Yosi Health

A digital intake and scheduling platform with a voice layer added, and the models are the smallest part of what a buyer is paying for.

Almost everything here is workflow. Pre arrival intake, form mapping onto a practice's existing documents, discrete data transfer into the record system, eligibility transactions against payers, scheduling, payment collection, two way texting and survey management are all deterministic software. The company's own strongest differentiators are of that kind: not requiring practices to change their forms, and producing both structured data and a signed document packet from the same submission. Those are integration and document handling achievements rather than model achievements, and they are why customers describe the product as easy to adopt.

Autoscoring questionnaires sit in the same category on the evidence available. Presenting a screener by appointment type and totalling a score is configuration, not inference, and nothing published indicates otherwise.

The voice automation is the model claim and it is real but bounded. The company describes it as an always on layer for routine scheduling and triage that reduces hold times and routes urgent calls appropriately, explicitly without replacing the human touch. Handling a spoken call and deciding where it should go requires speech recognition and intent classification, and that is genuine model work.

Graded D rather than C, one band below Clearwave, because the voice capability is described in a partner marketplace announcement rather than as a headline product with volumes attached, and no scale, deployment or performance figure for it was located anywhere.

The autonomy is modest, the framing is honest, and one capability is described in language the company should be more careful with.

The honest part first. The voice automation is positioned as an always on layer for routine scheduling that reduces hold times, and the company states explicitly that it works without replacing the human touch. Bounding an automation claim in the announcement that introduces it is unusual and it is the correct instinct. Check in, intake and payment are patient led self service rather than autonomous decision making: the patient does the work, the software collects it, and no model decides anything about them.

The careful part is the word triage. The company describes the voice layer as handling routine scheduling and triage and ensuring urgent calls are routed appropriately. Triage is a clinical term with a specific meaning, and two records in this index, Keona Health and Clearstep, use it for products built on licensed telephone triage protocols with defined escalation levels. Nothing here indicates protocol content, acuity logic or clinical review behind the routing. If what is meant is call routing by topic, then triage is the wrong word for it. If acuity is genuinely being assessed, then the absence of any described clinical basis is a material gap. Either way a buyer should establish which it is.

Nothing published describes escalation criteria, whether a caller can reach a person on request, coverage hours for human fallback, or what the voice layer does with a caller describing something urgent outside office hours.

Autoscoring carries a smaller version: a screener score presented to a clinician is an input, and nothing describes whether thresholds trigger anything automatically.

Graded C.

The platform mechanics are described with unusual clarity and the models are not described at all.

On the mechanics the company is specific and the specificity is the selling point. Form mapping is explained as a proprietary platform capability that maps patient answers onto a practice's existing forms so nothing has to be rebuilt, producing both discrete data into the record system and a completed, signed and dated document packet. Check in is described across three distinct mechanisms, tablet kiosk, quick response code and location detection. Intake presentation is described as varying by appointment type, patient category, scheduling provider and specialty. Integration is characterised as direct, real time, two way and fully automated, without a separate dashboard. A buyer can picture exactly how this works.

On the models nothing. The voice automation is named as a capability in a partner announcement and never described: no architecture, no provider, no scale, no accuracy, no versioning, and no explanation of how routine scheduling is distinguished from an urgent call that needs routing. That last mechanism is the one carrying the clinical weight of the triage claim, and it is entirely opaque.

Autoscoring is likewise undescribed, with no statement of which instruments are supported or whether scoring follows published scales.

The asymmetry is the finding. This is a vendor that explains its integration architecture in more detail than most and says nothing whatever about the one component that makes it eligible for this index.

Graded D.

No upstream model dependency is named, and for a seed stage company the presumption runs strongly toward integration rather than in house development.

The voice automation requires speech recognition, intent handling, dialogue management, speech synthesis and telephony. A company of this size, funded at seed and competing on integration breadth rather than on research, is very unlikely to have built that stack itself. Nothing names a provider for any layer, and each unnamed provider is a party handling patient call audio with its own retention and training terms. That gap also leaves the certification scope question open in a practical way: HITRUST and service organisation control certification cover the company's own environment, and whether an integrated voice provider sits inside that boundary is exactly what a buyer would need to establish.

Two further third party paths are implied by the product and never enumerated. Real time eligibility across payers runs through clearinghouse or transaction network infrastructure the company does not operate, carrying protected information outward. Card payment processing requires an acquirer and gateway, and payment card compliance is claimed without naming any of the parties in that chain.

One dependency category is named and it is the record systems, with six identified. That is integration rather than supply chain, and it does tell a buyer which external systems hold and receive their data.

