INTEGRATION DEEP DIVE · EPIC · 7 min
Epic AI Scribe Integration: What "Epic-Integrated" Actually Means
Every ambient AI scribe vendor claims "Epic integration." The phrase covers everything from copy-paste-the-draft-yourself to a full Epic Pal program partnership with bidirectional write-back into Haiku, Limerick, In Basket, and the encounter chart. This guide explains the tiers that hide behind the same phrase — and which vendors actually sit where.
A small set of AI vendors with the deepest Epic-sanctioned integration. Abridge was the first.
Copy-paste, browser sidecar, Pal-program in-EHR, full Epic-sanctioned partner with write-back to multiple modules.
The Epic mobile / web surfaces where ambient capture matters most for many ambulatory and rounding workflows.
Routing the draft into the clinician's In Basket workflow rather than only the encounter — separates deep integrations from shallow ones.
The four integration tiers
"Epic integration" can mean any of these. The procurement RFP should require the vendor to specify which tier they sit at, because the operational implications differ substantially:
| Tier | What it is | Clinician experience | Operational cost |
|---|---|---|---|
| Tier 1 — Copy-paste | Vendor produces a draft note in their own app; clinician copies it into the Epic note field. | Manual context-switching. Friction. | Lowest. No Epic configuration required. |
| Tier 2 — Browser sidecar | Vendor renders a panel alongside Epic via browser SSO; one-click "insert draft into note" via clipboard or simple API. | Less switching; still some clipboard mediation. | Low. SSO + browser extension; no Epic-side integration. |
| Tier 3 — Epic-app (in-EHR) | Vendor app embedded inside Epic via the Epic App Orchard / Connection Hub. Reads encounter context; writes draft into the note field via Epic APIs. | One workflow inside Epic. Draft appears in place. | Medium. Epic-side configuration. App Orchard certification cycle. |
| Tier 4 — Pal program partner | Epic-sanctioned deep integration. Bidirectional write-back across multiple modules — encounter note, In Basket, problem list, orders, mobile surfaces (Haiku / Limerick / Canto). | Native-feeling experience across the full Epic clinician workflow. | Highest. Multi-quarter integration. Epic relationship required. |
Which vendors sit at which tier
Approximate placement based on the WalledCare directory's vendor profiles. The exact tier varies by customer deployment configuration; what follows is the typical default for new customers:
| Vendor | Typical Epic tier | Notable surfaces |
|---|---|---|
| Abridge | Tier 4 — Pal | First Epic Pal. Abridge Inside in encounter; Abridge Inside for Inpatient; In Basket integration for prior auth. |
| Microsoft Dragon Copilot | Tier 4 | Nuance heritage — "fully embedded in Epic" per Healthcare IT News coverage. Microsoft 365 surfaces beyond Epic. |
| Nabla | Tier 3-4 | Epic Haiku integration documented. 6 EHRs total; Epic is the deepest. |
| Ambience Healthcare | Tier 3-4 | Strong Epic integration with AutoRefer write-back; Oracle Health support. |
| Suki | Tier 3 | Epic plus 6+ other EHRs; voice-command surface differentiates. |
| DeepScribe | Tier 3 | 9 EHRs supported; bi-directional write-back on Epic, athena, Cerner. |
| Commure Ambient | Tier 3-4 | Inherits Augmedix Epic depth; HCA Healthcare reference indicates Pal-level operational maturity. |
| Heidi Health | Tier 2 | Browser + lightweight push; strong clinician-led adoption but not Epic-embedded. |
| Freed | Tier 1-2 | Browser-first; athenahealth integration deeper than Epic. |
| Augmedix (legacy) | Tier 3-4 | Long Epic operating history; now part of Commure Ambient. |
What the Epic Pal program signals
Epic's Pal program (formally introduced in the early 2020s and expanded for ambient AI in 2024-25) is Epic's tightest third-party-AI integration tier. Pal status involves:
- checkEpic-validated technical integration. Pal partners pass Epic's deeper validation cycle covering security, performance, workflow placement, and data-handling defaults.
- checkJoint customer-facing positioning. Pal partners appear in Epic's customer-facing AI landscape materials and at Epic conferences (UGM, XGM).
- checkMobile surface access. Pal-tier integration unlocks deeper access to Epic Haiku (iPhone), Limerick (Apple Watch), and Canto (iPad) than non-Pal third-party apps typically receive.
- checkIn Basket and multi-module write-back. The technical capability to drop drafts not just into the encounter note but also into the clinician's In Basket workflow, the patient communication queue, and (in some configurations) prior-auth and orders surfaces.
- closeNot a guarantee of clinical quality. Pal status indicates integration depth, not note quality. The published evidence (UCLA NEJM AI RCT) treats integration tier and time-savings as different variables.
Questions to ask vendors about Epic integration specifically
- checkAre you an Epic Pal? If yes, since when. If no, what is the path to Pal status and what is the timeline.
- checkWhich Epic versions do you support? Epic is on a major-release cadence (May / August versions). Lag in vendor support of new Epic versions is common; can produce operational gaps for hospitals on the leading edge.
- checkDo you write back to In Basket? The differentiator between an encounter-note tool and a clinician-workflow tool. Vendors that integrate with In Basket can route prior-auth drafts, patient-communication drafts, and referral letters into the clinician's existing follow-up queue rather than leaving them in a separate app.
- checkDo you support Haiku, Limerick, and Canto? Mobile and tablet surfaces matter for hospitalist rounding, ED workflow, and outpatient settings where the clinician is not at a desktop.
- checkWhat is the integration timeline at our specific Epic build? Even a Pal-tier vendor typically requires several weeks to several months of customer-specific Epic configuration. Get the specific timeline in writing, with the gating dependencies named.
- checkWhat happens when Epic upgrades or you upgrade? Both sides upgrade. The contract should address compatibility maintenance and the SLA for restoring functionality after either party's upgrade.
What "deep Epic integration" does not solve
Three categories of pilot failure that look like integration problems but aren't. Most procurement committees diagnose them wrong because the deeper-integration vendor seemed like the safer bet:
- closeClinician adoption. Tier 4 integration produces a smoother experience but does not by itself produce adoption. The deeper the integration, the easier it is for clinicians to not use it without anyone noticing — versus a separate app where non-use is visible.
- closeNote quality. The integration tier and the note quality are largely independent. The UCLA NEJM AI RCT (Nabla −9.5% vs DAX −1.7%) showed two Tier 3-4 integrations with substantially different time-savings results. Don't infer quality from integration depth.
- closeCompliance posture. Tier 4 integration depth does not change residency, retention, or training-data defaults. The privacy review applies the same way whether the integration is browser sidecar or full Pal.
Where this fits in the WalledCare directory
This guide pairs with the vendor side-by-side comparison (EHR depth column), the RFP questions checklist (EHR integration section), and the smaller-EHR integration guide for the non-Epic equivalent. The AI Scribes category guide covers the workflow-level evidence above the EHR-integration layer.