Microsoft Dragon Copilot
The only ambient scribe with a published SDK, an extension marketplace and ten countries of availability including Belgium and the Netherlands, and the only one whose flagship randomised trial found no significant documentation-time saving at all.
PRICING
| Per-user licence (from 01/05/2026) | Quote only | single standard licence, full enterprise feature set, reduced list price across all 10 geographies; Physician Practice SKU discontinued |
| Consumption fee | Quote only | billed per encounter for ambient and generative AI, replacing per-note and per-AI-action billing |
| Historic reseller range | $369-$830+/mo | per provider, third-party reseller listings before the May 2026 reprice; treat as a range for procurement, not a quote |
| Individual / self-serve | Not sold | no self-serve checkout and no published price for a solo clinician |
Microsoft publishes no list price for this product in any currency. The dollar ranges here come from resellers and third-party reviews, not from Microsoft. Available for physicians in the US, Canada, UK, Ireland, France, Germany, Austria, Belgium, the Netherlands and Switzerland; nursing capabilities are US-only and the radiology experience is in US preview.
checked 2026-08-27 · vendor pricing page
Element scores
Strengths
Two things separate this from every other ambient scribe. First, it is the only one with a publicly documented developer surface: a JavaScript SDK, an Android ambient SDK, DAXKit for iOS, embeddable web apps and partner APIs for data exchange and session control, all on Microsoft Learn where anyone can read them before signing anything. On top of that sits an extension marketplace where partners such as Regard, Optum, Humata Health and Canary Speech receive the patient context, audio turns, note and transcript, and push adaptive cards back into the clinician's view. Second, it is available for physicians in ten countries including Belgium, the Netherlands, France and Germany, which is unusual in a category where almost everything is US-only. The product itself is broad rather than deep: 58-language ambient capture, order capture straight into Epic, ICD-10 specificity prompts, referral letters, after-visit summaries, nursing flowsheet capture, and a radiology experience with PowerScribe One. Governance is Microsoft-grade, with SOC 1 and 2 Type II, C5, NEN7510, 90-day data retention through Purview, and UKCA Class I medical device marking in Great Britain. Adoption is real: more than 100,000 clinicians daily as of March 2026.
Honest dings
The evidence does not match the marketing. The one vendor-independent randomised trial (UCLA Health, 238 physicians across 14 specialties, roughly 72,000 encounters, NEJM AI 26/11/2025) put DAX Copilot, Nabla and usual care head to head. Nabla cut time-in-note by 9.5% versus control. DAX moved it 1.7%, which was not significant at p=0.66. Both arms improved on burnout and task load, and those are self-reported secondary endpoints. So on the primary outcome the product is sold on, one trial that nobody paid for found nothing. A separate NEJM AI study at Atrium Health concluded that widespread implementation in its current form is unlikely to generate appreciable productivity gains. Real-world observational work is kinder (a 2026 emergency-department study found 28% less on-shift documentation time) but shows adoption skewed hard, with the top 10% of users accounting for 70.5% of ambient encounters. Then the buying problems. There is no published price in any currency, in any of the ten countries, so a Belgian hospital cannot compare it against anything without opening a procurement cycle. The marketplace and the nursing capabilities are US-only, so the European availability is thinner than the country list suggests. And a customer gets no eval suite, no model choice, no cheaper tier and no MCP or A2A, so measuring or moving the output is not something you do yourself.
Sources (14) — every claim traceable
Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).
