Tripletime

Every input path runs through models: snipped text and uploaded referral packages are parsed and placed into the correct note sections, ambient speech fills the history of present illness and plan, dictation is sorted rather than transcribed verbatim, phone photographs are read by OCR, and the assistant layer generates referral letters, after visit summaries and billing suggestions. No human scribe tier and no services layer underneath.

Graded B on a design behaviour that is demonstrated rather than claimed, and it is the most encouraging thing on this record. In the published pre chart example the system populates what it can source from prior documents and leaves what it cannot know as explicit placeholders, reading history of presenting illness to be documented during visit, examination pending, results pending, social history pending.

Marking absent information as absent instead of producing plausible filler is precisely the failure mode that error research in this category, most rigorously by Tortus, identifies as the dangerous one, and a product that visibly declines to invent is easier to review safely. The note is prepared for the clinician to complete and sign.

Held at B rather than A because no confidence threshold, error rate, hallucination measurement or abstention policy is published, and because content pulled from snipped prior reports enters the note without any described verification that the snip was read correctly.

No model card, named model, accuracy figure, error rate or evaluation methodology located for any of the four input paths, which is a notable gap given that OCR of a photographed blister pack and mid consultation language switching are both harder problems than ordinary transcription and both are sold as capabilities. Accuracy is asserted through the phrase medical grade, applied to both the OCR and the bilingual handling, which describes an intention rather than a measurement. The single quantified claim, up to three hours saved per clinician per day, carries no denominator or method.

The boundary is stated as cleanly as anywhere in this lane and no party is named. Two unqualified commitments cover what this axis asks about content: the vendor states it does not train its models on customer data, with none of the third party or shared model carve outs that weaken the same claim elsewhere in this index, and that it does not store any audio.

Together those foreclose accumulation of the most sensitive artefact and foreclose the training question outright, which is a better position than most of this category reaches. On enumeration there is nothing: no model or model family, no foundation model provider, no hosting arrangement and no sub processor list was located. One gap sits between the two halves and matters because the product does retain content even though it does not retain audio.

Photographed documents and synced material persist across devices, and no hosting region or storage location is stated for any of it, so a buyer knows what is not kept without knowing where what is kept actually sits. For a vendor claiming compliance across three regimes at once that is the question, since the answer determines which of the three governs. Ask for the storage region for notes and images, and for a sub processor list.

No study, controlled evaluation, third party rating, institutional customer or deployment count located. Testimonials are a step better than most of this tail because individual physicians are named with their cities, in Toronto, Montreal and Calgary, rather than reduced to initials, but they remain vendor selected quotes from individual clinicians rather than an organisation attesting to a deployment. The claim that clinicians save up to three hours a day is unsupported by any published measurement.

Two unqualified commitments that together cover the questions this axis exists to ask. Tripletime states plainly that it does not train its models on customer data, with no third party carve out of the kind that held Tali to a B, and that it does not store any audio. Compliance is claimed across three regimes rather than one, HIPAA, PIPEDA and GDPR, which is consistent with a Canadian vendor serving cross border users. Doximity holds an A on this axis for an audio commitment alone with the training question left open, while Tripletime closes both.

The remaining gap is recorded rather than penalised here, because it belongs on the residency axis: no hosting region or storage location is stated for the notes and images that are retained, and the product does retain photographed documents and synced content across devices.

Compliance is claimed across HIPAA, PIPEDA and GDPR, displayed prominently and addressed directly in the vendor's own FAQ, which also references NHS requirements. Naming the Canadian and European regimes alongside HIPAA is more precise than the single regime claim typical in this lane and matches where the company appears to sell.

Held at B because no business associate agreement terms, availability or tier gating are published, and because a compliance claim across three jurisdictions is a broad assertion that a buyer should ask the vendor to evidence for their own regime specifically.

A watchlist phrasing this index has caught before, and it should be read carefully. Tripletime states it uses SOC 2 and ISO 27001 certified providers at all stages. That certifies its suppliers and infrastructure vendors, not Tripletime's own controls, processes or software, which is what a security review actually asks for.

RXNT was graded C on the same distinction for hosting in SOC 2 Type II certified data centres, and the full watchlist in this lane now also includes SOC 2 aligned, verified by a third party auditor unnamed, military grade security and bank level encryption.

Graded C rather than Not Rated because real certifications exist somewhere in the stack and the vendor is being reasonably precise about whose they are. Ask whether Tripletime itself holds a SOC 2 Type II report, and if not, when.

No clearance claimed and none required for documentation itself. No United States device pathway attaches to drafting a note a licensed clinician reviews and signs, and the signature is the control point. Three features sitting around that documentation core each raise a different question, and the earlier assessment of this record saw only one of them.

The assistant suggests billing codes from documented procedures and diagnoses. That is the coding boundary: a system proposing the codes a note will be billed under shapes reimbursement in a way note drafting does not, and it creates an incentive gradient the documentation product alone does not have.

