Uncovr

The computer vision models are the entire product. Uncovr reads raw operating room video, segments the procedure into steps, instruments and events, and derives both the operative narrative and the codes from that segmentation. There is no video management platform, archive or workflow tool underneath that would retain value without the models, and the company was founded around the vision capability rather than adding it to an existing product.

A stated hard gate plus the most auditable output in this space, held back by one unanswered question. Uncovr states that output is clinician reviewed before every submission, so nothing files autonomously, and it staffs senior surgeons and expert medical coders against the output. The auditability is the standout: every CPT code is presented with linked evidence, meaning the passage of the report supporting it AND the video timestamp it was derived from.

That is the Linked Evidence property this index credits in Abridge, applied to a billing code and anchored to a second of footage, which is a stronger form of justification than any vendor in the coding category offers. The unanswered question is WHO reviews.

Clinician reviewed before every submission does not say whether the reviewer is the operating surgeon who performed the case or a reviewer employed by Uncovr, and those are entirely different oversight models with different accountability. Also missing: any published rate at which the reviewer disagrees with the draft, any confidence threshold, and any statement of behaviour when the video is obscured, which is routine in laparoscopy.

The product interface demonstrates what the system outputs, showing step segmentation with named instruments and procedural phases against a timeline, but the mechanism behind it is undescribed. No model is named, no architecture is given, there is no description of training data or how procedures were annotated, and no evaluation methodology or accuracy figure for the vision models is published. References to our models and our agents carry no further detail. The chief technology officer's background in autonomous tracking systems at the European Space Agency and ETH Zurich is credible provenance but it is not disclosure about this product.

One architectural commitment bounds the chain and nothing names it. Anonymisation is stated to run on device before any processing occurs, so if it performs as described, whatever parties sit downstream receive content from which identifiers have already been removed rather than receiving identifiable operating room video and being trusted with it.

That is the same class of structural protection this index credits where redaction precedes transmission, and for video it matters more than for text, because a video frame cannot be selectively redacted after the fact the way a transcript can. Customer controlled retention and end to end encryption sit alongside it. On enumeration there is nothing.

No model is named, no architecture is described, no training data source or annotation process is given for the vision models, and no hosting arrangement or sub processor list was located. References to the company's models and agents carry no further detail, and a founder's background in a different field is provenance rather than disclosure about this product.

The annotation question is worth asking specifically: procedural step segmentation with named instruments implies a labelled corpus of real operations, and whose operations those were, and on what basis, is a supply chain question rather than a technical one. Ask what the on device anonymisation removes, who processes the anonymised video, and where the training corpus came from.

Uncovr does something this index rewards, publishing denominators alongside its figures, and it still lands at C because of what the figures measure. Credit first: the company labels its foregut analysis with N equals 100 and cites an HCA Healthcare series at N equals 1,011, where most vendors publish percentages with no denominator at all. The problem is that the two numbers are of different kinds and appear under one heading reading real results.

The foregut figure, 16 percent of revenue lost to poor documentation and undercoding, is Uncovr's own deployment finding with no stated methodology, cohort definition or comparison arm. The HCA figure, 70 percent of recommended detail missing correlating with reoperation, readmission and surgical site infection rates, measures how poor operative reports are in general.

It characterises the PROBLEM Uncovr addresses, not Uncovr's PERFORMANCE against it, and presenting the two together invites a reader to attribute both to the product. Beyond that: no peer reviewed publication, no accuracy or completeness measurement of Uncovr's own output against a reference standard, and the Moon Surgical deployment reports 20 cases. The company was founded in 2025, so a thin evidence base is expected rather than damning, but it is thin.

The most specific architectural commitment located for any video based product in this index, and it is four distinct claims rather than one slogan: a zero PHI architecture, on device anonymisation performed BEFORE any processing, end to end encryption, and customer controlled retention. If accurate, identifiable data never leaves the operating room, which addresses the central exposure of surgical video head on. Held at B rather than A for two reasons.

