Epic Systems made news on several fronts this week at its annual Users Group Meeting, announcing new artificial intelligence capabilities for its customers.
These include tools such as Art for Clinicians, a clinical assistant; Emmie, a patient-facing chatbot; and Penny, a revenue management assistant.
The new agentic AI offerings are designed to improve provider operations, increase patient engagement and reduce administrative burdens for clinicians.
Epic is also developing new generative medical event models called CoMET to help doctors use real-world evidence to improve patient treatment and care decisions, the company said.
The largest electronic health vendor in the U.S. also made announcements about other platforms and discussed its research on developing generative medical event models at scale as it kicked off UGM 2025, called Sci-Fi, on Tuesday.
"We brought you a shipload of cool stuff," Dr. Jackie Gerhart, the company's chief medical officer, said in a social media post about the opening discussions at UGM 2025.
"This year’s flight path took us through healthcare intelligence made real, AI agents working side-by-side with clinicians and Cosmos AI and the promise of predictive care."
AI charting at scale
One of those experiences aims to integrate Microsoft's Dragon ambient AI technology, introduced in March, into Epic's AI Charting tool, explained Microsoft's Joe Petro, corporate vice president of health and life sciences solutions and platforms, in a social media post.
Putting a comprehensive clinical AI assistant into workflows at an Epic scale is a step toward democratizing AI. Epic says that it is in use at 3,620 U.S. hospitals and health systems, which accounts for about 38% of the inpatient electronic health record market share and contains patient records for approximately 325 million people.
The new scribe and clinical assistant, Art, could serve clinicians with information about what’s needed for an upcoming patient visit, provide summaries and answer questions.
Two years ago, Epic leveraged generative AI to give doctors draft responses to patient messages they could review, edit and approve and provided demos of "in-basket" at the HIMSS Global Conference & Exhibition.
Deploying AI agents in Epic platforms
Integrating AI across applications and positioning agents improve clinical efficiency, Seth Howard, executive vice president for R&D at Epic, said ahead of HIMSS25, where representatives from Epic spoke with Healthcare IT News about working on new agentic AI and the philosophy behind it.
"We expect AI agents to help with pre-visit prep by chatting with patients about their needs, identifying missing tasks (such as labs), helping schedule and complete those tasks, and creating an easy-to-read summary," he said.
According to Forbes, a new digital concierge, called Emmie, will appear in the MyChart patient portal. Patients can ask Emmie questions, including about test results. As a patient educator, Emmie can also engage patients before their appointments and help prepare them, according to the story. Emmie can also work with Art autonomously to streamline and inform patient outreach tasks on behalf of providers.
There's also Penny – an AI assistant that is already live and supporting revenue cycle management. Penny is the virtual go-to for provider employees who need support on billing codes and other aspects of claims and financial management for practices. According to CNBC's coverage of the EHR vendor's user conference, Epic CEO Judy Faulkner said Penny can generate appeal letters for insurance claims that get denied.
Predicting patient health
Teasing its ongoing development of clinical decision support technology, Epic reportedly shared further details about a predictive analytics milestone the company reached in the development of its Cosmos AI engine for health risk prediction.
Cosmos, launched in 2019, is an EHR analytics platform and a repository of de-identified patient data collected across Epic EHRs.
"Using Epic Cosmos, a dataset with medical events from de-identified longitudinal health records for 16.3 billion encounters over 300 million unique patient records from 310 health systems, we introduce the Cosmos Medical Event Transformer (CoMET) models, a family of decoder-only transformer models pretrained on 118 million patients representing 115 billion discrete medical events (151 billion tokens)," said researchers from Epic Systems, Microsoft Research, Yale School of Medicine and the Cosmos Governing Council in a new report released over the weekend.
They noted Cosmos was created to use real-world data to generate real-world evidence at scale, but "answering a single clinical question requires crafting custom cohort definitions, feature-engineering pipelines and statistical analyses."
"To enable personalized medicine and RWE at scale for routine clinical decision-making, we need tools that can learn from the integrated patient record and can answer complex medical inquiries with flexibility, retrieving the right RWE to support clinical decision-making across a wide variety of contexts," they said.
The CoMET models, pretrained on Cosmos data, are "producing to date the largest scaling-law study on real-world patient journeys," some "three orders of magnitude larger in terms of the number of patients," according to the report.
The models could ignite the way doctors use real-world evidence to improve patient treatment over their lives, according to Epic. The researchers not only tested the models' abilities to diagnose and predict, but they said they also found success in scaling them.
"We show that scaling up model and dataset size predictably decreases training loss and that minimizing train loss consistently improves downstream evaluation scores."
How doctors might use the models
Epic's new foundational AI models encompass a diverse range of patient populations and health conditions, as noted by the researchers in their paper. The diversity of real-world evidence that healthcare AI tools rely on is essential for minimizing bias and engendering trust.
Previously, at HIMSS25, Gerhart sat down to share insight into how doctors could generally use AI in the future to make better care decisions during a larger conversation about Epic's more agentic developments with tools like MyChart and Slicer Dicer.
"When I think about what AI can do for medicine, efficiency is the low-hanging fruit," she said in an interview with Healthcare IT News in March. "But what excites me is that it'll help me understand a differential that I didn't think of."
At that time, Gerhart explained, she was working on a software project called Diagnosis Advisor that would enable physicians to create a precision cohort around patients.
"Let's say you're my patient, and you come in and you have heart failure and diabetes and kidney disease," she explained. "I might know what to use for heart failure and what to use for diabetes, and what to use for kidney disease. But I can search Cosmos for the group of people that have those three diseases and see what's actually worked best in those people."
Gerhart called the medication prescribing support tool Best Care Choices for My Patient and the diagnosis assistant Diagnosis Advisor.
In her example, the group of people that she would hone in on as a doctor using the Cosmos AI engine, all have five other features that are gathered from all areas of their chart and pooled together, and they point to something that other people were diagnosed with earlier.
"It'll make me think, 'Maybe I should think, pancreatic cancer in this patient, even though they don't have jaundice and abdominal pain.'
"It is a cool time to be in medicine," she said.
Andrea Fox is senior editor of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.