The assistant also generates after visit summaries in plain language for patients. Patient facing output carries a different safety posture from clinician facing output, because the reasoning that keeps clinical decision support outside device regulation rests on a professional being able to review the basis of what is produced, and a patient is not a professional. That the summary derives from a note the clinician has already signed mitigates it, but the artefact itself reaches the patient.

The question answering feature is described as answering any clinical question using the full context of the patient's notes. Grounding answers in the clinician's own record is more traceable than open ended inference and that is to the product's credit, but any clinical question is a wider scope than the earlier read of this record allowed.

Markets matter here. The published customer evidence is Canadian, the interface is bilingual English and French, the integrations span Canadian and United States record systems, and the vendor's own frequently asked questions raise NHS requirements alongside HIPAA and PIPEDA. Ambient documentation is treated as software as a medical device in the United Kingdom and the European Union. No classification or registration under that regime was located, so a buyer in those markets should ask for it directly.

Graded B for treating language variation as an engineering requirement rather than a caveat, which is the same basis on which Corti and voize earned this grade for accent handling.

Tripletime is the first vendor in this index to claim mid consultation code switching as a designed property. It states that a clinician can move between English and French part way through an encounter and the system auto syncs and still produces the note in the target language regardless of input language, with medical terminology held accurate in both, citing CABG and pontage coronarien. Code switching mid conversation is materially harder than supporting two languages, it is the daily reality of bilingual practice in Canada, and until now this index had recorded it only once, in an app store review written by a clinician rather than as a vendor claim.

The billing code feature also stays off the coding gradient. Codes are suggested from documented procedures and diagnoses with no revenue lift, capture or optimisation framing anywhere, which is the Wavo pattern.

Held at B because no performance breakdown by language, accent or speaker is published, and the whole bilingual claim rests on assertion.

The band is reached through the regime rather than through the vendor. Operating under Canadian federal privacy law with cross border users, the individual holds an enforceable right to challenge the accuracy of information held about them and have it corrected, with regulators behind it, so the recorded patient has a mechanism that does not depend on this company offering one.

The vendor's own contribution on this axis is thin and one gap is more consequential than a general absence of accuracy figures. The product takes four distinct input paths, and two of them are harder problems than ordinary transcription and are sold as capabilities: optical character recognition of a photographed medication package, and language switching mid consultation. Neither has a published accuracy figure, error rate or evaluation methodology.

The first of those is where an error becomes a medication error rather than a documentation one, since a misread strength or a misread drug name transfers directly into a record a prescriber will rely on, and the failure is silent because the photograph is not re read. Accuracy is asserted through the phrase medical grade, which describes an intention rather than a measurement and is not a standard anything is assessed against.

The single quantified claim uses the unfalsifiable up to construction. Ask for a character and field level accuracy figure on the medication capture path specifically, and for what the vendor commits to when a drug or strength is read wrongly.

The weakest part of the record, and the claim shape this index treats as a warning. The site says the product works across any EHR and displays a logo wall covering Accuro, Arya, athenahealth, Epic, Greenway, NextGen, Oracle, Telus and Meditech, a spread that usefully signals the Canadian market through Accuro and Telus Health.

But no integration mechanism is described anywhere: no API, no standard, no write back, no named connector, and no supported systems page. In this lane universal compatibility is normally a consequence of not integrating rather than of integrating well, which is the finding recorded against LucasAI, and a desktop application that produces a note for the clinician to move into the chart fits that pattern. Templates are imported from the EMR, which is a read rather than a write.

Establish what actually happens to the finished note before assuming these logos represent integrations.

The architecture is described clearly even though the residency is not. Tripletime ships a desktop application with a companion mobile app on iOS 15 and Android 5 or later, and recordings and photographs sync automatically from phone to desktop, which is a coherent design for a clinician moving between clinic room and workstation. What is absent is any statement of hosting region, storage location, residency option or sub processor list.

That gap is concrete rather than theoretical here, because the vendor claims compliance with PIPEDA and GDPR alongside HIPAA, and each of those points at a different answer about where data may sit. For a Canadian clinician, whether patient content crosses the border is a question the product does not answer.

A published pricing page exists and the trial terms are stated plainly on the landing page rather than buried: free for 14 days with no credit card required, plus a free mobile app download on both platforms. Removing the card capture step lets a clinician evaluate output on their own consultations before any commitment, which this index credits.

Held at B rather than A because the tier structure and what each tier gates were not confirmed, and because no enterprise or group terms are described for a product sold to specialist practices where multi clinician licensing is the likely purchase.

Deliberately narrow and genuinely deep, which this index prefers to a large unsupported specialty count. Coverage is cardiology, internal medicine and oncology, aimed at specialist consultation work rather than at primary care volume.

The cardiology depth passes the instrument test: templates are described as built by cardiologists working with guideline directed medical therapy, coronary anatomy and high sensitivity troponin, and the published sample note handles catheterisation findings, ejection fraction across serial studies, stress imaging and perfusion defects as structured longitudinal history rather than as narrative. Telephone and video consultations are supported alongside in person. Held at B because three specialties is a narrow footprint and no inpatient, procedural or emergency setting is addressed.