Nothing is independently verified and no attestation covers it. And the number that matters most is absent, exactly as this index records for C-SATS: there is no published de identification error rate, and a de identification failure on operating room video discloses PHI of an unusually intimate kind. Surgical video can carry faces, tattoos, distinguishing anatomy and the voices and images of theatre staff, and nothing published states what the anonymisation covers or how it was validated. Ask for the error rate and the scope before treating zero PHI as established.

Converted from Not Rated after a second search that reached the company's own site. The compliance position is unchanged and is now a tested finding rather than an untested one: no statement of compliance with the United States health privacy rule, no business associate agreement availability or terms, no European data protection position, and no stated lawful basis or data protection contact.

The axis reaches this vendor unusually directly. Operating room video is among the most identifiable clinical material that exists. It records the patient's anatomy, and depending on camera placement can capture faces, wristbands, monitor displays, paperwork and staff. The operative report and procedure codes derived from it are unambiguously protected health information in a United States setting. This is not a vendor where the question can be scoped away.

Both regimes plausibly apply and neither is addressed. The company is headquartered in Paris with an office in New York, has a named deployment at a French hospital, and cites a United States hospital series in its materials.

One distinction is doing a lot of work here and a buyer should not let it pass. The company does state that anonymisation runs on the capture device before any processing, so raw video is not uploaded. That is an architectural claim and it is a good one. A third party software directory reports this as a zero protected health information design. Those are not the same claim. Anonymisation describes what the software does; zero protected health information is a legal conclusion about what the resulting data is, and that conclusion depends on meeting a defined standard, either expert determination or removal of the enumerated identifiers under the United States rule, and true anonymisation rather than pseudonymisation under European rules. The company has asserted the architecture. It has not asserted the legal status, and nothing published shows the anonymisation has been tested against either standard. A directory collapsing the two is doing the vendor's legal work for it.

Four things to establish before contracting. What specifically the on device step removes, and whether that has been assessed against a recognised de identification standard by anyone outside the company. Whether the company will execute a business associate agreement and which entity signs it. What lawful basis and retention apply to European processing. And for any French deployment, whether hosting sits with a certified health data host, which is a country specific requirement rather than a general European one and is frequently missed by companies entering the market from outside it.

Note also that the company operates under two domains, uncovr.ai and a separate surgical branded site, with overlapping content. Confirm which entity and which domain the contract names.

The security section is headed enterprise grade security and controls, and enterprise grade is a marketing adjective that this index tracks on a watchlist rather than treating as evidence. What sits under it is better than the heading: four specific technical commitments covering anonymisation, encryption and retention. What is entirely absent is any third party assurance.

No SOC 2 of either type, no HITRUST, no ISO 27001, no penetration test summary, no trust centre and no report request path. Graded C rather than Not Rated because concrete controls are described, but a health system security review will ask for an attestation and there is none to point at.

No device authorisation and none apparently required, since generating documentation and proposing codes are not clinical determinations about a patient. The live regulatory surface is different and unaddressed. In the United States, procedural coding sits under False Claims Act exposure, and a product whose headline value is recovering revenue from previously unbilled steps places that exposure squarely on the customer.

The video linked evidence is a genuinely strong answer to that risk, discussed on the governance axis, but no compliance position, coding audit methodology or documented regulatory analysis is published. For the French entity, any drift from documentation into clinical claims would raise EU medical device questions; nothing indicates such a claim is being made.

No fairness, subgroup or performance variation disclosure was located. The more consequential governance question is the revenue gradient, and Uncovr occupies a genuinely new position on it that deserves stating fairly in both directions. Every vendor this index tracks on coding intensity infers what was done from what was documented, which is why upcoding risk is intrinsic to the category.

