Australian GPs can choose from a dozen AI scribes. Community pharmacy got none of them.
Scribe is a consult-note engine built for the pharmacy counter. Record the consult; it drafts the note, the claim document and the GP letter, for you to check, correct and sign.
Consult flow
- 1Record
- 2Transcribe on this machine
- 3Draft
- 4You review and export
Ten minutes of talking. One reviewed record.
-
1
Pick the consult type.
MedsCheck and Diabetes MedsCheck, immunisation, leave certificate, UTI, oral contraception, state pharmacist prescribing, or a standard SOAP note. Seven built in, plus any template you write yourself.
-
2
Pull up what they are actually taking.
Scribe reads the pharmacy's own dispensing system and puts the patient's history on screen before you start. It is labelled for what it is: dispensing history for the last 90 days, not a current-medications list. The prescriber line reads "most frequent prescriber, 12 months", because that is what the data supports.
-
3
Hit record and talk normally.
Speech is transcribed on this machine. Drug names are corrected against 4,843 PBS brand and generic names and 11,489 over-the-counter terms, so "glick-something" comes back as gliclazide.
-
4
Read the draft. Fix what is wrong.
Nothing leaves the app until you have read it. The review step is not a formality, it is the design.
The note is not the deliverable. The document is.
A note you retype into a claim portal has saved you nothing. Scribe finishes the job: the thirteen-section MedsCheck report, the immunisation record, the leave certificate, the UTI and oral contraception records, the state prescribing record, the GP or referral letter, and the patient handout. Nine document generators, exporting to DOCX, or PDF where LibreOffice is installed.
And each one names its own claim pathway. MedsCheck goes to the PPA portal. Immunisation reports to AIR, with a PPA NIPVIP claim only where the vaccine is NIP-funded. Leave certificates are a private service and no PPA claim exists. Scribe puts that on the document, so nobody bills for something that was never claimable.
What it will not do.
Most of the engineering in Scribe went into the things it refuses to guess at.
It will not check interactions.
No trustworthy source is wired in, and a half-built interaction checker in a clinical UI is worse than none.
It will not diagnose.
It does not grade severity, infer a condition from symptoms, or decide that something was a red flag. It records what was said.
It will not pick the patient's handout.
You choose the condition. Scribe fetches published resources from Australian health organisations and prints them with attribution. It does not choose these for you.
It will not fill a gap with a plausible guess.
The instruction it runs under is blunt about it: filling gaps with plausible clinical reasoning is the single worst thing it can do.
You carry the professional liability. Scribe is built so that it never quietly makes a clinical decision on your behalf.
Everything it does.
Works out of the box
| Capability | Detail |
|---|---|
| Consult recording and transcription | On this machine, faster-whisper |
| Structured note drafting | Seven consult types plus custom templates |
| Drug-name correction | 4,843 PBS names, 11,489 OTC terms |
| GP and referral letters | Notification or referral, drafted from the note |
| Dictate mode | Speak straight into any field, no consult required |
| Medication profile import | Upload a PDF or screenshot, per-row pharmacist sign-off |
| Patient resource library | Published AU peak-body resources, printed with attribution |
| Custom template editor | Define your own consult types in the app |
| Consult history and autosave | Every draft saved as you type |
| Model picker | Local model, or a cloud provider if you choose one |
| DOCX and PDF export | Nine document generators |
| Live transcript preview | Requires a GPU; disables itself on CPU |
Switch on when you are ready
| Capability | What it needs |
|---|---|
| Dispensing history from FRED or APSS | A read-only login on the pharmacy's own server |
| Dispensing history via the Podium warehouse | Multi-store, scoped per store |
| Hosted mode for multiple pharmacies | Logins, per-user storage, admin panel |
Where the data goes.
- Audio is transcribed on this machine and is never sent to a third-party AI service. That is true in every mode, and it is not retained past transcription.
- Note drafting runs on a local model by default. If you switch to a cloud provider in the model picker, the transcript text is sent to that provider, and the app shows a banner saying so for as long as it is on. No mode sends audio.
- Consult records are stored in the pharmacy's own database.
- Dispensing data is read-only by construction. Three independent layers prevent writes, and that was verified against the live warehouse rather than asserted.
Built by a pharmacist, for the pharmacy floor.
Sign in