The asymmetry on this record is worth stating plainly. This company publishes four security certifications, which shows it understands that enterprise buyers want verifiable third party assurance, and publishes nothing about the third parties inside its own product.

Graded D as an absence of disclosure rather than evidence of a problem.

Consistent operational figures and named integrations, all self reported, with genuine independent signals that are organisational rather than clinical.

The headline figure is stated consistently across the site at 84 to 85 percent of patients fully pre registering before arrival. Consistency across pages over time is worth something, and adoption is the sort of metric a practice can verify quickly in its own data. A documented 99.95 percent uptime figure is also published, which is rare in this category and is the kind of claim a customer would notice if it were false.

Customer evidence is qualitative and unusually specific in tone. Practices describe integration completed without pushback, customisation delivered as requested, improvements to patient flow and referrals, and responsive service. One review explicitly contrasts the product against a competitor on cost and service. Those are real practitioner voices rather than curated testimonials, and for a small vendor they carry more weight than a logo wall.

Independent signals exist and are about the company rather than the product: selection as a finalist in a healthcare innovation award programme in 2025, admission to a major cloud provider's advanced technology partner programme, and inclusion in a specialty record vendor's marketplace, which requires the partner to review the integration.

What is missing is measurement. No study, no controlled comparison, no error rate for the voice automation, no accuracy figure for form mapping or autoscoring, and no independent evaluation of any outcome. For a product whose voice layer routes urgent calls, the routing accuracy is the number that matters and it does not exist publicly.

Graded C.

Certified handling and no published stewardship position, on a data estate that is smaller than the neighbouring records but not small.

On handling the position is strong for a company this size. HITRUST certification is a validated assessment of controls over protected data rather than an assertion, payment card compliance covers the financial half, and both are stated together with the cloud partner programme credential. A practice's compliance officer has real material to work with.

On stewardship nothing was located. No retention period for intake submissions, insurance card images or demographic data, no statement on whether any of it informs product development, and no position on the voice automation's recordings or transcripts. The voice layer is the sharpest version of the question, because a call in which a patient describes why they need an urgent appointment is a symptom disclosure captured as audio, and audio cannot be de identified the way a form field can.

One design characteristic works in the vendor's favour and deserves credit. The platform is built to move data out rather than to accumulate it: intake syncs into the practice's record system in real time as discrete fields, and the stated purpose is eliminating transcription rather than building a data asset. A product whose value is delivering data to the customer's system has less structural incentive to retain it than a product whose value is analysing it. That is an inference from the design rather than a published commitment.

The document packet is a second unaddressed surface: a completed, signed and dated file is generated for each patient, and nothing describes where those live or for how long.

Graded C: certified handling, no published stewardship, on an architecture that points the right way.

The most complete certification stack of any small vendor in this session, with the contractual layer still unpublished.

Four credentials are claimed together and each is externally validated rather than self asserted: service organisation control type 1 and type 2 certification, HITRUST common security framework certification, payment card industry compliance, and admission to a major cloud provider's advanced technology partner programme, which the company notes carries stringent security requirements for entry. HITRUST is the credential health organisations ask for by name, and holding it alongside payment card compliance covers both halves of what this platform actually handles, since the product processes clinical intake data and card payments through the same interface.

That combination is the substantive point. Clearwave, on the neighbouring record, performs the same functions at far greater volume and publishes nothing. Yosi is a fraction of the size and has invested in exactly the credentials its data flows require.

The data flows are extensive and stated plainly. Real time two way integration writes intake data into the record system as discrete fields, insurance and demographic capture runs through the platform, and payments are collected pre and post visit. The vendor is unambiguously a business associate.

What is missing is the contract. No business associate agreement is offered or described, no terms are published, no protected data handling summary exists and no retention position is stated. Certification scope is also unenumerated, which matters less here than on a multi acquisition platform but still leaves a buyer unable to confirm the voice component sits inside the certified boundary.

Graded B.

Four externally validated credentials from a seed stage company, published plainly, with no apparatus behind them.

The stack is service organisation control type 1 and type 2 certification, HITRUST common security framework certification, payment card industry compliance, and advanced technology partner status in a major cloud provider's independent software vendor programme. Type 2 tests controls over a period rather than at a point. HITRUST is the credential healthcare buyers ask for by name. Payment card compliance covers the card data this platform handles directly. The cloud partner tier carries its own security review for admission. Together they cover both halves of what the product touches, clinical intake and payments, which is the right coverage rather than a collection of whatever was easiest to obtain.

