Lyrebird Health

Ambient capture and generation is the whole product and the company was founded to build it. No non AI business underneath and no platform the model is decorating.

Conventional draft and review. Notes and documents are generated ready for transfer into the record, with the clinician editing and approving before anything is saved, and Australian professional guidance reinforces that a clinician reviews every note. Held at B because no acceptance rate, edit burden figure or confidence threshold is published.

Worth flagging for buyers rather than for the grade: the RACGP recommends obtaining patient consent before using any AI listening tool, and medical defence organisations support that position, so consent workflow is a deployment obligation here regardless of vendor design.

No accuracy figure, no model card, no named models and no evaluation methodology located. The performance claim offered is latency rather than correctness, notes generated in under 20 seconds, which tells a buyer about speed and nothing about fidelity. Multilingual consultation support is claimed without any statement of which languages or how performance varies across them.

The boundary is well drawn and no party is named, which places this in the middle band. On the boundary the disclosures are unusually concrete: all information and data processing is conducted exclusively within Australia, records are stored de identified, data is retained for 24 hours and then deleted, and the encryption standard is named rather than described as bank grade.

Exclusive in country processing bounds the chain geographically even without listing who is in it, and a 24 hour retention window limits what any party in that chain could hold. On enumeration there is nothing. No foundation model provider, model class or version is named, no hosting or cloud arrangement is published, and no sub processor list was located in two passes. One gap sits between the two halves and should be closed in writing.

The short retention window implies raw audio is not available for training, but it settles nothing about anything derived from it, and no explicit statement that customer data is never used to train models was located. Ask for the sub processor list, and ask the training question about derived text specifically rather than about recordings.

No published study, controlled evaluation or independent assessment located. Adoption is substantial and the Best Practice partnership gives it distribution across a large share of Australian general practice, but distribution is not evidence and this index does not treat installed base as a substitute for it. Worth noting the contrast within its own market: its closest local competitor has a published multi site general practice trial behind it, and Lyrebird does not.

The most specific data lifecycle disclosure located in this category. Four things are stated concretely rather than gestured at: all information and data processing is conducted exclusively within Australia, records are stored DE IDENTIFIED, data is retained for 24 HOURS and then deleted, and AES-256 encryption is named rather than described as bank level or industry standard.

A published retention window is rare anywhere in this index, and a 24 hour one materially limits the exposure from any single breach. One gap keeps this from being unimpeachable: no explicit statement was located that customer data is never used to train models, which the short retention window implies for raw audio but does not settle for anything derived from it.

The earlier assessment treated this axis as not applicable, on the basis that the product is built for the Australian market under the Australian Privacy Act. That reading is superseded. The vendor now operates a United States facing site, prices in United States dollars and pounds sterling alongside Australian pricing, and its compliance description names the United States health privacy rule alongside the European data protection regulation and Australian law. Once the claim is made the axis applies.

What a counterparty can verify is thin. No business associate agreement template, no scope statement covering which processing activities the claim reaches, and no United States customer or deployment were located.

The residency question is where a United States buyer should press hardest, because this vendor's strongest and most repeated commitment is jurisdiction specific in a way that does not automatically extend. Everything published about data location says Australian data is processed and stored on Australian servers. That is an excellent answer for an Australian practice and it says nothing at all about where a United States customer's encounters are processed. The two possibilities carry very different consequences: protected health information processed offshore is permitted, since the rule imposes no localisation requirement, but the business associate agreement must reach a foreign processor and enforcement runs across a border.

Establish which instance a United States account runs on, where its data is held, and whether the onshore commitment is Australia specific, mirrored per region, or simply Australia only with other customers served from there.

And there is no certification for the health privacy rule, so the compliance claim is a self description rather than an independent examination.

A second pass located no SOC 2 report of either type, no ISO 27001 and no trust centre. What is published is architecture rather than examination: AES 256 encryption in transit and at rest, processing on Australian servers, audio transcribed in real time and not retained afterwards, and generated documents held briefly under a short default retention window that a customer can shorten or extend.

That is a coherent data minimisation posture and it is better than most in this tier. It is still a description of design rather than an independent examination of controls.

Two things make this absence better evidenced than the usual one. The vendor is described as pursuing certifications including SOC 2 and ISO 27001 rather than holding them, and pursuing is not holding. Separately, an independent Australian market comparison of scribe vendors identifies a competitor as the only Australian scribe carrying ISO 27001 certification. That is external corroboration of the gap rather than an inference from a failed search, which is a materially stronger basis than this axis usually rests on.

The segment comparison matters here because of where this product is sold. Australian buyer guidance in this market routinely tells practices to check for ISO 27001 or SOC 2 before adopting a scribe, so the certification is not an enterprise nicety in this segment but a published purchasing criterion that buyers are being told to apply.

The grade reflects what a counterparty can verify before contracting rather than a judgement that controls are absent. Ask which certification is held today, its scope, and its date, and whether it covers the non Australian deployments as well.

No device pathway attaches and none is claimed. This axis does not read as an absence, because a governing framework exists, the vendor names it, and it has built product to serve it.

