Six stages. Four human touchpoints. The rest is a machine.
Most clinics lose patients to operations, not medicine. This is what to systemise, what must stay clinical, and exactly what we build to close the gap.
If your clinic cannot name the day each patient runs out, nothing else you fix will hold
That single capability is the difference between a clinic that grows on repeat supply and one that refills a leaking bucket with paid acquisition. Everything else in this document, every stage, every automation, exists to make that one thing true and then keep it true.
If you already have it, the rest of this will sharpen a machine you have mostly built. If you do not, you are almost certainly losing compliant, happy patients every month without a single complaint to tell you it is happening.
Nothing in this is a clinical problem
A patient does everything right. Correct dose, no missed injections, good results at month three. In week nine of his second cycle his vial empties, weeks before his script expires, because nobody calculated the difference. He waits for a call that never comes, tells himself he will sort it next week, and quietly stops.
He never complains. He appears in a churn report nobody can explain, alongside a clinical standard nobody can fault.
That patient was lost by arithmetic, not medicine. So is most of the attrition in this category. The six stages below are where it happens, and each one has a specific, buildable fix.
The three reasons clinics tell us this does not apply to them
"We are too small for this."
Size is the argument for it, not against it. A clinic of two hundred patients can hold the whole picture in one person's head, right up until that person takes leave or the list doubles. The method costs least to install before you need it and most to retrofit after a year of workarounds have hardened into process.
"Our patients are happy. Retention is fine."
Almost every clinic we speak to believes this, and almost none can produce the number. The patients who leave for operational reasons do not complain and do not churn loudly. They go quiet. If your reporting cannot separate "stopped because treatment finished" from "stopped because supply ran out", retention is an assumption, not a fact.
"AI in a clinical setting is too risky."
Agreed, which is why nothing we build touches clinical judgement. No interpretation of results, no treatment recommendation, no dose adjustment, ever. The automation sits on logistics: booking, chasing, referrals, order entry, status, timing. The risk is not automating those. The risk is a clinician spending a third of every consult on retrieval while a patient somewhere runs dry.
One system, first search to fourth refill
Six stages, run end to end on one patient record. The final stage feeds back into the machine rather than terminating, which is why retention compounds instead of leaking.
Brass nodes are where a human is required. Everything between them runs without one.
Four touchpoints, and the reason for each
Everything else, and in most clinics that is roughly 27 manual actions per patient per cycle, is logistics. Capture, booking, chasing, referrals, preparation, order entry, status updates, renewal timing. None of it clinical. All of it currently done by someone who could be doing something better.
Demand
Leak · anonymous trafficMost clinic sites are built for the person ready to book, which is a small fraction of the people reading. Everyone else arrives, reads, leaves, and is never recorded. The clinic then pays to reach the same person again, because it never captured them the first time.
What we build
Included
- Patient record with source, campaign, first touch
- One form per intent, one record behind them
- Self-assessment tool, scoring you approve
- Booking surface for the ready
- UTM framework across every channel
- Attribution to booked calls, not form fills
Measure · booked calls per channel · cost per booked call
Qualify
Leak · the form-fill gapA form is an intention, not a booking, and intentions decay fast in a category built on people finally admitting something is wrong. Between the form and the call sits a task list and a working week. Most clinics lose more in that gap than anywhere upstream of it.
What we build
Included
- Scheduler on real clinician availability
- Booking writes to the record, source intact
- Confirmation and reminders
- Finite no-book path that terminates
- No-show recovery, one tap to rebook
- Suppression on every sequence
Measure · form-to-booking rate · median time to contact · no-show rate
Assess
Leak · the silent stallThe longest stage and the least visible. Bloods not booked. Results returned but unread. No named owner on the handover into clinical. Patients rarely complain here, they go quiet, and quiet is indistinguishable from patient until they are gone.
What we build
Included
- Lab result record, per-marker values and ranges
- Intake-to-cleared pipeline with owned stages
- Referral issue and chase
- Stall detection and an internal work queue
- Nurse-to-clinician handover rule
- Patient-facing progress, so silence never means fear
Measure · median days intake to cleared · stall rate by step · results unread beyond 48h
Prescribe
Leak · the cold consultClinician time is the most expensive and most limited input in the business. Spending the first third of every consult reconstructing context the clinic already holds is a data problem, not a scheduling one.
