Marvix AI

Ambient capture, pre charting synthesis, note generation, letter production and coding are all model output, with no services layer or platform business underneath.

Notes and letters arrive chart ready for clinician review, and the coding path is more transparent than most: E/M levels, modifiers and add on codes are presented with the medical decision making rationale supporting them, so a clinician can see the basis for a suggestion rather than accepting a bare code. That is the traceability principle this index credits in Abridge and Lime Health, applied to coding. Held at B because no acceptance rate, edit burden figure or confidence threshold is published, and no explicit sign off gate is described.

Names its techniques rather than gesturing at them, which is uncommon at this size. Personalisation is attributed to neural style transfer, a specific method rather than a claim that the system learns your style. Coding output carries explicit medical decision making rationale, so the reasoning behind an E/M level is inspectable rather than asserted. Specialty behaviour is described concretely, including laterality propagation in orthopaedics, which is the kind of detail only a team that has worked the specialty would name.

Held at B because no accuracy figure, model card or evaluation methodology was located.

The boundary is drawn better than most in this tier and no party is named. Two statements should be read together. The first is that customer protected health information is not used to train shared models, where shared would be a qualifier worth challenging on its own, since it leaves customer specific models outside the commitment. The second closes it: training and improvement workflows are stated to require explicit customer permission before any data use occurs.

Taken together that is a permission gate over training rather than a carve out, which places the decision with the customer and is a better position than most of this category offers. The cleaner formulation would drop the qualifier altogether, and it belongs in the contract rather than in marketing material.

On enumeration there is nothing: no model, model family or provider, no hosting arrangement and no sub processor list, despite genuinely specific technical description elsewhere on the record. The volume point matters here. Pre charting ingests prior notes, laboratory results, imaging and medications and the vendor describes surfacing three years of patient history, so whatever unnamed parties sit in this chain are handling far more historical content than a capture only scribe would send them. Ask for a sub processor list and for the permission gate wording in the agreement.

No study, controlled evaluation, accuracy benchmark, third party rating, named customer or deployment count located. Note also that a substantial share of the comparison content ranking for competitor names in this category is published by Marvix itself, so search visibility here reflects content investment rather than market presence.

The disclosures this axis measures are largely present, which is unusual for this tier, and the earlier assessment's concern about volume remains the right thing to press.

On training the vendor makes two statements that should be read together. The first is that customer protected health information is not used to train shared models. Read alone, shared would be a qualifier worth challenging, since it leaves customer specific models untouched by the commitment. The second closes it: model training and improvement workflows are stated to require explicit customer permission before any data use occurs. Taken together that is a permission gate over all training rather than a carve out, and it is a better position than most of this category offers. The clearer formulation would drop the qualifier entirely, and a buyer should ask for that wording in the contract rather than the blog.

On retention the position is real but softer than the best in this lane. Retention is described as configurable, with data minimisation and controlled retention practices, and audio handling following configured policies which may include automatic deletion after processing or storage for a defined period. Configurable without a stated default leaves the buyer to discover what happens if they configure nothing, and may include is optionality rather than commitment. A peer in this category publishes a specific default period and names the standard used to destroy data at the end of it. That is the bar.

The volume concern stands. Pre charting ingests prior notes, laboratory results, imaging and medications, and the vendor describes surfacing three years of patient history. So this product holds far more historical protected health information than a capture only scribe, and the retention policy should be understood to cover ingested chart data as well as encounter audio.

HIPAA compliance stated consistently across product and store listings. Business associate agreement terms are not published for inspection.

The earlier assessment recorded no attestation located. That is overturned. The vendor states it holds SOC 2 Type II alongside health privacy rule compliance, and describes encryption at rest and in transit, role based access permissions and logging of all access.

What lifts this above a claimed logo is that the vendor demonstrates it understands what the attestation means. Its published material explains that the Type 2 audit covered its controls across a full audit period rather than at a single point in time. That distinction is the entire difference between the two report types, it is the thing most vendors in this category blur, and several elsewhere in this lane assert compliance while their own guidance shows no grasp of it.

This record is therefore the counter case to a pattern that has recurred through this category. Three vendors in this lane publish buyer guidance naming SOC 2, HITRUST or ISO 27001 as the test, and hold none of them. This one publishes comparable guidance and meets the standard it sets. That is worth stating plainly, because the index should credit the vendor that answers its own checklist as readily as it flags the ones that do not.

Two things hold it short of the top grade. No trust centre carrying a dated report, an audit period or a scope statement was located, so a counterparty still cannot see what the report actually covers, and for a platform spanning documentation, pre charting from historical records, and coding, scope is the substantive question. And no ISO 27001 or HITRUST was located alongside it.

Ask for the report, its period, and the systems in scope.

No clearance claimed and none required for ambient documentation. No United States device pathway attaches.

What governs is payment integrity, and the coding surface here is broader than the earlier assessment recorded. The product is described as automatically generating procedural codes, diagnosis codes, evaluation and management levels, hierarchical condition category and risk adjustment factor scores, and full medical decision making rationale alongside the documentation.