Uncovr inverts the evidence: it proposes a code and attaches the video timestamp where the act occurs. That is the first evidentially grounded justification for additional codes the index has assessed, and it is a real defence rather than a marketing line.

Against it: the product is marketed heavily on recovery, with revenue lost framing, a revenue cycle audience and revenue recovery among its named uses, and a model optimised to surface billable steps carries an incentive gradient regardless of how good its evidence is. Video does not settle whether footage actually supports a given code, which remains a clinical and coding judgment.

Separately, and applying this index's unadjusted benchmarking finding from Theator and Caresyntax: timestamped step level data on named surgeons is precisely the substrate for comparative performance analysis, and nothing published states whether it will be used that way or with what risk adjustment.

Two passes located no accuracy figure or error rate for the vision models, no published limitations and no warranty, indemnity or remediation commitment, and the absent number is a specific one that carries more weight here than a general accuracy figure would. The product rests on a zero patient information architecture with anonymisation performed on device before any processing, and no de identification error rate is published.

That is the number the whole claim depends on, because anonymisation is a model making judgements and models make mistakes, so zero patient information is a design intent until a failure rate is attached to it. The consequences of a failure are unusually severe in this setting. Operating room video can carry faces, tattoos, distinguishing anatomy and surgical sites, and a de identification miss discloses patient information of an intimate kind that cannot be retracted once seen.

It also creates a category of affected person this index has recorded only once before. Theatre staff are recorded continuously, by voice and image, throughout every procedure. They are not patients, they hold no privacy right in the clinical sense, they may have had no say in the system's adoption, and they have no relationship with the vendor and no route to raise anything. Establish what the anonymisation covers, how it was validated and at what error rate, what happens to staff voices and images, and what the vendor commits to when a de identification fails.

Strong on the side that captures data, undocumented on the side that has to receive it. Uncovr states it connects to any operating room video source across laparoscopic, robotic and any device, and it has a demonstrated platform integration: in June 2026 it became the first third party AI application running on Moon Surgical's Maestro platform. That is real device layer interoperability.

But the output of this product is an operative report and a set of codes, and both must reach the electronic health record and the billing system to be useful. No EHR is named, no HL7 or FHIR write back path is described, and no reference integration is cited. The contrast inside this index is instructive: Theator, working with the same raw material, earned a top interoperability grade for native structured write back into Oracle Health. Until that path is documented, a buyer should assume the report needs to be moved manually.

Better described than most because the architecture forces a statement: anonymisation runs on device before any processing, so the first stage of the pipeline is edge compute inside the operating room rather than a cloud upload of raw video. The company also states that customer data is retained on the customer's terms, which implies configurable retention rather than a fixed vendor policy.

What is not published: where post anonymisation processing occurs, which cloud or region, tenancy model, and whether European and United States customers are served from separate infrastructure, which matters for a company operating from Paris and New York with customers in both markets.

No price, tier or pricing unit is published and every path routes through a booked demo. The unit question is unusually consequential here because the product's stated value is expressed in recovered revenue: pricing per operating room, per procedure, per surgeon or as a share of recovered reimbursement produce very different incentive structures, and the last of those would tie the vendor's compensation directly to coding intensity. Per this index's standing contingent pricing check, establish in writing whether any fee component varies with codes captured, claims value or revenue recovered.

Scoped to procedures that generate video, which is the natural boundary: laparoscopic, robotic and endoscopic surgery. Within that the reach is real, with a stated deployment pipeline above 400 operating rooms across the United States and Europe and thousands of hours of surgical video analysed.

Breadth across every surgical specialty is claimed but the evidence concentrates in one area: the published study is foregut surgery, and the named surgeon advocates are foregut, hepatopancreatobiliary, minimally invasive and robotic specialists. That is a coherent beachhead rather than a weakness, but a buyer in orthopaedics, gynaecology or cardiothoracic surgery should ask what has actually been validated in their specialty. Open surgery, which generates no scope video, is outside the model entirely.