If you run a US health system on Epic, you have access to four native AI products that didn't exist 18 months ago. Here's the honest landscape as of HIMSS 2026.
ART — clinician-facing AI Drafts progress notes from ambient listening, summarizes patient charts pre-visit, queues orders from visit conversation. The ambient scribe is GA but adoption is uneven — Penn Medicine, Mayo, and Riverside have published outcome data; many systems are still in pilot. Reported gains: 20-30% reduction in time on discharge summaries (Riverside), measurable reduction in after-hours EHR work.
PENNY- revenue cycle AI Drafts prior auth responses, generates appeal letters for denials, suggests billing codes. Numbers Epic publishes from Summit Health: 42% reduction in medication PA submission time, 92% of AI-drafted responses accepted without edits. At systems most actively using Penny, coding-related denials are reportedly down >20%.
EMMIE - patient-facing AI inside MyChart Answers patient billing and care questions, schedules appointments, suggests relevant screenings. Rush University Medical Center has published a sustained 58% reduction in billing-related customer service messages.
AGENT FACTORY — the platform layer Announced at HIMSS 2026. Lets health systems build, monitor, and govern custom AI agents that span Epic workflows. The strategic shift. Until now, Epic AI was features. Agent Factory positions Epic as the agent runtime for the enterprise.
What this means for digital strategy:
The build-vs-buy line just moved. Two years ago, prior auth automation was a custom RPA project with vendors like Olive AI or Notable. Today, if you're an Epic shop, Penny is the default starting point for medication PA. Custom builds need to justify themselves against a baseline that didn't exist before.
The integration tax dropped -but only inside Epic. Penny works because it lives in the same data plane as your charts, orders, and billing.
Validation is now your responsibility, not Epic's. Epic's AI Trust and Assurance Suite ships an open-source framework for local validation, but proving Penny's outputs are accurate for YOUR payer mix, specialty mix, and documentation patterns sits with your health system. Most CIOs haven't budgeted for this.
What I'd ask before assuming Epic AI is enough:
→ For Penny: which of our top 10 payers does PA automation cover end-to-end vs. which still require manual portal submission?
→ For Art: does our specialty mix (especially behavioral health, oncology, complex pediatrics) match the populations Epic's models were validated against?
→ For Emmie: what happens to messages it can't answer? Is that escalation path measured and SLA'd?
→ For Agent Factory: who in our org owns agent governance? CTO function, CMIO function, or new headcount?
Want the full Epic AI feature audit template I use with health system clients? Comment AUDIT and I'll DM the spreadsheet.