The comparison that matters sits on the adjacent record. Clearwave performs the same functions at vastly greater volume, describing 150 million patient visits a year, and publishes no security posture at all. Yosi is a fraction of the size and holds four credentials. On this axis the small vendor is straightforwardly better, and a buyer weighing the two should notice.

What is missing keeps this from an A. There is no trust centre, no described process for requesting reports under agreement, no audit period or auditor named, no certification scope enumerated, no penetration testing statement, no vulnerability disclosure policy, no subprocessor list and no incident notification commitment. The credentials are asserted in marketing copy on product pages rather than presented through any verification route.

Graded B.

No clearance is claimed and none is required for the platform as described. Scheduling, intake collection, eligibility verification and payment processing carry no clinical claim, and autoscoring a questionnaire the patient completes is arithmetic on their own answers rather than an assessment of them.

One capability sits closer to a line than the rest and is worth recording precisely. Autoscored screening questionnaires configured by specialty are, in many practices, standardised instruments such as depression or anxiety screeners, and a score presented to a clinician is decision support. It remains comfortably on the informing side, because the clinician sees the patient's own answers alongside the total and the arithmetic is transparent, which is exactly the transparency that keeps such software outside device regulation. Nothing published states which instruments are supported or whether any scoring is proprietary rather than a published scale, and the answer would matter.

The regulatory regimes that plainly apply are payments and telephony. Card payment handling is inside the payment card industry standard, and here the company does claim compliance, which is more than the neighbouring Clearwave record offers. Two way texting and appointment reminders engage telephone consumer protection rules covering consent and revocation, and the voice automation engages the emerging state requirements that a caller be told when they are speaking to an artificial agent. Neither telephony regime is addressed anywhere located.

Consent capture through digital intake is a third area, and the platform's production of a signed and dated document packet is a genuinely good answer to it, since it creates the evidentiary artefact a practice needs.

Graded C.

Nothing was located. No model card, no training data description, no accuracy figures, no subgroup analysis and no bias statement.

The exposure is the same access mechanism identified on the neighbouring intake records, and it is sharpened here by one of the company's own stated markets.

Yosi names community health centres and federally qualified health centres among its buyers. Those organisations serve populations with the highest concentration of limited English proficiency, low digital confidence and unstable phone access in United States healthcare. A mobile first, pre arrival, self service design assumes a smartphone, data, literacy and comfort completing clinical forms unaided, and the 84 to 85 percent pre registration figure measures who completed the process rather than who struggled or who was left to the front desk. The 15 percent who do not pre register are not a rounding error in a safety net clinic; they are likely to be the patients with the greatest need.

The voice layer carries the recognition problem. Speech recognition performance varies with accent, dialect, age and speech rate, and a caller the system cannot understand, calling a practice that has reduced front desk staffing, may simply not reach anyone. Nothing published describes language coverage for either the voice automation or the intake interface, and nothing addresses accessibility conformance for the forms.

Autoscoring adds a third: if screening instruments are scored identically across populations for whom the instruments were differently validated, the score carries an assumption nobody has examined.

Graded D on the absence, with the safety net exposure named because the company's own market selection makes it concrete.

Nothing published addresses responsibility for an automated outcome, and the sharpest exposure follows from the company's own choice of words.

The voice automation is described as handling triage and ensuring urgent calls are routed appropriately. If a caller with an urgent problem is routed as routine and comes to harm through the delay, the question of who carries that is unaddressed, and the vendor has used clinical language to describe the capability without describing any clinical basis for it. That combination is the specific liability concern on this record: a claim strong enough to create reliance, with no published protocol content, escalation logic, or position on responsibility behind it.

The other exposures are financial and administrative. An eligibility check returning wrong coverage produces a denied claim or an unexpected bill. Payment collection pre and post visit creates overcollection and dispute handling questions. Form mapping onto a practice's existing documents is a transcription function performed by software, so a mapping error puts a patient's answer in the wrong field of a signed legal document, and the platform generates that signed document automatically.

That last one deserves emphasis because it is unusual. The signed and dated packet is presented as a benefit, and it means an automated mapping error is preserved in an executed record rather than caught by the person who would otherwise have typed it.

No indemnity, limitation, performance warranty, accuracy commitment or recourse route was located, and no error rate is published for mapping, autoscoring or voice routing. The published uptime figure is the only service commitment of any kind on the record and it covers availability rather than correctness.

Graded D.

The best record system coverage of any small vendor in this index, and the direction of flow is what earns the grade.

