Epic used its 2026 Users Group Meeting in Verona, Wisconsin, to tell health system CIOs that the EHR is no longer just where clinical data lives - it's where the next wave of automation will be built, deployed, governed and measured. With the company fast-tracking agent tools, Cosmos data access and new APIs, the message to IT leaders is clear: move from pilots to platform adoption, and update your governance plans accordingly.
Epic founder and CEO Judy Faulkner said the company tracked 167 projects announced at UGM 2025 and delivered 84 on time, with seven completed and headed for special updates and 76 in progress and on track. "What's announced at UGM is very likely to happen," Faulkner said, according to Healthcare IT News.
That delivery cadence is the operational story. When the core EHR vendor ships AI features at conference tempo, the bottleneck shifts from vendor capability to validation, change control, clinical governance, and training capacity - all of which sit on the health system side.
Epic's AI stack consolidates around Ergo, Art, Emmie, and Penny
Epic positioned Ergo, an AI platform first announced in 2025, as "here" this year. The platform starts with human judgment and uses generative AI to find patterns and iterate quickly, built on Epic's data assets and workflows, according to Healthcare IT News.
Ergo's foundation runs on Cosmos, Epic's analytics and research platform. Epic said Cosmos is available to support 320 million deidentified patient records drawn from 23 billion patient encounters. Fierce Healthcare, citing an Epic spokesperson, separately reported than 325 million patients have medical records in Epic, which is installed in more than 3,700 hospitals.
Epic's copilots are designed to work as a system: Art helps clinicians, Emmie is a patient-facing assistant in MyChart, and Penny targets revenue cycle work, all presented as components that can function together as enterprise AI for healthcare, Fierce Healthcare reported.
Adoption data was a key proof point. Healthcare IT News reported that more than 60 organizations are using Chart with Art, and that clinician and nurse summaries generated by Art are deployed in more than 300 health systems. In an Epic community survey, 68% of respondents said visit summaries containing Care Everywhere notes surfaced new patient details they would otherwise have missed, according to Fierce Healthcare.
With Agent Factory, 2027 planning starts now
The biggest forward-looking takeaway for IT leaders was Epic's Agent Factory. Forbes reported that Epic is framing it as a no-code way for health systems to build their own AI agents, with broad availability targeted for 2027. The message for platform gravity: keep innovation inside Epic instead of building external tools that integrate through APIs and additional third-party contracts.
That changes budgets and operating models. If agent development happens "in the platform," line items shift from third-party subscriptions to internal capacity planning - who builds agents, who tests them, how token or usage costs are allocated, and who signs off when a locally-built agent changes clinical workflow.
Forbes framed the tradeoff for CIOs as timing and ROI uncertainty. Even if no-code reduces technical lift, health systems still have to account for staff hours, governance overhead, and model usage economics as agent concepts move from pilot to production. Those costs now hit this or next year's operating plans.
Revenue cycle pulls into the same train
Epic's AI story at UGM wasn't confined to the clinical side. A prior authorization API called Coverage Requirements Discovery (CRD) lets clinicians check payer requirements inside the EHR at order entry, rather than leaving the workflow for external portals or manual lookups, per Healthcare IT News. Ochsner Health is an early adopter, with additional implementations at Froedtert ThedaCare Health, Denver Health, and Summit Health. Payers involved include UnitedHealthcare, Network Health, and Aetna, with 16 payers testing CRD.
For revenue-cycle leaders, CRD is a measurable change, not a vision statement. If payer rules can be resolved without leaving the order-entry screen, the operational target becomes fewer rework loops - fewer staff touches to discover requirements, fewer incomplete orders that return, and fewer avoidable denials.
Epic also pointed to operational impact data from existing AI assistants. Ochsner reported Ask Emmie saved more than 1,000 call center staff hours over 10 months, and Community Health Network said billing-related Respond inBasket messages dropped 73% after going live in July. Tampa General, Ochsner, Penn Medicine, and Sutter Health have integrated Ask Art, with Sutter first to live with Ask Emmie in MyChart, per Healthcare IT News.
AI adoption is moving from clinician productivity stories to contact center throughput and message-volume management. That's where AI's cost case can be audited inside existing service-level metrics - and where deployment tends to move faster.
Epic also addressed security as AI expands. Chief Security Officer Stirling Martin emphasized planning for patching and said Epic is participating in Project Glasswing, an AI cybersecurity initiative run by Anthropic with about 150 organizations worldwide, as reported by Healthcare IT News.
Security and risk teams will be asked to approve model behavior, data access, and turning - not just features, but the speed at which AI-dependent features roll into production.
Why this matters for healthcare operations leaders
Feeds delivery tempo is set. If your organization plans an Epic upgrade in the next two years, map governance before the next release: assign owners for clinical validation, model behavior, and change control before features arrive in standard releases. That's decision capacity, not a vendor problem.
The Agent Factory and CRD timelines imply you should start scoping internal "agent product owner" capacity, test environments, and cost allocation for model usage now, ahead of the wider 2027 expectations - not after they're announced in your locality. For revenue cycle and contact center leaders, tie patient-facing AI to KPIs like call deflection or InBasket reduction, as our own organizations have done with Emmie and billing messages. If you're planning similar implementations, AI Learning Path for CIOs covers the decision points for governance, rollout, and cost. And for broader context on medical AI deployments, the AI for Healthcare tag has ongoing coverage of health system adoption.
Your membership also unlocks:
