TL;DR: An optometry practice automates its inbound fax inbox in three moves: replace the physical fax machine with a cloud fax number, add an AI layer that classifies and extracts data from every document, and connect that layer to your EHR so documents file themselves and generate the right follow-up task. Start by auditing two weeks of fax volume to learn your actual document mix, automate your two or three highest-volume types first, and run the system alongside human sorting until you trust the accuracy. The work that decides whether this succeeds is defining routing rules and owning the exception queue, not the technical setup.
Start by counting what actually arrives
Most practices guess wrong about their own fax mix, and the guess shapes every decision that follows.
Spend two weeks logging every inbound fax by type. You don't need software for this — a tally sheet at the fax machine works. Count referrals, outside imaging and test results, co-management records from ophthalmology, vision plan correspondence, medical payer prior auth responses, records requests, optical lab confirmations, and whatever else shows up. Note roughly how many pages each runs and how long someone spends handling it.
Two numbers come out of this. The first is total daily volume, which tells you whether automation is worth the effort at all. A practice taking 15 faxes a day has a different problem than one taking 120. The second is concentration, which tells you where to start. In most optometry offices, three or four document types account for 70% or more of volume. Those are your automation targets. The long tail can stay manual for a while without hurting anything.
You'll also learn something uncomfortable: how much of the pile is duplicate. Referral sources routinely fax the same packet twice, and confirmation faxes from labs often restate information already in your system. Deduplication alone is sometimes worth a meaningful share of the time savings.
Get off the physical fax machine first
Automation can't read paper. The first structural change is moving to a cloud fax number so documents arrive as digital files.
This is the least interesting step and the one practices delay longest, usually because the fax number is printed on referral pads across every PCP office in town. You don't have to change the number — most cloud fax providers will port your existing number, which means your referral sources notice nothing. Do this before you evaluate triage vendors, not after; every triage product assumes digital input, and arriving at a vendor conversation still on paper means you're buying two things at once and can't tell which one is failing when something goes wrong.
Confirm three things with whichever provider you choose: that they sign a BAA, that they support your expected page volume without throttling, and that they expose an API or webhook. That last one is what the triage layer plugs into.
Automate your highest-volume document types first
Resist the urge to automate everything in week one. Pick two or three types and get them genuinely working.
For most optometry practices the right starting set is:
- Inbound referrals — highest operational value, because a missed referral is missed revenue and a missed patient
- Outside imaging and test results — highest volume in practices doing significant medical eye care
- Vision plan correspondence — highest annoyance, because it routes to a different person than everything else
Leave records requests, subpoenas, and one-off correspondence on the manual path initially. They're low volume, high consequence, and not worth the tuning effort until the core is stable.
For each type you automate, define three things explicitly: what the system should extract (which fields matter for that document), where it should file (which chart location and document tag), and who gets the task (a named role, not "the front desk"). Practices that skip the third definition end up with beautifully filed documents that nobody acts on.
Define routing rules for how your office actually works
This is the step that determines whether the system feels helpful or feels like a second inbox to manage.
Routing has two dimensions in optometry. The first is by document type — imaging goes to the doctor who ordered it, vision plan mail goes to the optical or billing lead, referrals go to whoever schedules. The second is by doctor, which matters more in multi-doctor practices where each OD has their own patient panel and their own preferences about what they want to see versus what staff can handle.
Write the rules down before you configure anything. A simple grid — document type on one axis, destination and owner on the other — surfaces the disagreements in your office before the software does. And there will be disagreements, because in most practices these rules live in one long-tenured person's head and have never been written down.
Also decide your confidence threshold, which is how certain the system must be before it acts without a human. Set it conservatively at first. A system that escalates too much is annoying for a few weeks; a system that misfiles confidently erodes trust permanently, and trust is hard to rebuild once staff decide they need to double-check everything anyway.
Run in shadow mode before you cut over
Don't flip the switch and hope. Run the automation in parallel with your existing manual process for two to four weeks.
In shadow mode the system classifies and proposes a destination for every document, but a person still does the real filing and compares the two. This gives you an honest accuracy number on your documents rather than the vendor's, and it surfaces the specific failure patterns you'll need to tune — usually a referral source whose form layout confuses the classifier, or a patient-matching problem caused by how your practice handles duplicate records.
Track two metrics during shadow mode: classification accuracy by document type, and time to resolution for documents that land in the exception queue. If exceptions are piling up because nobody owns clearing them, that's a staffing problem the software won't fix, and it's better to discover it now.
Cut over type by type as each one hits an accuracy level you're comfortable with. There's no reason all document types have to go live on the same day.
Honey Health's Fax Triage agent is built around this flow — classifying inbound documents into 30-plus types, extracting structured patient and clinical data, and filing into the chart with the right document-type tag and routed follow-up task, with uncertain documents surfaced rather than guessed at. The specific vendor matters less than confirming that whatever you pick handles the classify-extract-file-route sequence end to end rather than stopping at "we got the fax into a folder."
Decide what the reclaimed hours are for
The change-management piece practices consistently underestimate is what happens to the time you free up.
Automating document intake at a practice with meaningful fax volume typically returns several hours per week per person who currently handles it. The 2025 CAQH Index estimates fully automated administrative workflows save around 70 minutes per patient visit across the full transaction set, and flags $21 billion still tied up in manual phone, mail, and fax handling. Document triage is one slice, but it's a slice your staff feels daily.
If you don't name where those hours go, they dissipate. The practices that get the most out of this decide in advance: recall outreach, prior auth follow-up, optical sales, or reducing the overtime that was quietly covering the document backlog. Tell the staff the plan, too. "We're automating the fax pile" without a second sentence reads as "we're eliminating your job," and you'll get quiet resistance that shows up as nobody clearing the exception queue.
Someone also needs to own the exception queue by name. It's usually 10–20% of volume, it takes a fraction of the old effort, and it silently breaks the whole system if it's everyone's job and therefore nobody's.
Frequently Asked Questions
How long does it take to automate a fax inbox?
Plan on six to ten weeks end to end for a single-location practice — two weeks for the volume audit, one to two for cloud fax migration and configuration, and three to four running in shadow mode before cutover. Multi-location groups take longer because each site's routing habits need reconciling.
Do we have to change our fax number?
No. Most cloud fax providers port your existing number, so referral sources keep faxing the number printed on their pads and notice no difference. Porting usually takes a few weeks, which is worth starting early since it's the long pole in the timeline.
What if our EHR doesn't have an API?
You can still automate. Vendors handle API-less EHRs through document-drop integrations or automation layers that file into the interface the way a staff member would. The result is usually shallower — less structured data write-back — but the classification, extraction, and routing benefits still land.
Should we automate every document type at once?
No. Start with the two or three types that make up most of your volume and get them working properly. Low-volume, high-consequence documents like subpoenas and records requests should stay manual until the core is stable, and some practices reasonably leave them manual permanently.
How do we measure whether it's working?
Track three things: hours spent on document handling per week, time from fax arrival to action taken, and the size of the exception queue at end of day. The first shows labor savings, the second shows patient impact, and the third tells you whether the system is actually trusted or quietly being worked around.
