Verbal

Here the models are the product, with nothing underneath them to sell. Verbal audits calls, video visits, chart notes and chat against an organization's own protocols and against external frameworks such as The Joint Commission, CHAI and URAC, for human staff and for AI scribes and agents alike. Judging whether an interaction met a protocol, or whether a note is complete and accurate, is the model's work; the Healthie integration page adds that the same AI generates clinical documentation such as SOAP, SBAR and therapy progress notes.

What surrounds the models exists to feed or display them: a desktop app that captures phone and video interactions, a Chrome extension, dashboards and search over findings. Conversational measures such as speaking pace are signal processing rather than learned judgment, and they are a minor part of what is sold. With the models removed, a buyer would hold call capture and an empty dashboard.

Nothing is decided for a patient; the product scores interactions and staff, and people act on the findings. Audit findings go to compliance and clinical operations leads as searchable, sortable results, and the company has described a step in which the customer validates a model's accuracy before it is released to users, which puts a human acceptance gate on the model itself. Staff can see their own scores; a contact center agent quoted on the home page describes looking at her score and working out what she did wrong.

What is missing is the other side of the score. Nothing says whether a staff member can contest a finding, whether scores inform employment decisions, or what the model does when a protocol is ambiguous. A customer quote on the home page says the checklist guarantees no subject goes missed on a call; the subject of that sentence is the staff member's coverage of topics, and it is the customer's sentence rather than the company's.

The mechanism is described in functional terms and nothing in it is identified. The company describes audits customized to each organization's protocols and to external frameworks, applied to calls, video visits, chart notes, chat and AI agent output in real time and retrospectively; automated redaction of personal identifiers as the first step in processing; and note generation for customers using the Healthie integration. It has described training models per customer against that customer's requirements, with the customer validating accuracy before release.

No model or provider is named, no aggregate accuracy figure is published, the validation method is not described, and nothing says how model changes reach a customer or what the model does when a protocol is ambiguous. Transparency, as described, runs to each customer about its own model rather than to the public.

Every component is described by function and none by name. The company describes a desktop app and Chrome extension capturing interactions, redaction of personal identifiers as the first processing step, transcription and audit models customized per customer, and production data hosted in the United States and managed by United States based staff. No hosting provider, speech recognition provider, language model provider or subprocessor list is published.

A product that transcribes calls and reads notes and chat across an organization almost certainly relies on at least a cloud host and a speech recognition service, and each is a party that may see content before or after redaction, under retention and training terms that are not published.

Vendor reported results with a stated source, and no measurement of the audit itself. The home page reports cost reductions of up to 90 percent from replacing manual audits and adherence improvements of up to 79 percent within 90 days, stating both are based on results from its customers in 2025, with no customer count, baseline or method. Quotes from named staff describe time saved and training value, and Fred Brown, a former chairman of The Joint Commission, is quoted as an adviser to the company. The company was named a qualified solution on Mayo Clinic Platform in October 2025 (as the company announced), which reflects that program's vetting rather than measured performance.

The number that matters most for a compliance product, how often its judgment agrees with a qualified human auditor, is not published, and no independent evaluation appears. The company was founded in 2024.

A categorical commitment on de identification, with retention and training left open. The security page says personal identifying information is removed from every transcript and that redaction is always the first step in processing, that access to identifying information is strictly controlled, and that production data is always hosted in the United States and managed by people based there. Data is encrypted in transit and at rest.

The exposure is wide: the product reads calls, video visits, chart notes and chat across an organization, continuously rather than on a sample, and retrospective audit requires keeping them. Nothing published states how long recordings, transcripts and notes are kept, whether customer interactions train models beyond the customer's own, or whether models trained for one organization are isolated from another's.

Every customer gets a business associate agreement, the company says, and a service description backs that up; the agreement itself is unpublished. The security page says the company signs business associate agreements with all its customers, that the platform is HIPAA compliant and HITRUST certified, that identifying information is redacted from every transcript as the first processing step, and that production data is hosted in the United States and managed by staff there. That states the status, the tier (every customer) and how protected information is handled in the service.

Nothing states which subcontractors the agreement binds, although transcription and model processing are likely to involve them; the terms page could address it. What the redaction step removes, and what it leaves, is not described.

