Resources
Epic’s first arc of generative AI was focused on efficiency: how we could rapidly improve people’s existing work. That arc has reached widespread adoption. Today, more than 75% of our health system customers use gen AI. At The Christ Hospital, for example, it has helped oncologists treat nearly 70% of lung cancer cases at stage I or II, well above the national average. Northeast Georgia Health System has seen a 46% improvement in accuracy and efficiency of coding. Nurses at Mercy, UNC Health, and Houston Methodist opt to use more than 90% of the end-of-shift notes drafted by Epic AI. These are just a few examples of how generative AI is already making a practical difference.
The term electronic health record no longer reflects what health systems have in scope: the ability to use a vast array of multimodal data to guide patients, clinicians, and operational professionals through every step of the healthcare journey. These inputs include voice, sight, text, genomics and advanced diagnostics, medical device information, and much more. AI agents, Epic’s second arc of generative AI and one that is well underway, will interact not only with this information, but also with users, sharing insights and receiving guidance that allows them to act toward a prescribed goal. A platform like this is the next generation of healthcare technology.
Epic was built differently: not assembled from parts, but designed as a single, integrated system. Every application runs on the same codebase and writes to the same record for each patient. That design—one that gives the full picture to people, helping them work efficiently—enables efficient AI agents, too. If you were to design a codebase for agents from the ground up, it’d look a lot like ours.
Our early agents are embedded into existing workflows, providing a seamless experience for clinicians and their patients. They’re capable of organization-wide actions, drawing from an integrated action library that spans clinical, financial, and operational workflows. They’re intelligent, capable of learning from a health system’s local practices, user settings, and community to determine optimal actions. Users Group Meeting 2025 will bring much more on this front.
AI is moving fast, and before long, the distinction between AI and technology writ large will be irrelevant. The governance of AI, in turn, needs to be equally dynamic.
One of our solutions is our open-sourced AI Trust and Assurance Suite. It gives the healthcare community the ability to validate the newest AI—our models and anyone else’s—across diverse populations and real-world workflows. Most recently, the suite expanded to include a tool to assess AI-generated patient summaries.
Alongside that infrastructure, Epic staff and customers are advancing peer-reviewed research on generative AI in practice. In February, Epic staff and University of Wisconsin-Madison researchers published an investigation of AI-generated patient summaries in Nature; in April, Epic staff contributed to a National Academy of Medicine assessment of gen AI adoption in The New England Journal of Medicine; in May, researchers from Epic and the University of Wisconsin-Madison published an article in JAMIA validating the newly open-sourced tool for evaluating AI-generated patient summaries.
While we have challenges that need to be addressed and questions about how we’ll do it:
The name “Epic” comes from epic poems like the Iliad and the Odyssey. Part of what makes poetry special is the concept-to-word ratio. The richness of a poem’s language often shares more than its word count seems capable of capturing. Medical records—dense in information, spanning lifetimes—share this trait. And, like poems, the analysis of medical records can reveal meaning that might otherwise go unseen. Healthcare is heading into a new age of insights.