Patient-Message Triage Copilot | Real Minds AI
Healthcare & Disability /Customer-service agent live field guide · 9 min

Patient-Message Triage Copilot

Sorts the patient inbox before it reaches a clinician — classifies each portal or email message, scores urgency, routes it to the right queue, and drafts a holding reply for staff to approve, while escalating anything that reads clinically urgent instead of answering it.

theater/demos/healthcare_patient-message-triage-copilot.html · sandbox · read-only
Open
FIG. 1

The live demo, running on fabricated data. Open it to step through the full flow — every output is shown for a person to approve before anything happens.

How it would work

Reads each inbound patient message against your triage policy, proposes an intent, urgency and route, drafts a grounded holding reply, and surfaces all of it — with any clinical concern flagged — for reception or a nurse to approve, edit, reassign or escalate before anything is sent.

Input 01
The patient inbox

Inbound patient-portal and email messages, each matched to the patient's record in your practice management system — sender, channel, timestamp and body text.

Agent 02
Reads, classifies, routes, drafts

Tags clinical and intent signals, classifies the request against your message taxonomy, scores urgency on your triage rules, picks a queue from your routing map, and drafts a holding reply grounded in your policy — no medical advice.

Output 03
A sorted message, working shown

A proposed classification, route, confidence score and drafted holding reply that reception or a triage nurse approves, edits, reassigns or escalates — a clinically urgent message is held with no auto-reply for same-day GP or nurse review.

Where it works well

It does the first sort on every message, every time, and shows why it routed each one.

  • Best where sorting is itself a job — a multi-GP practice, or an aged-care service fielding family enquiries, where someone triages the queue by opening each item.
  • It surfaces intent, urgency, a route and a grounded holding reply, so the act of sorting stops being the bottleneck.
  • At high message volume the recaptured time goes back into patient contact and the messages that need judgement, not into a smaller roster.

The slow, invisible cost of a patient inbox is not the urgent message — it is the forty routine ones a clinician opens to find the two that matter, because messages arrive in the order they were sent, not the order they count.

Where it works badly

It is confidently wrong when the record it matches against is stale — and the clean classification hides the bad fact underneath.

  • Useless where the inbox isn't really an inbox — patients who phone, or "messages" that are scanned PDFs and faxes with no machine-readable text.
  • Weak on the message that needs a clarifying question, not a route — "I'm not right since the new tablets" needs a person to ask what "not right" means; it will flag and escalate, not resolve it.
The honest test

Count last week's inbox and ask how many messages a competent receptionist could route in under ten seconds; if most need a back-and-forth with the patient, the sort isn't your bottleneck and this won't pay back yet.

If it reports "active patient, matched — last script 6 weeks ago" and that record is out of date, you get a confident sort built on a wrong fact, which is more dangerous than no sort at all.

What it doesn't do — and shouldn't

It drafts and sorts. A person triages and decides. That line is deliberate.

WHAT IT DOES
Surfaces a proposed intent, urgency, route and confidence score
Flags clinical-concern signals against a named triage-policy clause
Drafts a neutral holding reply with no medical advice
WHAT IT WON’T
Decide clinical urgency or make a triage call that binds care
Send a clinical answer or approve a repeat script
Auto-reply to any message carrying a clinical signal

Under the RACGP Standards for general practices a responsive triage system with a person assessing urgency is a mandatory requirement, and under AHPRA's standards the treating practitioner stays clinically responsible for the patient. A tool that confidently reassured someone their dizziness was nothing while sending their repeat script would be worse than no tool — the consequence lands on the clinician, so the clinician stays on the call.

What your data has to look like

Messages as readable text, a current de-duplicated patient record, and a written triage and routing policy.

44%
Typical readiness
across orgs we see, before the first job
Messages as machine-readable text
Usual weak point
A current, de-duplicated patient record
Needs shaping
A written triage and routing policy
Needs shaping
A message taxonomy to classify against
Usual weak point
Mailbox / portal integration
Usually ready
The real first job

The triage and routing policy is usually the real first job — writing down the rules your senior reception staff already carry in their heads, and getting patient and medication records clean enough to trust. That work is rarely about buying software, and on most engagements it is larger and more valuable than the AI layer that sits on top of it.

Right fit if…
Message volume is high enough that sorting the inbox is itself a job
You have a written triage policy and a routing map you can point to
Patient and medication records are current and de-duplicated
Messages arrive as text — portal and email, not phone or faxed images
Walk away if…
Patients mostly phone, or messages arrive as scanned PDFs and faxes
Your practice records are stale or full of duplicate patient files
Most messages need a clarifying question back to the patient, not a route
You want a tool that makes the clinical triage call for you
Open questions

The worried-buyer questions, answered straight

It never gives medical advice and never auto-answers a message with a clinical signal in it. When it detects symptoms reported alongside a request — like dizziness and headaches next to a repeat-script request — it holds the message, suppresses any reply, and raises an escalation banner for same-day GP or nurse review. The only thing it drafts in those cases is a neutral holding reply (“a nurse will call you today”) that a person approves before it goes out. A clinical concern routes to a human, not to an answer.
Short, mixed or ambiguous messages are normal, and it is built to flag rather than guess. A message it can’t match to an active patient record, or can’t classify above your confidence threshold, is routed to a human queue rather than auto-handled, and it shows a confidence score so reception can see when it is sure and when it isn’t. What it can’t fix is a message that genuinely needs a clarifying question back to the patient — that is a human reply, and it routes it as one.
No. It does the sorting — reading, classifying, routing, drafting — so reception and triage staff spend their time on the messages that need judgement and on patient contact, not on opening every email to work out where it goes. Under the RACGP Standards a person still assesses urgency and a clinician stays responsible for clinical decisions. The recaptured time goes back into care and the queue that matters, not into a smaller roster.
Current enough that the record it matches against is right today. If it tells a clinician “last script 6 weeks ago” or “active patient, matched”, that is only safe if your practice records and medication history are up to date in the system it reads. Stale data is the main failure mode: a patient who changed medication last week, or merged duplicate records, produces a confident-but-wrong classification. It reads your live practice system, so the freshness is whatever your front desk maintains.
It runs against your own systems — your patient records and Microsoft 365 mailbox — and message content is sent to the language model only to classify and draft, not to train it. Health information is sensitive information under the Privacy Act 1988 and the Australian Privacy Principles, so before any deployment we map exactly what data is sent where, what is logged and what is retained — set with your practice, not assumed. If your records are in My Health Record, the My Health Records Act 2012 applies to that source as well.
Every classification cites its source — the message itself, your practice records, your triage rules, or a named clause of your triage policy. The route, the urgency, the detected signals and the holding reply are all visible and attributed before a person approves them. That audit trail is what lets a nurse or GP trust the sort without re-reading every raw message from scratch.
What it takes to build
3–4 weeks · 4 phases
Reused from template~65%
Bespoke to this skin~35%
stack · Claude · Microsoft 365 · helpdesk queue
What it would cost

Fixed scope, fixed price, fixed dates.

01
Bite-sized first piece
One contained change, low risk
02
Pilot build
Most builds land here
03
Embedded support
Scale on proof

Considering this for your practice?

The honest place to start is a bite-sized first piece — one contained change, low risk. Tell us where the inbox hurts; we'll play it back, scope it, and show you what's possible.

More in Healthcare & Disability
No-Show Predictor
View →
Care-Plan First-Drafter
View →
Clinical Note Generator
View →
Complaint Response Agent
View →
How We Work Proof Talk to us
How We Work Proof Talk to us
Ask us anything