Ambient AI scribes listen to a clinical encounter and draft the note, and the category has drawn some of the largest funding rounds in health technology. Two questions follow every one of those rounds: what does the tool actually do, and what independent evidence backs it. This page answers both the only honest way — as a dated landscape. It sets each vendor's own disclosures beside the independent studies that exist, ties every cell to a source, and declines to crown a leader. As of July 2026. For the underlying concept, see our glossary entry on the ambient AI scribe.
How to read this landscape
A ranking answers "which is best." A landscape answers two separate questions — what a vendor documents, and what evidence exists — and keeps them apart, because in this category they diverge sharply. A tool can carry a nine-figure round and a thin evidence base; another can anchor the largest published deployment and disclose less. Funding measures investor conviction. Independent evaluation measures what happened when clinicians used the tool. Reading the two as one number is the single most common error in this market.
Three cautions frame everything below. First, every "documented" cell reflects a vendor's own materials, which are marketing as well as fact — read them as claims to verify, never as findings. Second, an empty evidence cell is a prompt to ask the vendor for a study, never proof the tool performs poorly. Third, disclosures, integrations, availability, and funding all change on the vendors' timelines, so each cell is a snapshot dated July 2026.
The landscape at a glance
Every filled cell cites its source. A dash means the attribute was absent from the sources cited here — read it as "ask the vendor," never as "the tool lacks it."
| Vendor | Documented design and scope (vendor's own materials) | Disclosed funding / ownership | Vendor-independent evaluation identified |
|---|---|---|---|
| Microsoft Dragon Copilot | Ambient capture of multiparty, multilingual conversations; drafts specialty-specific notes for clinician review; surfaces information with cited sources; generally available in the US from May 2025 67 | Microsoft product; built on the earlier Nuance ambient lineage 6 | Randomized trial, DAX Copilot arm: no significant change in time-in-note 1; surgical-resident evaluation 5 |
| Abridge | Ambient documentation deployed inside an integrated EHR; states partnering with more than 150 enterprise health systems 89 | $300M Series E, 24 Jun 2025 8 | Large real-world deployment: 303,266 encounters in 10 weeks; over 2.5M uses in year one 23 |
| Ambience Healthcare | Ambient platform documenting more than 100 ambulatory subspecialties, EDs, and inpatient specialties; Epic, Oracle Cerner, athenahealth 10 | $243M Series C, 29 Jul 2025 10 | — (only the category-level blinded evaluation 4) |
| Nabla | Ambient AI assistant; multilingual capture; reported 85,000 clinicians across more than 130 organizations 1112 | $70M Series C, 17 Jun 2025; total $120M 11 | Randomized trial, Nabla arm: −9.5% time-in-note, p=.02 1 |
| Suki | Assistant spanning documentation, dictation, coding, chart-data retrieval, and clinical question answering 13 | $168M raised to date after a Zoom Ventures investment, 30 Jan 2025 (on a $70M Series D, Oct 2024) 13 | — (only the category-level blinded evaluation 4) |
The most important column is the last, and its most honest feature is how empty it is. Two of the five best-funded tools have no vendor-specific independent study in these sources at all; the evidence that exists concentrates on a different two. That asymmetry — capital pooling faster than evidence — is the defining feature of the field as of July 2026.
The disclosed funding, dated
Funding rounds are drawn from each company's own announcement. They are dated because they age quickly, and they are labelled as disclosures because a self-reported round is a primary statement of fact from the vendor rather than an independent audit.
| Vendor | Latest disclosed round | Amount | Date | Source |
|---|---|---|---|---|
| Abridge | Series E | $300M | 24 Jun 2025 | 8 |
| Ambience Healthcare | Series C | $243M | 29 Jul 2025 | 10 |
| Nabla | Series C (total $120M) | $70M | 17 Jun 2025 | 11 |
| Suki | Strategic investment ($168M to date) | undisclosed increment | 30 Jan 2025 | 13 |
| Microsoft Dragon Copilot | — (Microsoft product) | n/a | GA US May 2025 | 6 |
Our healthcare AI funding and M&A tracker holds the wider set of rounds and their dates. The figure to carry from this table is a caution: a larger round tells you where investors are placing bets, and nothing about note accuracy, time saved, or safety in your clinics.
The independent evidence, dated
This is the other axis. Every row is a study, its design, the tools it actually named, and its headline — the discipline our guide on how to read an AI validation study sets out in full.
| Study | Design | Tools named | Headline result |
|---|---|---|---|
| Randomized trial 1 | 238 physicians, 14 specialties, no vendor funding | DAX Copilot, Nabla | Nabla −9.5% time-in-note (p=.02); DAX Copilot no significant change |
| Integrated-system deployment 23 | one large group, single EHR | Abridge | 3,442 physicians / 303,266 encounters in 10 weeks; over 2.5M uses in year one |
| Blinded quality evaluation 4 | 11 tools vs human note-takers, 5 standardized primary-care cases | 11 tools (unnamed here) | Human notes scored higher on the modified PDQI-9 across all five cases |
| Surgical-resident evaluation 5 | specialty-specific | DAX Copilot | May help reduce documentation burden in that setting |
Read the randomized trial 1 first, because it is the only vendor-independent head-to-head. On its primary outcome — time spent in the note — one tool achieved a statistically significant reduction and the other did not, even though both are marketed for the same job. One trial at one academic system is a single data point rather than a verdict, but it is a pointed argument against trusting any league table over a local test.