In Australia the obligations on an ambient scribe come from privacy law, from the therapeutic goods regulator's guidance on digital scribes, and from medical defence organisation position statements, rather than from a device authorisation. All three converge on documented patient consent before a scribe is used in a consultation. The vendor sets that out plainly in its own guidance for clinicians, describes the practical options as verbal consent recorded at the start of each consult or written consent captured once at enrolment, and notes that most practices end up combining the two.

What lifts this above the category baseline is that the consent obligation is met with a feature rather than an exhortation. The product timestamps verbal consent in a consent log. Peers in this category that engage with consent at all do so by supplying wording; this one produces the evidentiary record a regulator or defence organisation would actually ask to see. That is the right response to a regime where the duty sits on the clinician and the difficulty is proving compliance after the fact.

One gap is now open and should be tracked. The vendor has expanded beyond Australia, operating United States and United Kingdom facing offerings. Ambient documentation is treated as software as a medical device in the United Kingdom, where the regulator classifies these products accordingly and peers hold registrations. No United Kingdom classification or registration was located for this product, and the Australian position does not transfer.

A second search confirms the absence and sharpens it, because the vendor now makes the claim this axis exists to test.

No fairness statement, no subgroup analysis, no accuracy breakdown by accent, dialect, language or speaker population, and no evaluation methodology were located.

What was located is an explicit capability claim. The product is marketed as built to listen, handling varied accents, speech patterns and multilingual consultations. That is an assertion about precisely the property this axis asks about, published with no number, no methodology and no breakdown attached. The pattern is the familiar one across this category: the efficiency benefit is quantified and the robustness claim is asserted.

The reasoning from the earlier assessment stands and remains the most valuable thing on this record. Australia has substantial multilingual populations and Aboriginal and Torres Strait Islander populations whose speech patterns are least likely to be well represented in the data these systems are trained on. A scribe that performs less well for those speakers produces a worse record for those encounters, and nothing in the workflow surfaces it to the clinician reviewing the note. The claim to handle varied accents and speech patterns is exactly the claim that most needs evidence in that setting, and it has none.

The expansion into other markets widens the exposure rather than diluting it, since each new market adds speaker populations the training data represents differently.

This remains the single most valuable disclosure this vendor could add, and the methods exist: peers in this category have submitted note quality to published evaluation against validated documentation quality instruments.

The route into this band is jurisdictional, and this record extends a finding first recorded on the European vendors in this lane: the statutory correction right is not a European peculiarity. Australian privacy law gives an individual the right to have inaccurate personal information corrected, with a regulator behind it, which means the recorded patient has a mechanism that does not depend on the vendor granting one.

Exclusive processing within Australia keeps the product inside that regime rather than creating a transfer question that would complicate it, and that is a deliberate architectural choice rather than an accident of registration. The vendor side of the position is genuinely strong on data and silent on output.

A published retention window of 24 hours followed by deletion, de identified storage and a named encryption standard together limit what could be exposed and for how long, and a published window at all is rare across this index. None of it speaks to whether the note is right.

No accuracy or error figure was located, the only performance claim offered is generation in under 20 seconds, which describes speed rather than fidelity, and multilingual support is claimed without naming the languages or stating how performance varies across them, which is itself a liability question for any practice serving a multilingual population. No warranty, indemnity or remediation commitment was located. Ask for accuracy by language, and for what the vendor commits to when a note is wrong.

The deepest practice software integration located in this category, with the caveat that it is deep in one market rather than broad across many. Lyrebird runs inside Best Practice Premier at the click of a button, and the integration is bidirectional in the way that matters clinically: patient context including allergies, medications and history flows into the consultation rather than the scribe merely pushing text back.

It is described as the only scribe fully integrated with Best Practice, and Pro adds Genie and Gentu for specialist workflows plus Cubiko and BetterConsult. Read the grade as depth within the Australian estate; there is no evidenced integration with United States or United Kingdom record systems.

Unambiguous and rare. All data processing and storage is conducted exclusively within Australia on Australian based servers, with no audio retained after the consultation and other records deleted after 24 hours. Where Corti answers the where does the data go question by giving the buyer a choice including on premise, Lyrebird answers it by removing the choice and confining everything to one jurisdiction, which for an Australian practice under Australian privacy obligations is the simpler and more defensible answer.

Published, tiered and unusually granular. Rates are set separately for general practitioners and specialists versus allied health, and separately again for full time clinicians working 25 or more hours a week versus part time under 25 hours, quoted annually and excluding GST, with tiered discounting available for multiple clinicians.

Pricing to actual utilisation rather than a flat per seat fee is buyer friendly and almost unheard of in this category, and core scribing is now free to Best Practice Premier customers with Pro features charged separately. One caveat: published figures differ between sources and periods, with a partner listing showing 2,400 and 1,200 Australian dollars a year for full and part time clinicians and a later review citing 1,920 and 960, so the tier structure is the durable fact and the specific number should be confirmed at quote.

Narrow geographically, unusually deep in document types. Core coverage is Australian general practice with expansion into specialist and hospital settings and a separate allied health tier. What distinguishes it is that it drafts the jurisdiction specific administrative artefacts that consume clinician time rather than just the clinical note: chronic disease management plans, mental health treatment plans, Centrelink and WorkCover certificates and NDIS documentation, alongside referrals, summaries and certificates, plus a dictation mode for custom documents. No United States or United Kingdom equivalent scribe attempts local statutory paperwork of this kind. No specialty count published.