Dragon Copilot Integration With Epic: What Is Actually Embedded

The build uses Epic's ambient documentation framework, the same module other Toolbox vendors plug into, but with the additional depth of a first party partnership. The clinician selects the patient in Haiku or Canto, starts recording, and Microsoft's servers return a draft note into the encounter. The Hyperdrive build, which requires Hyperspace February 2025 or later and Haiku or Canto 11.5.3 or later on iOS and 11.6.5 or later on Android according to Microsoft's release notes, adds desktop recording and note review.

Microsoft's FAQ is specific about what sits on your side. Your Epic analyst must enable the ambient security points listed in Epic's Sherlock checklist, configure a user allowlist where several contributors edit the same encounter note, design note templates compatible with DAX SmartSections, and set a primary specialty in each user profile so the right model is selected. Processing happens server side; Wi-Fi is needed to select the patient but recording continues without connectivity. Session data, meaning audio, transcript and summary, is retained for 30 days and then deleted, per Microsoft's documentation.

Patient context comes from Epic, so this is not a FHIR app your integration team writes to. It is a partner feature you switch on. For the general shape of Epic's paths see our EHR integration questions guide.

In the mobile apps, Dragon Copilot supports ambient recording and gives access to Note, Orders and Transcript screens, per Microsoft's Epic mobile app article. The note lands in your templates through SmartSections. Newer releases, described in Microsoft's Dragon Copilot 3.x release notes, add diagnosis aware notes that structure assessment and plan content around the visit diagnoses already in Epic, and send that structure back to the chart.

Orders screens are the feature to review before enabling. A transcript that proposes orders for acceptance is decision support, and your governance process should look at it as such. Our governance committee playbook covers the staging. The documentation loop itself, recording to draft to signature, is the low risk part and the part clinicians feel first, which is why ambient documentation is usually the first Epic AI project; the AI scribe use case page describes what a good first ninety days looks like.

• The Epic version floor. If your Hyperspace release predates February 2025, the Hyperdrive build is not available until you upgrade; mobile only may still be possible. Check with Epic technical services.
• Ambient security points and the user allowlist, per the Sherlock checklist Microsoft references.
• SmartSection compatible note templates for each specialty in scope.
• Mobile device management: Haiku and Canto minimum versions, and microphone permissions on personal phones.
• The business associate agreement with Microsoft, and the retention and subprocessor terms in it. The HIPAA and AI page lists what to insist on.
• A change control slot and named validation clinicians.

Sites that already run Dragon Medical One often assume the ambient rollout is a licence change. It is not; it is an Epic build change with a queue behind it. Ask your Epic lead how long that queue is before you promise clinicians a date.

Price transparency is the first. Microsoft does not publish per user pricing for Dragon Copilot and sells through direct enterprise agreements and resellers, often bundled with an existing Dragon contract. Our Dragon Copilot pricing page sets out what buyers report and what to ask for, and the scribe pricing comparison puts it beside the alternatives.

The second is version gating. Because features arrive with Epic releases, a site behind on upgrades gets a narrower product than the demo shows. Ask which of the features you saw require which Epic release.

The third is that the market has caught up. Abridge is embedded under Epic's Workshop programme, and Ambience, Nabla and Suki sit in Epic's Toolbox for ambient voice recognition. Dragon Copilot's advantages are the Microsoft relationship, the Dragon dictation heritage and the documentation quality; they are real but no longer unique, so a comparison on your own encounters is worth running. See the AI medical scribe comparison.

• Which Epic release and mobile app versions does the build we are buying require, and which features are gated by version?
• What is the per user price at our volume, what is the minimum term, and how is it affected by our existing Dragon agreement?
• Is the 30 day retention in the documentation contractual, and are any recordings or transcripts used to train shared models?
• Who supports what: Microsoft, Epic technical services or our own analyst, and where is that written down?
• What does the implementation plan look like and how many Epic analyst hours did a comparable site spend?
• Can we speak to a reference site on our Epic release without the vendor on the call?

Search demand for "dragon copilot implementation partner" is high because sites underestimate the last two questions. Our vendor HIPAA question list covers the compliance side.

A Dragon Copilot decision is usually a licensing decision inside a much larger Microsoft relationship, which means the people negotiating it are not always the people who will use it. An independent review separates the two: we check the Epic build against your actual release, run a comparison on your clinicians' encounters against at least one other embedded vendor, and give you a number to negotiate with. We take no reseller or vendor commissions.

Our vendor selection service runs that comparison, and our deployment roadmap sequences the Epic work once you have chosen. Book a call with your Epic version and your current Dragon contract to hand.