ZenithMed AI
An ambient clinical AI scribe, built by AUXA.
ZenithMed AI is an ambient clinical AI scribe built by AUXA. This page describes what it does and the constraints that shaped it. It reports no clinical outcomes, no accuracy figures and no deployments, because none are published.
The problem it addresses
A clinician spends a substantial part of a consultation typing rather than listening, and the note written afterwards is reconstructed from memory. An ambient scribe listens to the consultation and produces a structured draft note, leaving the clinician to review and sign rather than to transcribe.
The hard part is not transcription. It is that the output is a clinical record, and a clinical record that is subtly wrong is more dangerous than one that is missing.
What the platform does
Ambient capture. Audio from the consultation, with consent handled before capture begins rather than assumed.
Structured draft note. Output shaped to the sections a clinical note actually has, not a wall of transcript.
Clinician review before anything is recorded. The draft is a draft. Nothing enters the record without a person accepting it.
Architecture and the decisions behind it
Review is a required step, not a setting. There is no configuration that lets a generated note enter the record unreviewed. Anything that can be switched off eventually is.
The transcript and the note are kept distinct. A note is derived from a transcript and is not the same artefact. Keeping both means a reviewer can see what was actually said rather than only what was summarised.
Data handling designed in from the first commit. Where audio is processed, how long it is kept and who can reach it are architectural decisions in health-adjacent software, not a review at the end. The Privacy Act 2020 governs handling of personal information in New Zealand.
Failure is visible rather than silent. If audio is poor or a section cannot be produced, the draft says so. A scribe that quietly omits what it could not hear is worse than one that stops.
Why this page is last
Clinical software is the most crowded category we work in and the one where evidence matters most. ZenithMed AI is presented here as a description of capability and nothing more. We have no published accuracy figures, no deployment numbers and no clinical results, and a page implying otherwise would be the exact failure the software itself is designed to avoid.
Related
The other platforms are AskHR360 and VisaFlow. The service line is described on our software development page.
Frequently asked questions
Is ZenithMed AI in clinical use?
We are not publishing deployment information. This page describes what the software does and how it is built.
How accurate is it?
We do not publish accuracy figures. Any number we gave would need evidence behind it that we are not in a position to share.
Does a clinician have to review the note?
Yes. Review is a required step in the software rather than a configurable one, because a setting that can be switched off eventually is.
Who handles consent?
Consent is obtained before capture begins. The software does not treat consent as an assumption.