HITRUST certification is claimed as the company's own, with nothing a buyer can open. The security page states that the company is HITRUST certified, that it regularly undergoes comprehensive third party audits of its security, and that regular penetration tests are run; it also describes encryption in transit and at rest, strict access controls, continuous monitoring and redaction of identifying information.

The HITRUST assessment type, date and scope are not stated, no report or trust center is published, and the penetration testing firm and cadence are not named. No SOC 2 report is mentioned.

No clearance, authorization or submission is claimed and none is needed. Auditing interactions and documentation against an organization's own protocols is administrative, and the notes the product generates through its Healthie integration are documentation rather than a diagnosis or treatment recommendation.

The regimes that matter are the ones the product sits inside: its outputs feed billing compliance, accreditation readiness against frameworks such as The Joint Commission and URAC, and regulatory adherence, so the accuracy of its judgments has consequences under payment and accreditation rules rather than device law. The failure is asymmetric. A false flag costs a reviewer's time; a missed one tells an organization it is compliant when it is not, and the organization carries that liability whatever the software concluded.

Nothing says how findings are positioned as accreditation evidence, or what happens when an audited interaction raises a patient safety concern.

Principles and one described control, with no evaluation published. The company says audits are customized to each organization's protocols and can apply external frameworks such as The Joint Commission, CHAI and URAC, and it has described letting the customer validate a model's accuracy before release, which is a governance mechanism. No subgroup performance, testing method, model documentation or drift monitoring is published.

The exposure has a particular shape because the model judges people. If transcription is weaker for a clinician or patient with a given accent or first language, the resulting compliance score is depressed for reasons unrelated to conduct, and the person carrying the score cannot see why. The same applies when the product audits AI agents and scribes: an auditor that shares the weaknesses of the system it audits will not catch them.

Findings can be checked and nothing stands behind them. Audit results are searchable and sortable down to the single interaction and staff member, and the underlying call, note or chat is kept for review, so a compliance lead can check a finding against its source. The company has described a customer validation step before a model is released.

No accuracy figure, warranty or remediation commitment is published; the terms may contain one. The failure that matters is asymmetric: a false flag costs a reviewer's time, while a missed breach tells an organization it is compliant when it is not, and the home page's own pitch is that a missed protocol or inaccurate note can cost millions. Nothing says who carries that when the audit missed it, or how a staff member wrongly scored can have the finding corrected.

One named record system with writing and reading described, through its marketplace. Verbal is listed on the Healthie app marketplace, and its integration page says notes and documentation generated by Verbal pass to the Healthie record, and that the Healthie scheduler syncs to Verbal so scheduled sessions appear in real time. A Chrome extension and a desktop app capture phone and video interactions without a telephony project.

Healthie is the only record system named, and the chart note audits the product sells on require reading notes from other record systems that are not named; nothing describes how notes are obtained elsewhere or which telephony and video platforms the desktop app supports.

Residency is committed; the rest of the architecture is not described. The security page says production data is always hosted in the United States and managed by resources based in the United States, and describes encryption, access controls and continuous monitoring. The product is a hosted service with a desktop capture app and a Chrome extension.

What is left open is the processing path: no hosting provider is named, nothing describes how customers are separated, and retention is unstated. Retention matters by design here, because retrospective auditing and trend analysis require keeping calls, notes and messages across an organization.

How the product is bought is clear; what it costs is not. The site offers free trials with no commitment, a demonstration route, and availability through the Healthie app marketplace; it enumerates the modules (interaction audits, chart note audits, audits of AI agents and scribes, patient insights, conversational analysis, analytics and reporting) and the buyers (digital health companies, payers and health plans, health systems and medical groups, and AI agent developers).

What no page states is the unit of charge (per user, per audited interaction, per note, per agent or a platform fee), the rate, the term, or how pricing differs for AI agent developers. The company was founded in 2024, and a compliance program becomes hard to replace once auditors rely on its history, so supplier continuity and export of audit history matter; no export terms are published.

One function applied across several buyer types, stated clearly and not yet validated by named organizations. The use case pages name digital health companies, payers and health plans, health systems and medical groups, and developers of AI agents and scribes, whose output the product audits as it audits human staff. Quotes on the home page come from a therapist, health coaches, a community health advocate, clinical operations leads, a contact center agent and a psychology group practice owner, which points to behavioral health, coaching and contact center work, but the organizations go unnamed, though a case studies page may name some.

The Healthie integration reaches virtual care providers, and the product also generates clinical documentation there. Geography is the United States.