Two of those deserve separate attention.

Risk adjustment scoring determines payment for a patient population under risk bearing arrangements. A published policy analysis found ambient scribes producing measurable increases in documented risk adjusted diagnoses per encounter and treated whether that reflects better documentation or more intensive coding as the open question for this category. A documentation product that calculates the score itself is operating on that mechanism directly rather than incidentally.

Generating the medical decision making rationale is the sharper item, and it is the second instance of this pattern in the lane. The rationale is the text that justifies the evaluation and management level selected. Where the same system both selects the level and authors the reasoning defending it, the rationale cannot serve as an independent check on the selection, because it was written to support it. Ask what the rationale is derived from and whether it can ever fail to support the level chosen.

The product also generates after visit summaries, referral letters and insurance documents. After visit summaries reach patients directly, so establish whether a clinician reviews them before they are sent.

Markets appear to be United States only, across a stated coverage of more than a hundred specialties and subspecialties.

On the coding gradient with a real counterweight. The gradient language is present, with marketing about improving E/M accuracy and capturing modifiers, and modifier and add on coding is where billing scrutiny concentrates. Against that, every code is delivered with explicit medical decision making rationale, which is the single most useful control on this axis because it converts a coding suggestion into an argument a clinician or auditor can evaluate rather than a number to accept. No fairness statement, subgroup analysis or accent and dialect disclosure was located, and no language coverage is claimed.

The mechanism that carries this grade is inspectable reasoning attached to the output that matters most. Coding output arrives with explicit medical decision making rationale, so the basis for a recommended evaluation and management level can be examined rather than accepted, and a coder or clinician can refuse a recommendation on its reasoning.

That is the same principle credited elsewhere in this index for mapping generated content back to its source, and it is more valuable in coding than in note text because a level recommendation determines what is billed and the exposure for a wrong one sits with the practice. Two smaller disclosures support the grade.

The personalisation technique is named specifically rather than described as the system learning your style, which tells a buyer what kind of adaptation is occurring, and specialty behaviour is described at a level of detail that indicates real domain work rather than templating. What is missing is any measurement or commitment.

No accuracy figure, error rate or evaluation methodology was located for either the note or the coding output, and no warranty, indemnity or remediation commitment was found. Ask for a coding accuracy result with a defined denominator, and for what the vendor commits to when a level recommendation is wrong.

Named, broad and genuinely two way. Integrations are listed across athenahealth including Athena Flow and Athena Practice, Veradigm, AdvancedMD, eClinicalWorks, ModMed, DrChrono and NextGen, and the direction of travel is both ways: the product pulls patient history, prior notes, labs, imaging and medications to build the pre charting recap, then pushes structured notes, letters and codes back. Read access at that depth is a harder integration than note push and is what makes the longitudinal proposition work. Release notes show integrations being actively extended rather than listed once.

No hosting region, residency option or subprocessor list was located, and nothing establishes whether a third party model service processes the encounter or what it retains.

What is better established than the earlier pass found is the integration mechanism, and it deserves crediting because a large part of this lane reaches the chart by less controlled means. Data exchange with the record system is described as happening through secure application interfaces, with section level field mapping, schedule synchronisation and structured note delivery into named systems. That is a permissioned integration rather than a browser extension or robotic process automation driving the user interface, so its actions carry their own identity and the record system's access controls apply to them. Five vendors in this lane reach the chart the other way.

The integration is also bidirectional in a way that matters for this axis. The product does not only write notes out; it reads historical records in, pulling prior notes, results and medications to build pre visit summaries spanning years of history. So the data flow runs both directions and the volume moving toward the vendor is considerably larger than an encounter recording.

That makes the missing residency detail more consequential rather than less. A buyer is not only asking where today's conversation is processed but where a multi year extract of their patients' records is held while summaries are generated, and for how long.

Encryption in transit and at rest, role based access and access logging are described and are the right controls, and they are covered by the vendor's attestation, which is more than most peers can say.

Ask for the hosting region, the subprocessor list, and the model provider.

No rate card. Pricing is described as customised by practice size and specialty, which is a quote process. What is published and genuinely notable is the trial: 30 days with FULL feature access AND full bidirectional EHR integration included. Most trials in this category withhold the advanced workflow features that determine whether a product works in a specialty practice, and Marvix points that out itself. A trial that includes the integration is the closest thing to an honest evaluation a buyer can get without a contract.

Among the deepest specialty coverage in this index and specific about what depth means. A claimed 135 or more specialties and subspecialties, with named deep support for neurology, oncology, orthopaedics and nephrology, and capabilities that only matter inside those specialties: laterality propagation for orthopaedic documentation, oncology staging and treatment plan handling, and composite notes for patients whose history spans years rather than one visit.

It also covers multi user workflows across medical assistants and providers with parallel rooming, vitals and pre charting, which reflects how specialty clinics actually run. Output extends to after visit summaries, referral letters and patient instructions.