RXNT Ambient IQ

The vendor's own pitch is the most telling evidence. RXNT markets Ambient IQ's advantage as native integration into a platform built since 1999 and as being backed by 26 years of clinical insights and millions of data points, explicitly contrasting itself with venture backed tools a year or two old. That is a moat claim about the dataset and the platform. The module is generative, and it is also sold standalone at a published price, so it does stand apart from the EHR as a product. On current positioning the moat is the EHR history, not the model.

The review step is described plainly and in the right order: the AI transcribes and produces a note, the clinician reviews and makes changes, and then signs, with RXNT customers getting a clinical encounter created automatically from the note and presented ready to be signed rather than filed. Signature as the gate rather than as a follow up is the safer pattern. No acceptance rate, edit burden figure or confidence threshold is published.

Nothing published gives an accuracy figure, a model card, named models or an evaluation methodology. The headline claim, documentation time reduced by up to 70 percent, uses the unfalsifiable up to construction. One marketing choice stands out: RXNT supports its case by citing a 2025 study reporting that physicians reached correct diagnoses in only 20 percent of complex cases while AI assisted tools were several times more accurate. That is a diagnostic accuracy finding about other systems being used to sell a documentation tool, and the two capabilities are not related. Borrowed evidence is not evidence.

The published customer agreement names a number of third parties that receive or touch data, so the architecture is described; what remains undisclosed is the model layer itself.

Named in the agreement: Kno2 LLC as the health information exchange partner for Carequality and network participation, which is expressly permitted to de-identify protected health information and use it for any lawful purpose; Intelligent Medical Objects for embedded clinical terminology; ID.me for identity proofing; Nuvei for payment processing; and a named clearinghouse for statement printing. Applications and data are stated to sit in externally certified data centers with Azure threat detection enabled.

That is a real disclosure, but it does not reach the model. None of those parties sits between a recorded conversation and a generated note. For the ambient capability specifically, no model or model family, no foundation model provider and no processing arrangement is named anywhere, and the agreement's artificial intelligence section describes what the features do without identifying what runs them.

One workflow feature makes that gap more pressing than for a capture-only product. The tool supports post-visit dictation as well as live capture, so audio exists as a file awaiting processing rather than only as a stream discarded on completion, and a file persists until something deletes it. The agreement's retention position is broad: information generated through the service is retained as necessary for financial reporting, insurance claims and other legal and business purposes, and de-identified data is claimed as the vendor's exclusive property for any lawful purpose.

Named physician testimonials and vendor claims only: documentation reduced by up to 70 percent and providers able to see up to 15 more patients a week, neither with a denominator, cohort or methodology. Inclusion in a magazine's top health technology company list is recognition rather than evaluation. There is no study, controlled comparison, accuracy benchmark or third party performance rating.

States that Ambient IQ is built to meet HIPAA standards with encrypted recordings and secure data storage, which is more specific than a generic compliance claim because it names the artifact being protected. No retention schedule for recordings or transcripts, no de identification practice and no statement on whether customer data trains models is published. Given that the product supports post visit dictation as well as live capture, audio may persist longer in the workflow than for capture only tools, so retention matters more here.

The agreement is published in full as Attachment D of the customer agreement, linked from the site footer and readable without an account, a request or a sales conversation.

Status, availability, tier and subcontractor terms are all covered. Business associate status is stated, with Networking Technology, Inc. trading as RXNT named as Business Associate. The agreement is available in full. The tier is universal: it is incorporated into the terms of service for every customer that is a covered entity, and binds on completing online enrollment, so there is no enterprise gate and nothing to negotiate. And subcontractors are bound in writing to substantially the same restrictions, which is above the regulatory floor of reasonable assurances.

The posture is strong, and three substantive terms are not generous.

De identified data is claimed outright. The vendor may de identify protected health information and use the result for any lawful purpose, and states that it is the exclusive owner of all de identified data. The customer agreement separately reserves the right to collect, use, publish, disseminate, sell and transfer aggregate data.

Retention survives the relationship by design. Information generated through the service is stated to be retained as necessary for financial reporting, insurance claims and other legal and business purposes, and the vendor may determine that return or destruction is infeasible where its own document retention policies require retention. Separately, on termination there is no obligation to return customer information at all, and it may be held for sixty days; exporting before termination is the customer's responsibility.

Breach notification runs to sixty calendar days, which is the regulatory maximum and is also the covered entity's own deadline from discovery, leaving a practice no margin to act within its own clock.

The wording does less work than it appears to. RXNT states that it hosts all applications and data in SOC 2 Type II certified data centers. That certifies the facility and its operator, not RXNT's own controls, processes or software. It is a meaningfully weaker claim than holding a SOC 2 Type II report covering the company itself, which is what a health system security review will ask for. A real certification is present in the stack, at the facility level.

No clearance claimed and none required for ambient documentation. No United States device pathway attaches to a note the clinician reviews and signs, and the vendor describes exactly that workflow: start the session, record the discussion, then review and sign the generated note.

The company makes an express scoping statement: the product is not a diagnostic tool, and diagnosis remains in the purview of the provider. Saying so unprompted is the right posture.

One thing sits awkwardly beside it, and claims of intended use are part of how device status gets assessed. The same launch material introduces that disclaimer by citing a 2025 study on diagnostic accuracy, reporting that physicians reached correct diagnoses in only a minority of complex cases while AI assisted tools performed several times better, then states that although this product is not diagnostic it helps bridge the gap. That borrows the credibility of diagnostic AI research to market a documentation product. The disclaimer which follows is accurate, but the surrounding framing gestures at a benefit the product does not claim to deliver, and a regulator reading marketing for evidence of intended use could read it the same way.

