Wavo Health

Speech recognition, note generation, the clinical assistant and patient material generation are all model output, with no services layer or legacy platform underneath.

The gate is stated as a three step sequence rather than an implied one: review, edit, then APPROVE the generated note before it is used. Naming approval as a distinct action after review is a small but meaningful precision, since it makes acceptance deliberate. Held at B because no acceptance rate, edit burden figure or confidence threshold is published, and because the clinical assistant answering medication and coding questions carries no described limits of its own.

Both performance claims are unfalsifiable as written: industry leading accuracy, which cites no comparison, and saving up to 50 percent of documentation time, which the up to construction satisfies at any value. No accuracy figure, model card, named models or evaluation methodology was located. The specialty vocabulary work described below is real capability rather than measurement.

The vendor closes a gap that several peers in this lane leave open, and it deserves specific credit before the criticism. It names the de identification methods it uses, and both are the recognised statutory routes. Two other records assessed in this lane state that de identified data may improve their systems and define de identification nowhere, which leaves the whole commitment resting on an undefined word; this vendor has actually answered it.

What sits against that is placement rather than substance. The homepage states plainly that patient data never trains the artificial intelligence, while a security page describes how data is de identified for training and improvement.

Both can be technically true, since information de identified by either statutory method is no longer protected health information, but the reconciliation requires a reader to know the rule well enough to supply it themselves, and most clinicians buying this product will not.

A headline commitment saying customer content never trains the model, sitting on a different page from a description of how customer content is prepared for training, is materially misleading to the audience it is written for. On enumeration there is nothing: no model, model family or provider, no hosting arrangement, no sub processor list. Ask which statement governs in the contract, and for the sub processor list.

Testimonials only, and unnamed ones, including a clinician reporting an hour long complex visit documented in two minutes. No study, controlled evaluation, accuracy benchmark, third party rating, named customer or deployment count was located.

The vendor publishes more on this axis than the earlier pass found, and two of its statements do not sit together.

The homepage states plainly that patient data never trains AI. A security page states that for AI training and improvement, data is rigorously de identified according to the safe harbour or expert determination methods.

Both can be technically true, since information de identified by either method is no longer protected health information and arguably no longer patient data. That reconciliation requires a reader to know the rule well enough to supply it themselves, and most clinicians buying this product will not. A headline commitment that customer content never trains the model, sitting alongside a page describing how customer content is prepared for training, is materially misleading to the audience it is written for. The fix is small: say that de identified data may be used for improvement, and say it in the same place.

One element deserves real credit and it is the thing this index has asked several peers for without success. The vendor names the de identification methods it uses, and both are the recognised statutory routes. Two other records in this lane state that de identified data may improve their systems and define de identification nowhere, which leaves the commitment resting on an undefined word. This vendor has closed that gap.

What is still absent is retention. No schedule for audio or transcripts and no deletion commitment were located.

Ask which statement governs in the contract, and for the retention schedule.

HIPAA compliance stated consistently across product and pricing materials. Business associate agreement terms are not published for inspection, and the product is self serve with a free trial so use can begin before any agreement conversation.

No named or dated attestation was located. A dedicated security page exists and describes the posture as enterprise grade, which names no standard and belongs with the constructions this index treats as marketing rather than assurance.

What is concrete is better than the tier norm and worth crediting. The vendor names its encryption standard rather than saying encrypted, states that it readily enters into business associate agreements with covered entities, describes adherence to both the United States and Canadian federal privacy regimes, and names the statutory de identification methods it applies. Those are checkable commitments rather than adjectives.

What is missing is anyone independent having examined them. No report of either type, no certification, and no penetration testing statement.

The buyer profile is the reason to press rather than excuse. This product is sold on a self serve free trial to solo practitioners and small practices alongside a custom enterprise tier. The self serve buyer has no procurement function, so nothing in the adoption path asks these questions, and published evidence is the only mechanism. The proportionality argument that a small vendor can defer certification because sophisticated buyers will ask directly does not apply where the model is designed so that most buyers never ask.

The vendor also publishes guidance on questions worth asking before choosing documentation software, which puts it in the group that has articulated the standard. It has answered more of its own checklist than most, and not this part.

Ask what external testing has been performed, and whether an attestation is scheduled.

No clearance claimed and none required for documentation. No United States device pathway attaches to a note the clinician reviews and exports, and the vendor describes strong editing controls as part of its workflow.

The flag from the earlier assessment stands and is the substance of this axis. Alongside the scribe the vendor operates an AI clinical assistant described as giving instant, evidence based guidance in response to questions, including about medications. Answering a clinician's medication question is advice rather than record keeping. It sits in the territory occupied by clinical reference products, and the reasoning that keeps such tools outside device oversight depends on the professional being able to see and evaluate the basis of an answer before relying on it.

So the questions are about grounding and presentation. What source material does the assistant draw on, is it a curated clinical knowledge base or a general model answering from training, are citations shown, and how current is whatever it draws from. Medication guidance in particular ages: interaction data, contraindications and dosing change, and an answer generated from a static model without a dated source is a different object from a monograph lookup.

One piece of restraint deserves recording here as elsewhere. Coding cues are offered without any revenue capture framing, which distinguishes this record from most of the category.

Markets appear to span the United States and Canada, and the vendor names both federal privacy regimes, so a buyer should establish the position under provincial law as well.

Ask what the assistant is grounded in, whether it cites, and how currency is maintained.

