Microsoft Dragon Copilot — Ambient Clinical AI Vendor Profile

Organizations using DAX

600+

Reported by Microsoft in March 2025 — more than 3 million ambient patient conversations a month across 600+ healthcare organizations on DAX, the ambient component now folded into Dragon Copilot.

Epic-embedded systems

150+

Hospitals and health systems using DAX Copilot directly inside Epic note workflows as of 2024 — Microsoft's longest in-EHR reference base, predating the Dragon Copilot rebrand.

UCLA NEJM AI RCT

−1.7% (n.s.)

238-physician pragmatic randomized trial (2024–25). DAX showed a −1.7% change in time-in-note versus usual care, not statistically significant (p=0.66). Nabla in the same trial cut time-in-note by 9.5% (p=0.02). Buyers should know this.

Microsoft survey

70% / 50%

Microsoft-reported clinician survey: 70% said DAX reduced feelings of burnout; 50% cut in time spent on documentation. Treat vendor-reported numbers as upper-bound — the peer-reviewed RCT is the cleaner reference.

What Dragon Copilot actually is

Dragon Copilot is Microsoft's unified clinical voice AI assistant, launched March 2025 and generally available in the U.S. and Canada in May 2025. It is not a new product so much as a rebranding and consolidation: the long-standing Dragon Medical One (DMO) structured-dictation product and the DAX Copilot ambient-documentation product, both inherited from Microsoft's $19.7B acquisition of Nuance in 2022, now sit behind one Copilot surface with shared models, identity, and Azure plumbing.

For the clinician the experience covers three modes on the same audio: hands-free dictation into structured fields, ambient capture of the patient encounter with a draft note generated back into the EHR, and conversational Q&A over chart context. A March 2026 update added a nursing-documentation mode (U.S. med-surg) and expanded third-party agent access; UK, Germany, France, and the Netherlands followed the original U.S./Canada general availability.

The Microsoft pitch is breadth, not RCT supremacy: Dragon Copilot is a single contract that subsumes dictation, ambient, and Copilot agents inside a stack where most of the buyer's other tools — Teams, M365, Azure identity, Power Platform — already live. That is the differentiator a competing scribe vendor cannot replicate.

Deployment posture

Dragon Copilot is Microsoft Cloud for Healthcare on Azure, with all inference and storage handled in Microsoft's HIPAA-aligned cloud regions under a Business Associate Agreement. There is no customer-tenanted on-prem deployment. For hospitals already running M365 / Azure under enterprise agreements with Microsoft, the procurement and identity model is the path of least resistance. For Canadian hospitals constrained by provincial residency, the question is which Azure regions Microsoft will commit to in the contract, and what audit telemetry the customer gets.

SURFACE

Embedded in Epic + Microsoft 365

Nuance's AI copilot was the first ambient scribe fully embedded in Epic. Dragon Copilot also surfaces inside Teams and the Microsoft 365 Copilot fabric — the strongest cross-product story in the category.

DATA PATH

Azure-only processing

Audio, transcripts, and generated notes route through Microsoft Cloud for Healthcare in Azure. Encryption in transit and at rest. HIPAA Business Associate Agreement; Microsoft publishes compliance attestations under the standard Azure Trust Center.

CONTRACT

Per-clinician subscription

Sold via Microsoft enterprise agreements or authorized resellers. Public list pricing for DAX-tier subscriptions sits in the high-tens-to-mid-hundreds USD per clinician per month; enterprise rates are negotiated and not published. BAA bundled.

SCALE

Largest installed base

The Nuance DMO + DAX user base — millions of clinicians on dictation, 600+ organizations on ambient — gives Dragon Copilot the broadest deployment surface in the category by a wide margin, even where the RCT result is unfavorable.

