For discussion — not a sales pitch

Every referral reaches a decision.
None sit in a pile.

A proposal for a referral triage assistant: software that reads every incoming referral, checks it against your intake rules, and hands your staff a ready-to-approve decision — so the backlog stops growing and nothing falls through the cracks.

Prepared by Alic Edgar Date July 2026 Status Concept — seeking clinic input

The problem we keep hearing

Triage is eating staff time, and some referrals never get looked at

Referrals arrive from systems that were never designed to work together — fax, eReferral portals, email, EMR inboxes. Someone on your team has to open each one, figure out what it is, chase missing information, and decide whether it even belongs at your clinic. That effort scales with volume, and when it can't keep up, a backlog forms.

The worst outcome isn't slow triage — it's the referrals that will realistically never be triaged. That hurts the patient waiting on the other end, hurts the referring provider's trust in your clinic, and quietly becomes a liability sitting in a drawer.

From what we've seen, a meaningful share of any backlog is not real triage work at all. It's referrals that are incomplete (missing information, unreadable, unsigned) or that don't fit the clinic in the first place — for example, motor-vehicle-accident cases at a clinic that doesn't take MVA work. Today a person has to read each of those just to discover it was never bookable.

What we're proposing

One queue, every referral pre-read, your staff always decides

The assistant sits in front of your existing systems — it doesn't replace any of them. Every incoming referral, from every source, lands in a single queue where the software has already done the reading:

CaptureYour referral fax line forwards to the service. Portal and email referrals are added with one click by staff as they view them — no new logins for referrers, nothing changes for the people sending you referrals.

Read & checkThe software reads the document — including scanned faxes and handwriting — and pulls out the key details: patient, referrer, reason, funding type, what's attached and what's missing.

ProposeIt checks the referral against rules you wrote in plain language ("we don't take MVA cases", "adults 18+ only", "a referral without a reason is incomplete") and proposes a decision: decline, request missing info, or accept — with the reason cited and a response letter already drafted.

You approveStaff reviews the proposal next to the original document and clicks approve — typically a few seconds per routine item. The reply goes back to the referrer over the same channel it arrived on, and every decision is logged.

Your triager's job changes from "read 60 faxes and work out what each one is" to "review 60 prepared decisions and click through them."

A concrete example: the MVA rule

Suppose your clinic doesn't accept motor-vehicle-accident cases. You write that rule once. Then:

Incoming fax · 3 pages · received 8:42 a.m.

Patient

J. D—, DOB 1981 · contact info present

Referrer

Dr. A—, family practice · fax-back number present

Reason

Neck and shoulder pain following rear-end collision, six weeks ago

Funding

Motor vehicle accident — insurer claim referenced

Proposed: Decline — clinic rule "No MVA cases". A return letter to Dr. A— is drafted, citing the rule and (optionally) suggesting clinics that do take MVA work. Your staff glances at the highlighted evidence, clicks approve, and the fax goes back the same morning — instead of that referral sitting unread for weeks.

MVA is just the proof case. The real value compounds as you add your rules — the age ranges, funding types, catchment, required documents, and specialty boundaries specific to your clinic. Each rule you add removes a category of reading from your staff forever.

And then: prioritization, not just filtering (phase two)

Once rejection filtering has earned trust, the same rules engine orders what remains: "post-surgical within two weeks", "falls over 65 jump the queue", "worsening neurological symptoms flag for same-day clinician review." Accepted referrals arrive pre-sorted by your policy with deadlines attached — so prioritization reflects your standards, not the order of the pile.

What it does

Reads every referral, from every source, into one queue
Flags incomplete and out-of-profile referrals with the rule cited
Drafts every reply — staff approves before anything is sent
Tracks follow-ups so "waiting on info" never becomes "forgotten"
Keeps a complete log: what arrived, when, who decided, why

What it never does

Never makes a clinical judgment — rules are administrative and yours
Never sends anything without staff approval (automation is opt-in, per rule, and earned)
Never contacts patients
Never replaces your EMR, booking, or eReferral systems
Never uses your data to train AI models or serve other clinics

Privacy & custody

Your clinic remains the custodian. Full stop.

This is health information, so the privacy design comes first, not last:

You stay the health information custodian. We act only as your service provider under a written agreement (PHIPA-style agent agreement), touching only referrals your clinic already lawfully receives.
No new access pathways. The system captures referrals at your fax line and at your staff's own screens — it never holds portal or EMR credentials, and nothing runs behind your staff's back.
Canadian data residency for stored referral data, encrypted, isolated per clinic.
AI processing under zero-retention terms — documents are read, not kept by the AI provider, and never used for model training.
Retention is your policy. Original documents are kept only as long as you require; the long-lived record is the decision log.
Every access is logged. We arrive with the paperwork drafted — agreement and privacy impact assessment — for your privacy officer to review, and a plain-language data-flow diagram for every source.

Alternatives & cost

What this has to beat

Any proposal like this should be judged against the realistic alternatives — including doing nothing. Here's the honest comparison as we see it. The figures are rough Ontario estimates for discussion; part of why we want your input is to replace them with your real numbers.

