Augmedix

The service includes labour, and how much depends on which tier a customer buys. At one end the mobile application drafts autonomously; at the other, trained documentation specialists work through the visit alongside the model.

That mix is the company's proposition rather than a shortcoming, and it is the same split this index recorded where separately saleable software sits beside a staffed service. Automation has been increasing, with stated investment in reducing the scribe effort behind each note, so the centre of gravity is moving toward the model over time.

A buyer comparing this against a pure software scribe is comparing a service to a product, and the price difference reflects exactly that.

This is the clearest example in the index of oversight sold as a tier rather than described as a feature.

The product line is arranged along a single axis of how much human involvement the customer wants. The autonomous mobile application handles lower complexity settings with no person in the loop. The assisted tier adds specialist review after the fact. The live tier puts a documentation specialist in the encounter as it happens, and independent commentary states it is positioned for higher complexity environments needing heavy human oversight.

So the organisation chooses its own oversight level, and the choice is explicit and priced. Held at B because no automation or escalation rate is published for the autonomous tier, and because nothing describes what happens when a visit assigned to the automated tier turns out to be complex.

Products and workflow are documented clearly and the technology is not. No model is named, no accuracy figure published for the automated tiers, and no description of how work is routed between model and specialist.

The productivity claims carried by the parent company, around two hours of documentation time saved per physician per day and a reduction of more than 80 percent, are stated without a methodology, a baseline or a cohort, and they cover the combined product families rather than a specific tier.

This record introduces a form of the axis that applies wherever documentation is produced with people in the loop, and it deserves naming because the ordinary version of the question misses it entirely. The supply chain here is not only models. A global documentation workforce reviews and produces clinical notes, so the parties who receive encounter content include human specialists, and the enumeration a buyer needs covers organisations and locations rather than model providers alone.

None of it is published. The countries where specialists work are not stated, whether visit audio or video crosses a border is not stated, and whether the work is performed by employees of the vendor or by contracted third parties is not stated.

On the conventional half the position is equally bare: no model is named, no foundation model provider, class or version is disclosed, no hosting arrangement is published, no sub processor list was located, and the routing between automated production and specialist review is not described, so a buyer cannot even tell which encounters a person sees.

Independent review does describe a control regime around specialist access, workstation security and review processes, which is aimed at the right risk and is better than silence, but a described control regime is not an enumeration. Ask for the list of entities and countries that handle encounter content, and ask which tiers of the product involve a human at all.

Substantial commercial corroboration, no measurement.

A group purchasing agreement awarded in January 2025 with an organisation serving more than 65 percent of United States acute care providers and 97 percent of academic medical centres is a meaningful procurement signal, since those agreements follow evaluation. More than 20 major health systems and hundreds of sites of care were served at acquisition, and independent review aggregators report consistently positive clinician feedback about reduced after hours charting.

What is absent is any study. The published time savings carry no methodology, and nothing measures note accuracy, coding accuracy or clinical quality across the tiers, which matters most here because the tiers differ precisely in how much human checking occurs.

One unusual disclosure: while listed the company published audited financials, reporting 45 million dollars of 2023 revenue against a 19 million dollar net loss, so the economics of running a hybrid scribe service at scale are a matter of public record in a way competitors' are not.

Graded on an honest basis, and the risk here is different in kind from a software only scribe because people see the material.

Independent review describes a global documentation workforce trained on compliance, with operational controls around specialist access, workstation security and review processes, alongside encryption, access controls and audit logging. That is a described control regime aimed at the right risk, which is better than silence.

What it does not settle, and what a buyer should establish directly, is where the specialists are located, whether visit audio or video leaves the country, how long recordings are retained after the note is produced, and whether patients are told that a person other than their clinician will hear the encounter. This index has raised the offshore human review question before; here it is central to the product rather than incidental.

A stated contractual position rather than an inference: independent review reports that the company signs business associate agreements with all customers and trains its documentation workforce on compliance.