The deployment record 23 answers a different question: can a scribe run at scale? For one integrated group using Abridge, the answer was plainly yes — past 2.5 million uses in a year. Treat that as the ceiling of a well-resourced rollout inside a single EHR, rather than the median any organization should expect; the AI scribe adoption statistics page holds the fuller deployment and time-savings set.
The blinded evaluation 4 supplies the sober counterweight the funding headlines omit: across 11 tools and five standardized cases, human-written notes still scored higher on a standard quality instrument. That is the accuracy question every buyer inherits, and it sits underneath every cell in the landscape.
What no source here provides is a ranked, like-for-like accuracy comparison of the named products under a single rubric. The head-to-head trial covers two tools on a time outcome 1; the blinded quality study covers 11 tools but reports them against human notes rather than against each other by name 4. So the honest state of the evidence is that the field has a scale record, one time-outcome comparison, and a category-level quality signal — and no published league table of note accuracy. Any vendor claim of being the most accurate is, as of July 2026, unbacked by a neutral head-to-head, and should be read as a prompt to run your own blinded comparison.
Each vendor in brief
Each of these captures a conversation and drafts a note for a clinician to review and edit. They differ in lineage, scope, disclosed capital, and how much independent study they have accumulated.
- Microsoft Dragon Copilot carries the lineage of the earlier Nuance DAX Copilot. Microsoft documents ambient capture of "multiparty, multilingual patient-clinician conversations" and drafting of "specialty-specific notes" for review, generally available in the US from May 2025 67. Its independent evidence is the DAX Copilot arm of the randomized trial, which showed no significant change on time-in-note 1, plus a specialty-specific resident evaluation 5.
- Abridge anchors the largest published integrated-system deployment 23 and disclosed the field's largest recent round, a $300M Series E, stating partnership with more than 150 enterprise health systems 89.
- Ambience Healthcare documents unusually broad scope — more than 100 subspecialties, EDs, and inpatient settings across three major EHRs — and disclosed a $243M Series C 10. No vendor-specific independent trial appears in these sources.
- Nabla documents a multilingual ambient assistant and reports 85,000 clinicians across more than 130 organizations 1112; its Nabla arm of the randomized trial is the one that cut time-in-note by 9.5% 1.
- Suki documents the widest functional scope on paper — documentation, dictation, coding, chart-data retrieval, and question answering — and reported $168 million raised to date 13. No vendor-specific independent trial appears in these sources.
The shared design, and why it matters
Across every vendor the documented design converges on one fact: the tool drafts a note, and a clinician reads, corrects, and signs it. That shared human-in-the-loop step is what keeps a drafting error from silently entering the record, and it is the structural reason ambient scribes generally sit outside the FDA's medical-device pathway. The agency's clinical decision support guidance turns in large part on whether a clinician can independently review the basis for the software's output rather than rely on it 14; because a scribe produces a draft a clinician edits before signing, it typically falls on the non-device side of that line. That framing can shift as products add ordering, coding, or recommendation features — see our glossary entry on the clinical decision support system — so confirm the current status of any specific tool rather than assuming the category answer holds. For the deeper attribute-by-attribute view, our AI scribe head-to-head comparison lays the capability matrix out in full.
How to read this landscape
Four cautions travel with everything above. First, the two axes are independent: disclosed funding and independent evidence measure different things, and a strong number on one says nothing about the other. Second, the evidence is lopsided — two tools carry the studies, and absence of a study is not proof of poor performance. Third, self-disclosed capabilities and customer counts are the vendors' own statements, useful but unaudited. Fourth, every figure here is a snapshot dated July 2026; rounds close, availability expands, integrations change, and clearances and device status can move — reconfirm each cell against the vendor before you rely on it. Because parts of this page touch procurement and regulatory status, confirm current device classification and contract terms with your own compliance or legal function before acting.
Sources and method
This landscape draws on one vendor-independent randomized trial 1, a ten-week and one-year deployment record from a single integrated system 23, a blinded multi-tool quality evaluation 4, and a specialty-specific evaluation 5, set beside each vendor's own product and funding disclosures 678910111213 and the FDA's clinical decision support guidance 14. Funding figures are each company's own dated announcement; every evaluation figure is drawn from the primary study cited beside it. We revisit this page on a 180-day cycle and whenever a vendor discloses a new round, a change of ownership, or a documented change in capability, or a new vendor-independent trial lands. As of July 2026.