Option	Rough annual cost	What it solves	What it doesn't
Do nothing	"Free"	No change, no effort	Backlog compounds; untriaged referrals are patient harm and medico-legal exposure that grows quietly
Add admin staffing part-time hire or overtime	$20–35k+ (15–20 hrs/wk at $22–30/hr, plus onboarding & turnover)	More reading capacity; human judgment throughout	Scales linearly with volume; doesn't fix missed items, inconsistency between triagers, or the audit trail; hardest role to keep staffed
Clinician does triage status quo at many clinics	Highest real cost (clinician hours at $60–100+/hr spent on paperwork)	Best judgment on genuinely clinical calls	Most expensive person in the building reading incomplete faxes; time comes out of patient care
Outsourced intake service virtual admin / intake BPO	$15–35/hr as used	Flexible capacity without hiring	Third people handling PHI with less oversight; no clinic-specific rules memory; quality varies; you still manage them
eReferral / EMR built-in features	Often already paid for	Structured forms help within that one network	Only sees its own channel — fax, email, and other portals remain a pile; little clinic-rule filtering; you have several of these systems, which is the problem
This proposal	Pilot: free · then roughly $3–6k (a few hundred per month — final pricing is one thing we're here to test)	Every channel in one queue; your rules applied consistently; drafted replies; nothing unaccounted for; complete audit trail	Genuinely clinical judgment stays with your people (by design); portal referrals need a one-click step until deeper integrations land

The comparison we'd ask you to make is not "software cost vs. free" — it's software cost vs. the hours your team currently spends and the clinician time that triage pulls away from billable care. If your clinic spends even 10 staff-hours a week on referral handling, the status quo already costs more than this would.

Disruption

What actually changes at your clinic — and what doesn't

New systems fail in clinics because of workflow disruption, not missing features. This is designed to be adopted in days and abandoned in a day if it isn't working:

Changes on day one

Your referral fax number forwards to the service (a one-line request to your fax provider — reversible the same day)
Staff opens one new browser tab: the triage queue
Portal/email referrals get a one-click "send to triage" step while staff views them as usual
About an hour of training, total

Stays exactly the same

Your EMR, booking system, and eReferral accounts — untouched
How referrers send you referrals — they notice nothing
Who makes decisions — same people, same authority
Your existing paper process — it keeps running in parallel during the pilot, so there is zero risk of a missed referral while trust is being established

The rollback plan is part of the offer: un-forward the fax line, export every document and decision log to you, and you're back to today — nothing held hostage, nothing lost.

The pilot

Prove it on your real backlog before changing anything

We'd propose earning the rollout in stages, with a no-commitment proof step first:

Weeks 1–3

Backlog proof — no workflow change at allWe take a sample of your existing backlog (redacted or under agreement), run it through the system offline, and sit down with your triager to compare. The deliverable is a number: "the proposal matched your triager on X% of items and would have cleared Y% of the backlog automatically." If the number isn't compelling, we stop there and you've lost nothing.

Weeks 4–12

Live pilot — rejection filtering on incoming referralsFax forwarding on, your rules loaded, staff approving every action, existing process running in parallel as the safety net. Weekly check-ins; we measure staff minutes per referral, time-to-first-response, and backlog burn-down. Free during the pilot.

After

Phase two — rule-based prioritizationWith filtering trusted, your prioritization policies order the accepted queue with deadlines and escalation. Only then do we talk about ongoing pricing, informed by measured results.

What we'd need from you

A sample of backlog referrals for the proof step, under an appropriate agreement
An hour with whoever does triage today, to write the first rules in plain language
Your privacy officer's review of the agreement and data-flow docs we bring
Honest feedback, especially when the system gets it wrong

Your input

What we're asking you to react to

This page exists to be challenged before anything gets built. Answer any, all, or none — every box below sends your note directly to us. Nothing here asks for patient information; please don't include any.

Question 1 · Volume & today's cost

What does referral handling actually look like at your clinic?

Roughly how many referrals a week, arriving how (fax / portal / email)? Who triages them and how many hours does it take? Is there a backlog today — and is any of it realistically never getting triaged?

Question 2 · The rules

Which rejection rules would you write first?

We used "no MVA cases" as the example. What are yours — funding types, age ranges, services you don't offer, required documents? And what share of incoming referrals would those rules catch?

Question 3 · Cost bar

What would this need to beat to be worth paying for?

Is the honest alternative another admin hire, clinician overtime, an outsourced service, or living with the backlog? Does "a few hundred dollars a month" feel obviously worth it, borderline, or expensive against that?

Question 4 · Disruption tolerance

What part of the rollout would worry you most?

Forwarding the fax line? Staff learning a new queue? The parallel-run period? Something about referrers noticing? Where's your line on acceptable disruption for a pilot?

Question 5 · Privacy sign-off

Who has to say yes, and what would they need to see?

Who acts as your privacy officer? Would the custodian/service-provider structure described above satisfy them — and what's missing from the privacy section that they'd ask about first?

Question 6 · After filtering

Is rule-based prioritization the right second act?

Once incomplete and out-of-profile referrals are being handled, what's the next most painful part of triage — ordering the queue, chasing missing info, booking handoff, reporting, something else?

Overall reaction

The blunt version is the useful version. Would you pilot this? What did we get wrong?

Private proposal

Every referral reaches a decision.None sit in a pile.