Canary Ambient Now Generally Available in Dragon Copilot

Ambient documentation solved a real problem. The microphone is on, notes are generated, and clinicians get time back. But the conversation being captured has the potential to provide so much more than a note-it can turn the conversation into a source of objective, clinically relevant health signals.

That is the gap Canary Speech closes: others capture the conversation, Canary extracts the signal.

And as of August 21, 2026, bringing that capability into Dragon Copilot no longer requires a custom integration project. Microsoft announced the general availability of Dragon Copilot AI Apps and Agents for physician workflows, with Canary Speech named in the launch portfolio for cognitive and behavioral health screening.

Read Microsoft's announcement

Canary Ambient for Physicians is now generally available through Microsoft Marketplace and can be deployed within a health system's Dragon Copilot environment with the appropriate licensing and implementation support from Canary Speech.

What Actually Changed

Canary and Microsoft have been building toward this for a while. The collaboration was announced in March 2025, Canary was featured in Microsoft's first Dragon Copilot partner ecosystem post in October 2025, and the technology was demonstrated live in the Microsoft booth at HIMSS26.

General availability is a different kind of milestone. It is the difference between a technology that can work inside a clinician's workflow and a product that health systems can evaluate and procure through an established enterprise channel.

Practically, three things are now true:

1. Discovery and procurement are available through Microsoft Marketplace. Canary Ambient for Physicians is a listed offer, giving health systems a familiar channel for evaluating and procuring the solution. Enablement of the Canary solution still requires engagement with the Canary team. Organizations interested in using Canary Ambient should contact the Canary Speech team at [email protected] to obtain the appropriate license and access.
   
   
2. Deployment can take place within the Dragon Copilot environment. Canary is designed to work as part of the Dragon Copilot ecosystem rather than requiring clinicians to open a separate application during the encounter. Health systems can work with Canary and their internal teams to configure and deploy the solution within their existing Dragon Copilot environment.
   
   
3. The integration burden is reduced. Microsoft's extensibility model gives partners a standardized path into Dragon Copilot workflows, reducing the need for bespoke integration work. For clinicians, the goal is straightforward: no new hardware, no separate application to manage during the encounter, and no additional workflow step simply to capture the voice signal.

What Canary Contributes to the Encounter

Dragon Copilot captures what was said, Canary measures how it was said.

Canary's models analyze acoustic and linguistic properties of natural speech to identify patterns associated with specific health conditions. Rather than relying solely on what a patient says in response to a questionnaire, vocal biomarkers can provide an additional, objective signal from speech that is already part of a routine clinical interaction.

Inside the Dragon Copilot ecosystem, that includes:

Canary Cognitive: screens for signals associated with mild cognitive impairment and Alzheimer's disease. Cognitive changes can go unrecognized in primary care, even when patients are seen regularly. A voice-based screening signal offers another opportunity to identify patients who may warrant closer evaluation or formal cognitive assessment.

Canary Behavioral identifies vocal signals associated with depression and anxiety. These conditions are often the least likely to be volunteered in a fifteen-minute appointment and most likely to be missed when screening depends on a patient completing a form honestly.

Both are clinical decision support: Canary surfaces an objective signal for a clinician to interpret alongside everything else they know about the patient. It does not diagnose or make clinical decisions. The goal is to give clinicians another piece of actionable information without asking them to fundamentally change how they conduct the visit.

Built to Be Evaluated

Health systems evaluating anything in this category should ask hard questions, and the ones that matter are not primarily about accuracy claims (though those matter too, see our peer-reviewed research here). They are about intended use, consent, audio retention and deletion, whether patient audio is used to train vendor models, subgroup performance across accent and age and language, and what triggers retraining after go-live.

We have written about how to approach that evaluation-including the questions health systems should ask of vendors-in our guide to evaluating ambient voice biomarker technology.

Canary Speech participates in the Coalition for Health AI's ambient AI work group, and we believe emerging frameworks for responsible AI and ambient technology are good for the category. As voice AI moves deeper into clinical workflows, standardized approaches to evaluation, governance, and responsible deployment will become increasingly important.

Running within the Dragon Copilot ecosystem adds another layer of technical and operational requirements. Partners in the program must meet Microsoft's applicable requirements before their solutions are made available through the platform.

Getting Started

If your organization already uses Dragon Copilot, Canary Ambient for Physicians is available through Microsoft Marketplace and can be deployed within the Dragon Copilot environment.

View Canary Ambient for Physicians in Microsoft Marketplace

Marketplace availability does not mean self-service activation. Organizations interested in accessing Canary Ambient should contact the Canary Speech team at [email protected] to discuss licensing, access, and deployment.

If you are earlier than that-still determining which service lines would benefit or which models fits the population you are trying to better understand-start a conversation with our team.

And if you will be at HLTH USA in Las Vegas this November, we will be there and welcome an in-person conversation.

The visit is already happening, the conversation already being captured. The question is what else it can safely tell a care team.