The evidence base

Dragon Copilot's published evidence is dominated by one peer-reviewed trial — and that trial is the most important number a buyer should weigh. Microsoft has additional vendor-reported numbers; treat them as upper-bound.

closeUCLA NEJM AI pragmatic RCT, 2025. 238 outpatient physicians across 14 specialties, randomized 1:1:1 to DAX Copilot, Nabla, or usual care. DAX users showed a −1.7% change in time-in-note vs control, not statistically significant (p=0.66); Nabla in the same trial showed a 9.5% decrease (p=0.02). DAX was used in 33.5% of 24,696 visits. Both scribes improved Mini-Z, physician task load, and professional-fulfillment scores. Clinically significant inaccuracies were rated similarly across the two (DAX 2.7, Nabla 2.8 on a five-point Likert). NCT06792890.
checkMicrosoft / Nuance survey data. Of clinicians using DAX in the EHR, 70% reported reduced feelings of burnout and 50% reported less time on clinical documentation. Self-report, not blinded.
checkSingle-clinician reports (Microsoft case studies): ~2 hours of charting saved per 12-hour nursing shift; similar magnitudes from physician case studies at Stanford Health Care, WellSpan Health, Atrium Health. These are vendor-curated and not generalized.
checkMulticenter ambient-scribe meta-analyses (2025–26) consistently bundle DAX into the ~1–3 hour-per-day documentation-time-savings range observed across the category. The NEJM AI RCT is the cleanest head-to-head and the one that breaks the average.

What the RCT does not mean: that DAX Copilot does not work. Many clinicians prefer it, the burnout signal is real, and the second-generation model that Microsoft has shipped since the trial may close the gap. What it does mean: any buyer hearing "ambient scribes save two hours a day" from a Microsoft sales conversation should ask what the peer-reviewed minute-counts say, and whether the pilot will include both Nabla and DAX as a head-to-head.

Privacy and compliance posture

Dragon Copilot runs on Microsoft Cloud for Healthcare on Azure. Microsoft signs a HIPAA Business Associate Agreement, holds the standard Azure compliance attestations (SOC 2 Type 2, ISO 27001, HITRUST), and provides regional residency choices to the limit of where Microsoft offers Azure healthcare regions. Audio retention, training-data use, and customer-controlled deletion are governed by the Microsoft Cloud for Healthcare data terms; default settings have shifted multiple times since the Nuance acquisition, so the contract — not the marketing — is the controlling document.

What it does not offer: a customer-tenanted on-prem deployment. The trust model is "Microsoft Azure, audited, contractually bound." For most U.S. enterprise Microsoft customers this is unremarkable. For Canadian hospitals constrained by Ontario PHIPA, Alberta HIA, Quebec Law 25, or BC FIPPA, the binding question is which Azure region the data lands in and whether the contractual residency commitment matches the regulator's interpretation of cross-border PHI flow.

Strengths and limitations, plainly

STRENGTHS

Where Dragon Copilot wins

Largest installed user base in the category. Unified product covering dictation + ambient + agentic Copilot inside one Microsoft contract. Deepest Epic and Microsoft 365 / Teams integration. Mature enterprise security, identity, and compliance program already proven in healthcare procurement. Best fit when the organization is already Microsoft-standardized and wants one vendor instead of three.

LIMITATIONS

Where it does not fit

Cloud-only — no on-prem tenant. The single largest peer-reviewed RCT did not show a statistically significant reduction in time-in-note for DAX, and a direct competitor in the same trial did. Pricing is opaque. Like every ambient scribe, hallucination and omission rates are non-trivial and "signature is review" is not a safe workflow. Not the right answer if the constraint is "PHI cannot leave hospital infrastructure" or if peer-reviewed minute-counts outweigh ecosystem fit.

