for a more equitable distribution of care—a challenge of alignment that O’Brien addresses later.
Navigating the Hurdles of Trust, Regulation, and Adoption
The discussion acknowledged that healthcare has historically been slow to adopt new technologies, often lagging about 10 years behind other sectors. This cautious pace is driven by major barriers, including the critical need to build trust with both practitioners and patients, and the challenge of operating within a heavily regulated environment governed by laws like HIPAA.
However, O’Brien presented a compelling counter-argument: these high stakes also create powerful “filters” and accountability mechanisms. Because patient well-being is on the line, best-in-class AI companies are forced to prove their value, safety, and compliance with incredible rigor. Furthermore, healthcare organizations themselves have an extremely high bar for vetting new technologies before deployment, as their own reputations and regulatory standing are at risk. These built-in guardrails do more than just mitigate risk; they provide a practical foundation for solving the much larger, more abstract challenge of “AI alignment.”
The Future of Healthcare Work and AI Alignment
The automation of the vast administrative tasks discussed earlier is not just about cost savings; it directly addresses the future of the healthcare workforce. While menial and repetitive administrative tasks are likely to be automated, this shift will create a new class of higher-skilled jobs. These emerging roles will require critical thinking, domain expertise, and the ability to design and analyze the complex workflows that AI agents will execute. The goal is to allow highly trained professionals to “work at the top of their license” rather than being bogged down by administrative duties.
Finally, the discussion grounded the abstract concept of “AI alignment” in the practical realities of healthcare. O’Brien explained that the industry already possesses robust, data-driven frameworks—such as population health analytics and the study of “social determinants of health”—that can be used to rigorously test AI tools for bias and measure their real-world impact. This provides a clear, established method for ensuring that AI produces positive and equitable outcomes for all patient populations.
Ultimately, the conversation revealed a powerful paradox: healthcare’s greatest weakness—its slow, risk-averse nature—may be its greatest strength in the age of AI. The high-stakes environment, while a barrier, forces a level of responsibility and accountability that makes it the ideal testing ground for developing safe, effective, and truly aligned artificial intelligence.