What we build
Included
- Auto-scheduling on stage change
- One-screen consult view: history, protocol, results
- Structured prescribing: drug, strength, dose, route, repeats
- Protocol write-back to the record
Measure · consults held per available hour · days from cleared to consult
Dispense
Leak · the re-keyed scriptThe clinical decision lives in one system and the order in another, joined by a person typing. Nobody designed it that way. It accumulated one workaround at a time, and every hop is a delay and an error surface.
What we build
Included
- Prescription record as structured data
- Order record with line items and state
- Prescribe-to-order flow, zero re-keying
- Pharmacy connection
- Payment at the point of order
- Patient-facing order status
Measure · hours from prescription to dispatch · manual touches per order
Retain
Leak · renewal driven off the expiry dateThis is the most valuable correction in the method, and almost every clinic has it wrong. Two dates exist and they are not the same one.
Script expiry is administrative. It is set by regulation and repeats. True run-out is derived from the dose actually prescribed and the supply actually dispensed. Work backwards from run-out through compounding, dispensing, delivery and the lead time on a review appointment, and the renewal conversation has to start weeks earlier than most systems start it.
Which means a fully compliant patient runs dry while his script is still valid and his renewal reminder is still in the future. That is not a service failure anyone can see. It is arithmetic nobody did.
What we build
Included
- Dose-derived run-out calculation
- Lead-time model: compounding, dispensing, delivery, consult
- Renewal trigger with escalation
- Self-serve reorder and repeat supply
- Lapsed-patient recovery
- Run-dry reporting, so it stops being invisible
Measure · patients who ran dry before renewal · renewal rate · reorder rate
Four records, or none of the above works
A clinic split across a spreadsheet, a dispensing system and an inbox cannot answer "when does this patient run out". Every capability in this method depends on these four existing properly, on one record, in one place.
What we will not automate
Any claim about AI in healthcare is worth exactly as much as the list of things it refuses to do. Nothing we build interprets a blood result, recommends a treatment, adjusts a dose, or tells a patient what their numbers mean.
Where an assistant speaks to patients it works from a fixed vocabulary, never offers a clinical opinion, hands to a human the moment a conversation turns clinical or distressed, and logs every word. It runs in shadow mode first, where your team reads everything it would have said before it is permitted to say anything.
Automate the logistics. Never the medicine.
The advertising gate is an asset
Prescription medicines cannot be advertised to the public in Australia. Most clinics treat that as a handicap and write vague websites that convert badly.
Invert it. If treatment detail can only appear after a medical questionnaire, the gated side of your clinic becomes the only place a patient gets a straight answer. That makes the portal the most valuable surface you own: a reason to register, a reason to return, and the one place you can be specific.
Public pages educate and qualify without naming or pricing prescription products. Treatment detail sits behind the assessment. Outcome claims stay inside what evidence and regulation permit. Every patient-facing clinical word has a named clinician against it. Build to those constraints from day one and you never unpick it later.
Three clinics, three different leaks
Anonymised, because the work is confidential. What matters is not the logo, it is which stage was leaking and what closing it changed structurally.
Could not name a single patient's run-out date
Renewal ran off script expiry. Patients on a standard supply were finishing their medication weeks before anyone contacted them, and the clinic read the result as ordinary churn.
We built dose-derived run-out across every active script, subtracted compounding, dispensing and consult lead time, and moved the renewal conversation to the point where there was still room to act.
Patients disappearing between intake and the doctor
Referrals went out by email, results came back to an inbox, and the handover into clinical lived in one person's memory. Nobody could say how many patients were stuck, or where.
We gave the stage owned steps with due-by dates, put results on the patient record, and built stall detection that surfaces anyone who stops moving.
Enquiries decaying, scripts re-typed by hand
Every enquiry became a task for a human to chase, and every prescription was manually re-entered into a separate ordering process before anything shipped.
We put live booking at the point of intent and made the prescribing output flow straight into a payable order with no re-entry.
Outcomes described structurally rather than as performance claims. Specific figures are shared under NDA on a call, against the clinic they belong to.
Where is your machine leaking?
Twelve questions, two per stage. You get a score, a band, and the one stage worth fixing first.
The build, itemised
One engagement, delivered in stages, starting wherever your scorecard says the leak is. You are not buying software. You are buying the machine assembled around your clinic.
We do not hold your clinic hostage. It is built on platforms you own, documented so your team can run it, and we would rather you did not need us afterwards.
Start with the teardown, not the contract
You do not need to decide anything today. The next step is thirty minutes where we go through your six stages and tell you where the leak is, whether or not you ever work with us.