That is more than most records in this index can show, and it clears C on the strength of the commitment being explicit. Held at B rather than A because no independent attestation is named and the terms are not public, and because the acquisition raises the usual question of whether agreements now sit with the parent.

Recorded honestly and provisionally: the dedicated trust and security search this index requires was not run in this pass, and no attestation was encountered incidentally.

The control set described by third parties, encryption, access control and audit logging alongside workstation and access restrictions for documentation specialists, addresses the right risks and is not the same as an assessed certification. A buyer should ask which attestations the parent holds and whether they cover this subsidiary.

No device pathway applies and none is claimed. Documentation is administrative.

The regulatory question specific to a human in the loop model is employment and location rather than device law: a documentation workforce handling protected health information across jurisdictions engages data transfer rules and the terms of the business associate arrangement rather than any clinical regulator.

Nothing published on evaluation, monitoring, error rates or performance variation.

The governance question created by the tiering is one no other record in this index raises. If accuracy depends partly on how much human review a customer purchased, then note quality varies by contract tier, and the patients seen in the settings that bought the cheaper automated tier are documented by a system with less checking. Nothing published compares accuracy across tiers, and that comparison is the one a buyer choosing between them most needs.

The speech equity exposure applies as it does to every product here, and the human element cuts both ways: a specialist may catch what a model mishears, which is a real mitigation, and it is undescribed and unmeasured.

Two passes located no accuracy or error figure for the automated tiers, no published limitations, no warranty, indemnity or remediation commitment. The productivity claims carried at group level, around two hours of documentation time saved per physician per day and a reduction of more than 80 percent, are published without a methodology, a baseline or a cohort, and they describe combined product families rather than a specific tier, so they cannot be attached to whatever a particular customer buys.

The recourse question also has a dimension here that software only scribes do not raise, and it concerns the patient rather than the buyer. A person other than the treating clinician hears or reads the encounter as part of how the product works. Whether the patient is told that, and by whom, is a consent question rather than a courtesy, and nothing published addresses it.

Where a recorded conversation is reviewed by a specialist in another country, the patient has no relationship with that party, no way to know it happened and no route to raise it, which is the harmed party problem this axis tracks in its most concrete form. Establish what patients are told, who tells them, and what the vendor commits to when a note produced through the reviewed tier is wrong.

The note reaches the record during or immediately after the visit, which is the operative requirement for a scribe, and the parent platform is reported to integrate with more than 50 record systems including the major ones.

Independent guidance is candid that integration with the largest record vendors involves project work, workflow design and ongoing change management rather than a switch being flipped, and advises budgeting for it separately. Held at B because no interface standard or certification is published for this product specifically.

Cloud delivered, with capture through a smartphone or a lightweight headset rather than dedicated room hardware, which keeps deployment light on the clinical side.

What is not published is where processing occurs, where the documentation workforce is located, or how long audio is retained. For a service in which people in other places listen to consultations, residency is a staffing question as much as a hosting one, and neither half is disclosed.

The vendor publishes no price, and third party analysis is unusually explicit about both the band and its own uncertainty, which is worth using carefully.

Independent estimates place hybrid scribe services at roughly 300 to 700 dollars or more per clinician per month, with the automated tier positioned below that and the live tier above it, and state plainly that these are estimates rather than official figures, that no per clinician list price is published, that implementation and support fees are not itemised, and that post acquisition bundle pricing is undisclosed.

The structural point is the useful one. This product costs more than pure software scribes because people are included in the service, so a buyer is explicitly paying for oversight. That makes the comparison against a cheaper software only competitor a question about how much checking the organisation wants to buy rather than a straight price comparison.

Deepest where documentation is hardest. Stated roots are in emergency medicine, hospital medicine and specialty outpatient care, and there is a dedicated emergency department variant, which is unusual: most scribes in this index are built for the scheduled ambulatory visit and emergency documentation is a different problem, with interruptions, multiple patients in parallel and a legally consequential record.

Reach spans more than 20 major health systems and hundreds of sites of care, extended by a group purchasing agreement covering most United States acute care providers. Held at B because coverage is United States only and everything sits in documentation and chart preparation.