The product itself is narrow. Transcription and note generation, an encounter summary feature for spotting charting discrepancies, and specialty templates. No coding engine, decision support module or patient facing product is described, which keeps the regulatory position simple.

United States only, sold both natively inside the vendor's own record system and standalone.

No fairness statement, subgroup analysis, accuracy breakdown by accent, dialect, age or speech characteristic, or evaluation methodology is published, and no language coverage beyond English is claimed.

The familiar imbalance is present. The time saving is quantified precisely and repeatedly at up to seventy percent, while accuracy is asserted throughout, with claims that the system documents critical details which might otherwise be overlooked and reduces errors and omissions. The efficiency figure has a number attached. The accuracy claim does not.

This product faces a sharper and more specific version of the domain question than a general purpose scribe does, because of what it markets itself for. The vendor names psychiatry and behavioral health among its specialty templates. Those are precisely the encounters where speech recognition meets its hardest conditions. Pressured or disorganized speech, flattened prosody, medication induced dysarthria, long pauses, distress and thought disorder are clinical features of the population being documented rather than noise around it. A scribe that degrades under those conditions degrades specifically for psychiatric patients, and the record it produces is the one that follows them.

English only coverage compounds it. A behavioral health encounter conducted in a patient's second language is already harder to document accurately, and nothing in the workflow reveals to the reviewing clinician where the system struggled.

No evaluation of note quality against a validated documentation quality instrument is published, though such instruments exist and are in use. The gap is a choice rather than a technical impossibility.

The published agreement addresses artificial intelligence output directly.

Absent are any accuracy or error figure and any warranty, indemnity or remediation commitment attached to the output. The headline claim, documentation time reduced by up to 70 percent, uses an up to construction that any result at all satisfies.

What is published is a dedicated artificial intelligence section, and it is candid even though every clause runs away from the vendor. It states that generated content cannot guarantee specific outcomes, that results may not be correct even on accurate and complete data because of inherent limitations and the potential for hallucinations, that the customer is solely responsible for evaluating output and making final care decisions, and that the vendor is not liable for damages arising from use or misuse of it. Elsewhere accuracy, reliability, completeness and currency are expressly not warranted, and total liability is capped at the fees paid in the preceding twelve months.

The correction mechanism exists and reaches one party. The business associate agreement obliges the vendor to make records available for amendment under 45 CFR 164.526 and to notify the practice promptly if an individual requests an amendment directly, with any denial remaining the practice's responsibility. The same agreement states that it creates no rights in third parties. So a mechanism exists, nothing stands behind the output, and the patient has no route.

One marketing choice is more misleading than an absent number. The vendor supports its case by citing a study reporting that physicians reached correct diagnoses in only a small minority of complex cases while artificial intelligence assisted tools were several times more accurate. That is a finding about diagnostic systems, produced by other parties, deployed to sell a documentation tool. Nothing about a scribe's ability to transcribe and structure an encounter follows from a diagnostic model's performance. Borrowed evidence is not evidence, and the effect is to import credibility the vendor has not earned rather than merely to omit a measurement.

Two tiers, described candidly, which is the notable part. Inside RXNT the integration is native and produces a signed ready clinical encounter rather than text to move. Outside it, Ambient IQ works standalone and system agnostically by copying the note into any record system, and RXNT says so directly rather than implying parity, even writing that third party integrations elsewhere mean results may vary. The deep path exists only within its own platform, and the standalone path is transfer based.

Cloud delivered with 24 hour access and mobile applications for iOS and Android, and hosting is described as being in certified data centers. No region, residency option or subprocessor detail is published: the facilities are described as audited, and their location is not given.

Pricing is published in full, and the contract goes further than the price page.

RXNT publishes its full price list on its own website, flat and per provider per month, so the cost can be worked out without a sales call. Ambient IQ is included at no additional cost for EHR and Full Suite customers, with unlimited sessions and no cap on transcription or note creation, and is sold standalone at $75 a month after a free three month trial. Setup, onboarding, training, support and upgrades are stated as included rather than quoted separately. It also discloses which features standalone buyers do not receive.

The published pricing policy goes further than the price page. Ancillary services are individually priced in public: controlled substance prescribing per provider per year, secure direct email per practice and per address, an immunization interface with a stated setup fee and monthly rate, payment processing per provider per month, patient statements and paper claims per item, and custom form work per form with revision rates. Data storage, data import, data export within the service, and support are stated at no charge. Very little is left to a quote.

Three commitments sit alongside the rate, none of which appear on the price page. The subscription runs a minimum one year term and renews automatically unless canceled at least sixty days before renewal. Monthly billing is described as a financing convenience that does not shorten that term. And all fees are nonrefundable and noncancelable, including on deactivating a user partway through the term.

Ambulatory across practice sizes from solo providers to multi specialty multi location groups, described as specialty and system agnostic, with typical users including family medicine, pediatrics and urology. Two capabilities widen the fit: custom encounter templates by specialty, and support for both live capture during the visit and post visit dictation, which suits clinicians who cannot record in front of patients. One honest limitation the vendor states itself: the specialty template customization is available only to EHR and Full Suite customers, not to standalone users. English only.