Health systems are beginning to deploy autonomous artificial intelligence agents into clinical workflows, with early implementations at Advocate Health and ECU Health illustrating both the opportunities and operational requirements for adopting the technology. Built using Epic’s no-code Agent Factory platform, the AI agents have been designed to automate administrative tasks while remaining fully integrated within the electronic health record environment.
Industry experts have cautioned that successful deployment depends on more than financial investment. Healthcare organizations require robust data governance, security, compliance processes and AI expertise to support implementation, while ongoing monitoring is needed to address issues such as data drift that can affect model performance. Organizations pursuing AI without these foundations may struggle to realise meaningful value.
Advocate Health has introduced autonomous agents supporting inpatient pharmacy operations and infusion chart preparation. The pharmacy agent reviews a patient’s medical history, medications, clinical activity and progress notes to generate concise summaries that assist pharmacists in verifying medication orders. Additional agents automate elements of infusion scheduling and patient communications through MyChart, helping prepare patients before treatment while reducing administrative workload for clinical staff.
ECU Health, a regional academic health system in North Carolina, has deployed two Epic-developed AI agents focused on patient transfers and discharge planning. One agent generates concise summaries for transfer centre nurses, reducing the need for manual chart review. According to the health system, the tool has saved approximately 20 hours of chart review each week while producing no hallucinations since deployment. A second agent supports multidisciplinary discharge planning by summarising patient records ahead of progression-of-care rounds, allowing case managers to focus discussions on barriers to discharge rather than reviewing extensive documentation.
Both organisations emphasised the importance of clinician feedback during implementation. ECU Health refined its discharge planning agent immediately after launch when users indicated that summaries were too lengthy, resulting in substantially improved user satisfaction following prompt adjustments. The experience highlights how AI agent deployment requires continuous optimisation alongside technical integration.
The early projects demonstrate how healthcare providers of different sizes can begin incorporating autonomous AI into frontline care delivery, while reinforcing that governance, workflow design and clinical oversight remain central to achieving sustainable adoption.
Click here for the original news story.