Where Dragon Copilot fits versus an on-prem alternative

The honest framing for a directory reader weighing Dragon Copilot against a hospital-owned stack:

Dimension	Dragon Copilot (cloud SaaS)	On-prem (WalledCare)
Time to first clinician value	Days to weeks for existing Nuance / Microsoft customers; longer for greenfield Microsoft procurement.	30–90 day pilot on hospital-owned hardware; longer if GPU procurement is from scratch.
Deployment surface	Microsoft Cloud for Healthcare on Azure only.	Hospital network only; no outbound API.
Data residency	Limited to Microsoft's available healthcare Azure regions; contractual residency commitments must be confirmed in writing for Canadian provincial constraints.	Inside the hospital's own data center.
Ecosystem fit	Best-in-class if the buyer is already standardized on Microsoft 365 / Teams / Azure.	EHR-agnostic; HL7 / FHIR + sidecar; no Microsoft dependency.
Published evidence (time saved)	UCLA NEJM AI RCT: −1.7% time-in-note, not significant.	Internal evaluation against the same rubric; no public RCT yet.
Pricing posture	Per-clinician subscription, negotiated, bundled inside Microsoft enterprise agreements.	Capex + ops; predictable past year one, shared across multiple workflows.

Dragon Copilot is the right answer for hospitals whose primary constraint is "we already run on Microsoft; give us scribe inside the same contract." It is the wrong answer when the constraint is "PHI never leaves the building" — or when a peer-reviewed time-saved number is the deciding metric, in which case a head-to-head pilot against Nabla deserves a written line in the RFP.

Where this fits in the WalledCare directory

Dragon Copilot is the second commercial reference most buyers benchmark in the AI Scribes category — alongside Abridge for Epic-first U.S. systems and Nabla as the peer-reviewed-evidence counterweight. Among enterprise-platform plays, the closest peer is Commure Ambient, which positions an AI-native EHR rollup around the same buyer. The Microsoft ecosystem story is unique; the RCT story is a real signal in the other direction. Both deserve to be in the conversation. The canonical permalink for this page is /directory/vendors/dragon-copilot/.

If your team is at "we are a Microsoft shop and need a unified clinical voice product," Dragon Copilot belongs on the shortlist. If your team is at "we cannot route encounter audio outside our hospital network," skip the demo and read our reference stack for hospital-owned clinical AI instead.

Quick facts

Owner	Microsoft (Nuance Communications acquired April 2022 for $19.7B). Headquartered in Redmond, WA, with Nuance heritage in Burlington, MA.
Category	Unified clinical voice AI: structured dictation (Dragon Medical One) + ambient documentation (DAX) + agentic Copilot in one product.
Deployment	Microsoft Cloud for Healthcare on Azure. HIPAA, BAA, SOC 2 Type 2, ISO 27001, HITRUST.
Primary EHR	Epic (fully embedded). Also Oracle Health (Cerner), Meditech, athenahealth, eClinicalWorks, and other major systems.
Geographic availability	U.S. and Canada generally available May 2025; UK, Germany, France, Netherlands rolled out subsequently.
Reference customers	Stanford Health Care, WellSpan Health, Atrium Health, Greater Baltimore Medical Center, Tampa General Hospital, plus 150+ Epic-embedded health systems on DAX.
Headline evidence	UCLA NEJM AI three-arm RCT (2025): DAX time-in-note change −1.7% vs control, not significant (p=0.66). Nabla in the same trial: −9.5% (p=0.02).
Recent capability adds	Nursing documentation (U.S. med-surg, 2026). Third-party agent ecosystem expansion (2026).
Website	microsoft.com/health-solutions/clinical-workflow/dragon-copilot

Make Dragon Copilot a head-to-head, not a default

Microsoft's installed base will get Dragon Copilot into nearly every healthcare AI shortlist. The right move is to put the RCT result in the RFP, evaluate it side-by-side with a peer-reviewed comparator, and decide whether ecosystem fit or measured minute-savings is the binding criterion. WalledCare's pilot process scopes the comparison concretely — same workflow, same evaluation rubric, real numbers for both options.

send Request a WalledCare pilot arrow_back All vendors