No fairness statement, subgroup analysis, accent or dialect performance disclosure or evaluation methodology was located, and no language coverage is claimed despite the site being structured for multiple locales.

The locale structure is the specific thing to press. A site built with language paths implies either an existing multilingual user base or a plan for one, and both raise the question of how the product performs on speech in languages other than the one it was principally developed for. Nothing published addresses it, and a buyer cannot even establish which languages are supported.

The stated user base widens the question further. The vendor describes supporting physicians, mental health clinicians, allied health professionals and clinical teams. Those settings differ in vocabulary, session length and speech character, and mental health sessions in particular present the harder acoustic and linguistic material this index has noted elsewhere: emotional speech, silences, and clinical terminology that general models handle poorly.

One point from the earlier assessment is worth preserving and stating positively. Coding cues are offered with no revenue lift or capture language attached anywhere in the material. That keeps this vendor off the coding gradient this index tracks across the category rather than at its mild end, and it is a deliberate looking restraint given how consistently competitors reach for that framing.

The realistic ask is scaled to the company. Not a published subgroup study, but a statement of which languages are supported and what the underlying speech model is, since a commercial speech provider may already publish performance data the vendor could point to.

Ask for the language list and the speech model.

Two passes located no accuracy or error figure, no published limitations and no warranty, indemnity or remediation commitment. Both performance claims are unfalsifiable as written and they fail in different ways, which is worth separating because buyers read them as equivalent. Industry leading accuracy is a comparative assertion made without publishing a measurement of this product or of any competitor, so it claims a ranking that rests on nothing a reader can check.

Saving up to 50 percent of documentation time is satisfied by any result at all, including none, so it commits the vendor to nothing while reading as a quantified benefit. Neither tells a buyer where the system degrades or what to monitor for.

The vendor does demonstrate real domain understanding elsewhere in its material, and demonstration is not measurement: knowing that a phrase is ambiguous in a specialty shows the problem has been understood, not how often the system resolves it correctly. The de identification disclosure credited on the supply chain axis is also not a commitment about output. Ask for an accuracy figure with a defined denominator and reference standard, and for what the vendor commits to when a note is wrong.

The earlier assessment recorded the absence as total and said it determined whether this is a documentation tool or a parallel record. The second pass answers that, and the answer is the second one.

The vendor lists exports among the outputs it supports, alongside templates, coding cues, patient context, summaries and letters. Nothing describes a connection to a record system: no named platform, no interface standard, no write back mechanism, no marketplace listing. So documentation is produced inside this product and the clinician moves it out.

That is a legitimate model and it should be stated plainly rather than treated as a deficiency, because it suits the buyer this product is sold to. A solo practitioner or small practice adopting a tool on a free trial is not going to commission an integration, and an export based workflow means nothing has to be configured before value arrives.

It has consequences a buyer should hold in view. The product accumulates patient context and visit history in its own store, so a second clinical record exists alongside the official one, with its own retention, access and export questions. Notes not transferred remain only in this system. And the vendor's environment holds the encounter for whatever interval elapses before the clinician moves the output, with no institutional visibility of what was captured.

The practical questions are therefore about exit rather than integration. What formats are supported, can everything be extracted at once rather than note by note, does the export include the accumulated patient context, and what happens to the store if the subscription lapses.

Ask whether any record system integration is planned, and what the bulk export path is.

The vendor addresses this axis directly, which most of this category does not, and the answer contains a hedge that has to be read carefully.

Its security material states that it is transparent about where data is stored, that data for United States clients is typically stored within the United States, and that specifics can be found in the agreement and terms of service. Pointing to the contract for the definitive answer is the right instinct and worth crediting: that is where a residency commitment becomes enforceable rather than aspirational.

Typically is the word to press. It signals that United States storage is the normal case rather than a guarantee, and it leaves unstated what the exceptions are, when they apply, and whether a customer is told. A residency commitment with an undefined exception is not one a privacy impact assessment can rely on. Ask for it without the qualifier, in the agreement.

The second gap follows from the vendor's own compliance claims. It states adherence to both the United States rule and the Canadian federal privacy statute, so Canadian customers exist. Nothing states where their data sits. This index has recorded the same shape repeatedly: a residency promise naming one jurisdiction establishes nothing about the others, and Canadian buyers face provincial expectations that differ again by province.

Also unresolved: no subprocessor list, and no model provider. The vendor operates an AI clinical assistant answering medication questions alongside the scribe, so more than one workload may be involved.

Ask for residency per jurisdiction without the hedge, the subprocessor list, and the model provider.

Partial. A seven day free trial of the Pro tier is published along with a pricing page and plan FAQ, and custom enterprise plans are described for larger practices, hospitals and health systems. No rate, tier figure or per clinician price was retrieved in this pass, so a buyer knows the shape of the offer and not the cost.

Fewer specialties than several competitors claim, specified far more concretely than most. Rheumatology support documents tender and swollen joint counts and assists with DAS28, CDAI and RAPID3, which are validated disease activity instruments rather than note headings. Oncology handles TNM staging, chemotherapy, radiation and immunotherapy regimens, and tumour board summaries.

Internal medicine reaches inpatient territory with admission histories and physicals, daily progress notes, discharge summaries and transitions of care, which most outpatient scribes do not attempt. Gastroenterology and physical therapy with functional outcomes are also covered. Naming the instrument rather than the specialty is the mark of genuine depth, in the same class as Marvix's laterality propagation and Lime's OASIS handling. Held at B because the specialty list is short and no language coverage is published.