- microsoft.com/en-us/microsoft-cloud/blog/healthcare… — Official HIMSS 2026 announcement 05/03/2026: more than 100,000 clinicians daily; Marketplace AI apps and agents from Canary Speech, Humata Health, Optum, Regard; Work IQ context via M365 Copilot; 58-language capture; ICD-10 specificity suggestions; pull-forward; nursing flowsheets; radiology preview (accessed 2026-08-27)
- learn.microsoft.com/en-us/industry/healthcare/drago… — Official developer documentation: JavaScript SDK, embeddable desktop and mobile web apps, Android Ambient SDK, DAXKit for iOS, partner APIs for data exchange, feedback, licence checks and recording sessions; non-production access requires an established Microsoft partnership (accessed 2026-08-27)
- learn.microsoft.com/en-us/industry/healthcare/drago… — Official extension documentation: AI apps and agents receive patient context, audio turns, clinical note and transcript, and return adaptive cards; explicitly limited to the US market (accessed 2026-08-27)
- learn.microsoft.com/en-us/industry/healthcare/drago… — Official security and data FAQ: 90-day retention via Microsoft Purview, TLS 1.2/1.3 and AES-256, Entra ID, SOC 1 and 2 Type II, C5 Type II, NEN7510, ISO 27001 and 27701 certification pending, HIPAA/PIPEDA/GDPR alignment, de-identification for model improvement (accessed 2026-08-27)
- learn.microsoft.com/en-us/industry/healthcare/drago… — Official intended-use whitepaper: explicit unsupported uses (no autonomous clinical decision-making, no multi-patient sessions), UKCA Class I medical device marking in Great Britain only, customer responsible for EHR filing, no real-time collaborative editing (accessed 2026-08-27)
- microsoft.com/en-us/health-solutions/clinical-workf… — Official product page: physician availability in the US, Canada, UK, Ireland, France, Germany, Austria, Belgium, the Netherlands and Switzerland; nurses US-only; radiologists in US preview; trained on more than 15 million encounters; custom templates and saved AI prompt library; Microsoft Fabric integration (accessed 2026-08-27)
- pmc.ncbi.nlm.nih.gov/articles/PMC12768499 — Primary independent evidence. 'Ambient AI Scribes in Clinical Practice: A Randomized Trial', NEJM AI, 26/11/2025, NCT06792890 — 238 outpatient physicians, 14 specialties, three arms (DAX Copilot / Nabla / usual care), 04/11/2024 to 03/01/2025. DAX -1.7% time-in-note vs control (95% CI -9.4% to +5.9%, P=0.66, not significant); Nabla -9.5% (P=0.02). DAX used in 33.5% of 24,696 visits. Mini-Z +2.83 and PTL -39.9 for DAX. One grade 1 adverse event; occasional clinically significant inaccuracies in both arms. Funded by UCLA, not by a vendor (accessed 2026-08-27)
- learn.microsoft.com/en-us/partner-center/announceme… — Official Partner Center announcement 01/04/2026: reduced per-user list price effective 01/05/2026 across all ten geographies; Physician Practice offer end of sale 01/05/2026; consumption fees move to per-encounter rather than per note or per AI action (accessed 2026-08-27)
- plaud.ai/blogs/articles/dragon-copilot-review — Independent review 17/08/2026 (treat pricing as estimate): historic reseller range roughly $369 to over $1,000 per provider per month, Physician Practice briefly at $150-160, 60% rural hospital discount, named deployments at Mount Sinai and Tampa General, nine countries claimed (accessed 2026-08-27)
- pmc.ncbi.nlm.nih.gov/articles/PMC13154711 — Independent scoping review of ambient documentation in emergency medicine: Preiksaitis et al. (2026) retrospective study of 8,740 academic ED encounters found 11.2% overall DAX Copilot adoption with the top 10% of users accounting for 70.5% of ambient encounters, 28% less on-shift documentation time (2:45 vs 3:50) and 16% less total EHR time; Guo et al. (2026) UCI Health mixed-methods found about 14% less daily documentation time pooling DAX and Abridge (accessed 2026-08-27)
- pubmed.ncbi.nlm.nih.gov/41366572 — Independent peer-reviewed assessment, Surgical Endoscopy: DAX Copilot notes scored 46.91/50 (SD 1.35) for accuracy, thoroughness, comprehensibility, succinctness, synthesis and internal consistency, free from hallucination and bias; organisation and usefulness scored 4/5 because default notes still need editing for surgical style (accessed 2026-08-27)
- veroscribe.com/blog/nuance-dax-review-2026 — Competitor-published review 20/03/2026 (treat pricing and framing as partisan): cites the NEJM AI Atrium Health study of 112 clinicians concluding widespread DAX implementation in its current form is unlikely to generate appreciable productivity gains; 600+ organisations; $369-$830+ per provider per month (accessed 2026-08-27)
- schneider.im/microsoft-dragon-copilot-price-decreas… — Independent licensing analysis of the 01/05/2026 change: one per-user licence with no functional differences between former offer variants, ambient billed per encounter, generative AI billed per encounter, price reduction across Germany, Austria, Switzerland, Belgium, the Netherlands, France, UK, Ireland, US and Canada (accessed 2026-08-27)
- learn.microsoft.com/en-us/industry/healthcare/drago… — Official nursing documentation 19/03/2026: ambient flowsheet capture into the EHR, narrative nurse notes, transcript querying, information guidance from organisation-approved references (accessed 2026-08-27)