Clinical Note Generator
Turn the consultation you just had into a structured SOAP note before the next patient walks in — drafted from the audio, reviewed and signed off by the clinician.
It listens to the consult, drafts the Subjective/Objective/Assessment/Plan note with suggested codes, and waits for the clinician to edit and approve before anything saves.
Where this works well
The slow, invisible problem this makes visible is the after-hours pile of unfinished notes — the documentation that gets done from memory at 7pm, or not at all, and quietly drifts from what was actually said. When the draft is built from the consultation audio while the conversation is still fresh, the clinician is editing a near-complete Subjective/Objective/Assessment/Plan note instead of authoring it from a blank box. The range-of-motion figure, the pain score moving from 7/10 to 4/10, the medication change from twice daily to daily — the things that are easy to lose by evening — are already captured against the right structure.
It earns its keep for high-volume, structured-consult clinicians: GPs, physiotherapists, and specialists running follow-up-heavy clinics where the note follows a predictable SOAP shape. The more repeatable the consultation pattern, the better the draft and the lighter the edit.
Where it works badly
It will be confidently wrong on coding nuance. The demo itself shows the trap: it narrates "impingement" in the assessment but emits ICD-10-AM code M75.1 (rotator cuff syndrome), when impingement maps to M75.4 — and the two are mutually exclusive under Excludes 1 rules. A model that has produced a fluent, well-formatted note can still attach a code that does not match the reasoning, and the fluency makes the error easier to wave through. A clinician who skims rather than reads is the real risk, not the model.
It also degrades on the consults that need the most help: overlapping speech, three complaints crammed into one visit, strong accents, or specialty shorthand outside its vocabulary. Here it produces a thin or scrambled draft that takes as long to fix as to write. The honest test: if your typical consult is a clear two-person exchange about one or two issues, this is for you; if it is a crowded, interrupted, multi-problem visit, expect heavy editing and weigh whether the draft is saving you anything.
What it doesn't do — and shouldn't
It surfaces a draft note, suggested clinical terms, and candidate ICD-10-AM codes with a per-section confidence indicator. It does not diagnose, it does not finalise a code, and it does not write to the patient record. The clinician edits the note, confirms the assessment and plan, accepts or changes each code, and only then approves it to the practice management system.
That boundary is deliberate and it is not negotiable here. Under AHPRA's Good Medical Practice the practitioner is responsible for accurate and adequate clinical records (section 10.5), and a clinical note is a medico-legal document. The judgement — what the diagnosis is, what the plan should be, which code is correct — stays with the person who is registered and accountable for it. The tool accelerates the writing; it does not make the clinical call.
What your data has to look like for this to work
This needs three things in good shape, and most practices have only one or two. First, clean consultation audio with documented patient consent — a single clear feed, not a phone on a desk across a noisy room. Second, an accurate, current patient record to write into: the demo leans on the previous goniometry measurement and the live referral balance (two of five physio sessions remaining), and if those fields are wrong or absent the draft compares against the wrong baseline. Third, a SOAP (or equivalent) note structure your clinicians actually follow consistently, plus the specialty vocabulary and the ICD-10-AM subset your clinic really uses, so suggestions land in your language rather than a generic one.
Getting there is usually about how information is captured at the point of care — consent prompts, mic setup, keeping referral and measurement fields live — not about buying another tool. That capture and structuring work is typically the bigger, more valuable first job, and it is the part Real Minds AI helps you get right before the AI layer is worth switching on.
Could it put the wrong diagnosis or the wrong code in the note?
Yes, and that is the failure mode to design around. In the demo it codes a shoulder presentation as M75.1 (rotator cuff syndrome) while the assessment text also names impingement, which maps to M75.4 — and under ICD-10-AM Excludes 1 rules those two are mutually exclusive. The draft is a suggestion only. The clinician reads it, corrects the code and the clinical reasoning, and approves before it saves. Nothing reaches the record on the model's say-so.
Our consults are messy — patients interrupt, two issues at once, accents, background noise. Will it cope?
Partly. Clean single-issue follow-ups transcribe and structure well. Overlapping speech, heavy accents, multiple complaints in one visit, and shorthand the model has never heard will produce a thinner, sometimes scrambled draft. It surfaces a confidence indicator per SOAP section so a weak Objective or Assessment is visible, but the honest test is whether your typical consult is a clear back-and-forth or a crowded one — the crowded ones need more clinician editing, not less.
Does this replace the clinician writing up the note?
It replaces the typing, not the clinician. Under AHPRA's Good Medical Practice (section 10.5) the practitioner remains responsible for keeping accurate, adequate records. The tool drafts; the clinician owns the assessment, the plan, the codes, and the signature. The recaptured time goes back into the consultation and into seeing the next patient, not into removing the clinician from the note.
How current does the patient context need to be?
Current enough that the note builds on the right history. The demo references the previous range-of-motion measurement and the remaining physio sessions on the existing referral — if those fields are stale or missing, the draft will compare against the wrong baseline or invent a referral status. The audio is captured live, but the record it writes into has to reflect this patient as they are today.
Where does the consultation audio and the patient data actually go?
That is the question to settle before any pilot. Health information is governed by the Privacy Act 1988 and the Australian Privacy Principles, you need explicit patient consent before recording, and the data should be processed and stored on Australian servers. Depending on configuration a digital scribe can also fall under the TGA's software-as-a-medical-device rules. We scope the data path, consent flow, and retention with you rather than assuming a vendor default.
Will it record without the patient knowing?
It shouldn't, and a compliant setup makes that impossible by design. AHPRA guidance is that practitioners inform patients about their use of AI and address any concerns raised, and consent to record is obtained at the start. The consent step is part of the workflow, not an afterthought — if a patient declines, the consult proceeds without the scribe.
Estimated build: 4–6 weeks. Most of it is template work we've already done.
Fixed scope, fixed price, fixed dates.
The cost band reflects the engagement shape, not a per-feature line item. We work on fixed scope, fixed price, fixed dates — see the services catalogue for what falls inside each band.
Considering this for your org?
The honest place to start is a bite-sized first piece — one contained change, low risk. Tell us where it hurts; we’ll play it back, scope it, and show you what’s possible.