AI medical scribe Epic integration 2026: HL7, FHIR, Hyperdrive, and what actually works
Epic Systems holds about 36% of US hospital EHR market and ~70% of academic medical centers. If you're shopping AI scribes for an Epic shop, the question isn't "does it work" — it's "which write-back path, and will Epic's app market gatekeep us." Here's the 2026 reality.
The four integration paths (in order of friction)
1. Manual copy-paste (no integration)
Provider reads scribe-generated note, copy-pastes into Epic SmartText or progress note. Works with any scribe, no Epic involvement. Steals 30–90 seconds per visit, defeating much of the time savings AI scribes promise.
2. Browser overlay (Hyperdrive companion)
Scribe app sits as a side panel in Epic Hyperdrive (Epic's 2024+ desktop framework, replacing Hyperspace). When user clicks "insert," scribe injects text via clipboard or accessibility API. Heidi Health, Suki, Abridge, Nuance DAX all support this in 2026. No Epic technical approval needed — treated as user-side automation. Best path for solo shops and small clinics.
3. App Orchard / SMARTcfg apps
Real Epic-blessed integration. Scribe vendor publishes app on Showroom (Epic's 2024 rebrand of App Orchard for SMART-on-FHIR apps). User launches scribe inside Epic context (patient ID, visit, encounter passed via SMART launch). Notes write back via FHIR R4 DocumentReference and Encounter resources. Requires Epic IT approval and a hospital-side license fee ($5k–$50k/year depending on size). Suki, Abridge, Nuance DAX, and Ambience all have Showroom presence in 2026; Heidi is on Showroom for select tiers.
4. Custom HL7v2 / Bridges integration
Old-school: scribe vendor sends HL7v2 ORU/ADT messages to Epic's interface engine (Bridges). Used in larger health systems where Showroom is overkill or where the vendor isn't in Showroom. Requires interface analyst time on hospital side, typically 4–12 weeks of work. Disappearing gradually as FHIR R4 becomes standard.
Which path each major scribe vendor offers (May 2026)
| Vendor | Hyperdrive overlay | Showroom app | HL7 / FHIR | Notes |
|---|---|---|---|---|
| Suki AI | Yes | Yes | FHIR R4 + HL7 | Strongest enterprise integration; $349/mo per provider list |
| Abridge | Yes | Yes (deep) | FHIR R4 | Often bundled with Epic Showroom enterprise deal |
| Nuance DAX Copilot | Yes | Yes | FHIR R4 + HL7 | Microsoft channel; M365 bundle pricing |
| Ambience Healthcare | Yes | Yes (2025+) | FHIR R4 | $243M Series C 2026; aggressive Showroom rollout |
| Heidi Health | Yes | Limited / select tiers | FHIR R4 | Strongest small-clinic story; Showroom path for enterprise |
| Augmedix | Yes | Yes | HL7v2 + FHIR | Older HL7 footprint, FHIR added 2024 |
| DIY (Whisper + Claude) | Manual paste, or browser extension you build | Not available (you'd need vendor partnership with Epic) | Possible via Epic's open FHIR R4 endpoints |
What "Showroom" actually buys you
- Single sign-on. Scribe launches with patient context already loaded; provider doesn't paste patient ID.
- Clean write-back. Note lands in the right encounter, signed by the right provider, billable from day 1.
- Audit trail. Epic logs every note origin; compliance officers see "AI-generated, signed by Dr. X" not "free-text mystery."
- HIPAA chain inherited. Epic's contract with the hospital includes the scribe vendor as a business associate via the Showroom listing; no separate negotiation.
Where Showroom doesn't help (and you still need legwork)
- Multi-tenant clinics on a single Epic instance. Scribe must respect tenant boundaries; not all vendors handle this without configuration.
- Custom note templates. If your group uses heavily customized SmartText/SmartPhrases, scribe output may not match. Pilot it.
- Specialty modules (oncology, cardiology, OB). Pre-built scribe outputs target primary care. Specialty-specific structures require either custom prompting or a specialty-tuned variant.
The decision tree for an Epic shop
- < 5 providers, no IT staff? → Hyperdrive overlay. Pick Heidi, Suki, or Abridge. Skip Showroom — no negotiating leverage.
- 5–25 providers, IT manager, 1 instance? → Pilot 3 vendors with overlay. If pilot wins, negotiate Showroom for the winner. Budget $10k–$25k/year app fee.
- 25+ providers, multiple specialties? → Showroom from day 1. Insist on FHIR R4 write-back, not just overlay. Negotiate volume pricing — vendors will discount 30–50% off list at this scale.
- Academic medical center / 100+ providers? → RFP with Suki, Abridge, Nuance DAX, Ambience. Demand SOC 2 Type 2, BAA, sub-processor list, FHIR R4 evidence in pilot. Don't sign before pilot in your specific Epic build.
Pitfalls in 2026 contracts
- Per-provider licensing creep. Vendor counts every Epic user, including residents and locums. Negotiate tiered or named-license pricing.
- Note-volume caps. Some Showroom contracts cap notes per provider per month. Confirm your typical volume fits.
- Multi-year lock-in with annual price increase. 7–12% annual escalation is standard but negotiable. Ask for cap.
- Data residency. Audio + transcripts; ask whether they leave the US, especially for vendors using Anthropic, OpenAI, or Google sub-processors.
- Termination clause. What happens to historical notes if you cancel? Insist on data export at no cost.
The DIY-on-Epic path (works, but rarely worth it)
Epic exposes FHIR R4 read APIs on Hyperdrive (read patient, encounter, observation). Write APIs are gated by hospital IT — you'll need a SMART-on-FHIR app registration, which generally requires Epic Showroom membership. Going DIY for write-back means custom HL7v2 work with the interface team, 6–12 weeks of integration. For most shops, this is more work than the cost savings justify.
Where DIY-on-Epic shines: a tech-comfortable solo provider on a small Epic instance (Epic Connect or Community Connect) who wants to use Whisper + Claude for the AI part and Hyperdrive overlay for write-back. $40–$100/month total cost vs $349 for Suki, identical UX.
Need clean transcript before write-back?
$0.05/min, BAA on request, no subscription floor.
Transcribe a file →