Share via

Integrating with DAX

Shivi 0 Reputation points
2026-03-26T18:05:38.2066667+00:00

Hi, I am looking for a way to see if my product can be built on top of DAX. Is there a way to share the ambient listening and have DAX return whatever it does return, and then have our tool return with information we return?

Nuance | Other
Nuance | Other

Additional Nuance products, services, or issues

0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. Q&A Assist
    2026-03-26T18:05:53.6333333+00:00
    AI generated content. This question contains an answer created with AI from Q&A Assist. Learn more

    Yes. The Dragon Copilot / DAX ecosystem provides several integration patterns that allow a product to work alongside DAX ambient listening and then consume what DAX returns.

    At a high level, the pattern is:

    1. Use an SDK or API to handle ambient recording and send audio to Dragon Copilot / DAX.
    2. Receive transcripts, notes, and structured data back via webhooks or retrieval APIs.
    3. Let the product process that returned content and present additional insights in its own UI.

    Key options from the documented components:

    1. Use Dragon Copilot SDKs and embedded apps for ambient recording
      • Ambient recording can be integrated via:
        • DAXKit SDK (iOS) or Ambient SDK (Android) to add recording and secure upload into an existing mobile app.
        • Dragon Copilot web app for desktop/mobile embedded in an EHR or other host app, which provides the full Dragon Copilot UI and capabilities.
        • Batch upload audio service (preview) if the product manages recording itself and then uploads completed audio for processing.
      • These options let the product “share” the ambient listening flow by either embedding Dragon Copilot or by sending audio to it programmatically.
    2. Consume what DAX / Dragon Copilot returns
      • Dragon data exchange (partner API) delivers AI output to an integration via webhooks. The payload can include:
        • Transcript of the encounter
        • AI-generated note (clinical summary)
        • Discrete clinical data and document sections
      • The product can subscribe to these webhooks, receive the JSON payload, and then:
        • Store it
        • Enrich it with proprietary logic
        • Display combined results (DAX output + product output) in its own UI.
    3. Additional capabilities that can be combined with a product
      • Feedback service: send user feedback on AI-generated content back to Microsoft, while also using the same feedback to tune product behavior.
      • License service: check user licensing from a mobile app.
      • Ambient session service: start and update ambient recording sessions programmatically.
      • Audio service (preview): upload and process audio that the product records.
    4. Data integration with Microsoft Fabric (optional but powerful)
      • If DAX Copilot is integrated with Microsoft Fabric, all DAX Copilot data (audio, transcripts, notes) is stored in OneLake, organized by date and encounter ID.
      • A product can then:
        • Read this data from Fabric
        • Run analytics or AI on top of it
        • Combine those insights with what DAX already produced.
    5. Planning the integration
      • Choose the primary integration surface:
        • Embedded Dragon Copilot web app if the goal is to host the full DAX/Dragon Copilot experience and layer product UI around it.
        • SDKs (DAXKit / Ambient SDK / JavaScript SDK) if the product needs tighter control over recording and UI.
        • Partner APIs (Dragon data exchange, ambient session, audio service) if the product wants a backend-to-backend integration and to own most of the UX.
      • Work with a Microsoft representative to establish a partnership and get non‑production environments and credentials for the partner APIs.

    In all of these models, ambient listening remains handled by Dragon Copilot / DAX (via SDKs or embedded app), DAX returns transcripts/notes/structured data, and the product can then add its own processing and present combined results.

    References:

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.