At HIMSS26, Epic introduced its Agent Factory, a no-code platform that allows hospitals to build AI agents directly inside their electronic health record (EHR) systems. This is not just another software update; it marks a significant pivot in how enterprise technology is being built. Instead of relying on generic, horizontal AI tools that sit on top of existing software, Epic is moving toward a vertical-native model where the infrastructure itself is designed to be an active participant in clinical work.
With its software already installed in over 3,700 U.S. hospitals and managing records for more than 325 million patients, Epic is in a unique position to redefine the EHR. The goal, as envisioned by Judy Faulkner, is to transform the EHR from a passive repository of patient data into a predictive space. By embedding agent creation into the very environment where high-stakes medical decisions occur, Epic is attempting to move from simple data storage to real-time, active guidance for clinicians.
The shift here is from assistive AI—like the chatbots we are all familiar with—to what Epic calls collaborative AI. Phil Lindemann, Epic’s VP of Data and Research, describes this as enabling workflow-level thinking. These agents are designed to handle complex, multistep processes rather than just summarizing text or suggesting a single sentence. This is a move toward maturity in enterprise AI, where the software does more than just assist; it collaborates on the actual work.
We are already seeing the impact of this approach. Art, an agent designed for clinicians, has improved the speed of nursing documentation by 85% across 300 organizations. On the administrative side, an agent named Penny has helped reduce coding denials by 20%. These are not just abstract metrics; they represent a tangible reduction in the administrative burden that often overwhelms healthcare workers.
Early adopters are seeing real-world results. At ECU Health, agents deployed in the transfer center and for discharge planning are saving staff roughly 20 hours per week. Meanwhile, Advocate Health is preparing to launch agents for inpatient pharmacy and infusion chart preparation. These deployments show that when AI is integrated into the workflow, it can reclaim significant time for staff who are otherwise buried in paperwork.
Epic’s ability to execute this strategy relies on a massive foundation of data. Through Epic Cosmos, which aggregates 16.3 billion encounters from over 310 health systems, and the CoMET foundation models—which represent the largest scaling-law study on real-world patient journeys—the company has a proprietary data advantage that horizontal competitors find difficult to replicate. While platforms like Kore.ai or Microsoft Copilot Studio offer powerful horizontal capabilities, they often lack the deep, native integration into the clinical workflow that Epic provides.
This no-code approach also changes the power dynamic within hospitals. By making it possible for clinical domain experts—rather than just centralized IT departments—to build and customize these agents, the pace of innovation could accelerate. If a nurse or a department head can build an agent to solve a specific bottleneck, the hospital becomes more agile.
The market for healthcare agentic AI is expected to grow from $3.9 billion in 2026 to $24.6 billion by 2036, representing a compound annual growth rate of 21.5%. Because healthcare is so highly regulated, it is becoming the primary proving ground for compliance-by-design agent infrastructure. The tools developed here, supported by initiatives like Epic’s open-source AI Trust and Assurance Suite, may eventually set the standard for how other industries deploy autonomous agents.
However, there is a significant caveat to this progress. As agents move from suggesting notes to executing multistep clinical processes, the margin for error narrows. The industry still needs to prove that these systems can maintain safety and accuracy at scale, particularly when they are operating on the most sensitive patient data in the world.
The rise of vertical-specific infrastructure suggests that the future of enterprise AI is not just about having the smartest model. It is about embedding that intelligence exactly where the work happens. Whether these agents can reliably handle the complexities of clinical life remains to be seen, but the shift toward native, workflow-integrated AI is clearly underway.