Six record systems are named for integration: Epic, Oracle Health, Allscripts, athenahealth, AdvancedMD and Elation, spanning enterprise, ambulatory and independent practice systems. Each carries its own dedicated integration page rather than appearing on a logo list, which indicates real implementation work rather than aspiration.

The integration is bidirectional and writes discrete data. Intake forms sync into the record system in real time as structured fields, eliminating transcription and scanning at the point of care, while patient demographics and history flow the other way to populate the pre arrival experience. Writing discrete clinical intake into the chart is the hard half, and it is what separates this from a forms product that emails a document to the practice. The company also produces a signed and dated document packet alongside the structured data, which covers the medico legal requirement most structured data approaches drop.

The design principle stated is worth crediting: the interface requires no separate dashboard, so staff work in their record system rather than in a second application. Vendors routinely underestimate how much that determines adoption.

Marketplace presence adds independent confirmation, with the platform admitted to a specialty record vendor's marketplace in October 2025 following that vendor's review.

What is absent is any published application programming interface or named interoperability standard, so a customer wanting to build on this data has no documented route. That is a small deduction against genuinely broad and deep coverage.

Graded A.

Two useful facts are published and the specifics are not, which places this above most records in this session on an axis where almost nobody discloses anything.

The first fact is the infrastructure relationship. The company states admission to a major cloud provider's advanced technology partner programme, describing it as a highly selective programme with stringent security requirements for entry. That identifies the underlying platform and, more usefully, indicates the provider has reviewed the architecture rather than merely hosting it.

The second is availability, and it is the only such figure in this session. A documented 99.95 percent uptime is published and repeated across the site. That matters operationally for exactly the reason it matters on the neighbouring record: when intake, eligibility and payment are down, a practice that has reduced front desk staffing has no manual fallback at scale. Publishing a specific figure rather than a vague reliability claim is a commitment a customer can hold the vendor to.

What is not stated is everything granular. No region, no residency option, no subprocessor list, no retention position, no export or contract end position, and no disaster recovery description.

The deployment surface is lighter than Clearwave's because the design is mobile first. Patients use their own devices for most interactions, with tablet kiosks as one option among three check in mechanisms rather than the primary estate, so there is materially less vendor hardware sitting unattended in waiting rooms. Nothing published describes tablet provisioning, session handling or local data persistence for the kiosks that do exist.

Graded C.

Nothing is published by the vendor. No rate card, no unit of charge, no tier structure, no implementation fee, no contract term and no minimum was located on the company's own site.

Third party software directories list the product and carry pricing sections, which is the usual route by which a small vendor's pricing becomes semi public, and no figure was retrievable from them either.

The unit of charge is unstated and matters here for the same reason it does on the neighbouring Clearwave record: the platform collects pre and post visit payments, so whether the vendor earns a percentage of collections, a flat fee, or both is the question that determines both cost and incentive alignment. Payment card industry compliance is claimed, which confirms card data is handled directly rather than passed to a separate processor the practice contracts with, so the commercial relationship around payments is real and undescribed.

One genuine commercial signal appears from an unexpected direction. A customer review comparing Yosi against a competitor cites that competitor's cost as a reason for switching, which indicates the product competes on price in this segment. That is a third party observation rather than a vendor disclosure and it does not substitute for a number.

The absence is more consequential for a small vendor than a large one, because the buyers in this segment are independent practices making a discretionary purchase from an operating budget, and they are precisely the buyers least able to run a lengthy procurement to discover a price.

Graded D.

Wide organisational coverage for a company this size, with the intake layer genuinely configurable by specialty.

The stated buyer types run across specialty clinics, urgent care centres, primary care, health systems, and community health centres and federally qualified health centres. The last of those is worth noting: safety net organisations are a segment several competitors in this session ignore, and their needs around language, low digital confidence and complex coverage are distinctive.

Specialty depth sits in the intake configuration rather than in clinical content. Forms and screening questionnaires with autoscoring are configurable by appointment type, patient category, scheduling provider and specialty, so an orthopaedic new patient and a dermatology follow up receive different questionnaires scored differently. That is real specialty adaptation and it is the same depth Clearwave offers.

Channel coverage is broad and deliberately mobile first: self scheduling from the practice website, pre arrival links, tablet kiosks, quick response codes, location based check in, two way texting and telehealth. Covering both the phone in the patient's pocket and a kiosk in the office is more than most small vendors manage.

What holds this at B is scale rather than design. No large health system deployment was located despite health systems being a named segment, no customer count or visit volume is published, and the company is seed stage with a small team. The coverage claim is credible for practices and unevidenced at enterprise scale.

Graded B.