hellofax.ai / The fax workflow
The fax line doesn’t slow down because the front desk is busy. hellofax turns the incoming pile into a worked queue: each document classified, named after its patient, carried through statuses until someone finishes it — and every step on the record, HIPAA compliant from storage to screen.
Three things happen to every fax, in order, without anyone standing at the machine.
From your RingCentral fax lines or a drag-and-drop upload, straight into private, practice-scoped storage. No shared inbox, no paper tray, no PDF folder on somebody’s desktop.
A model classifies it into your practice’s own folders — the themes your faxes actually arrive in — and lifts out the patient’s name, date of birth, phone and insurance. Every fax is named after its patient, everywhere it appears. Uncertain reads are routed to human review instead of guessed.
Call attempts are tracked on the fax itself. Patients get a text when their referral is filed or a callback keeps missing them. Referring offices and pharmacies get their letters back by fax — composed from your templates, sent in a click.
A fax is a task wearing a coversheet. hellofax treats it that way.
Every practice’s fax mix is different, so your folders are built around the themes that actually arrive on your line — a psychiatry practice sees Spravato and TMS folders; yours will see its own. Filing is automatic from day one, and when a reviewer moves a fax, the correction teaches the classifier your practice’s patterns.
The queue reads like a patient list, not a pile of timestamps. One naming rule everywhere — on screen, in search, on the downloaded file — so the same fax never wears two names in two places.
Every referral sits in exactly one bucket — To call, Patient will call back, Needs PCP auth, Unable to contact, Completed — moved with one click, and a patient who rings back lands in To call on their own. The dashboard shows the day’s arrivals against the day’s calls, so “did we get to everyone?” has an answer by end of day — and you can walk back through any past day’s numbers, frozen as that day ended.
Approve, deny, or route to an appointment. Write on the request in the viewer and fax the annotated page straight back to the pharmacy — and when the answer is “come in first”, the patient is texted to book without anyone dialling.
Every fax carries a status — needs review, active, completed — and moves through the office like a task, not a document: bulk actions, notes threads on the fax itself, and an activity log of who did what. And what happens to each category — the statuses it moves through, the letters and texts it can trigger — is configured to your practice’s own workflow, not a one-size template.
Turnaround, referral funnel, volume by folder, team activity — and a plain-language Ask box that rides on every page of the app and understands date ranges, so “how many referrals did we finish last month?” is a question, not a spreadsheet project. The same figures can arrive by email daily, weekly or monthly, so nobody has to remember to look.
A referral and a refill are each half of a conversation with somebody outside the building — the office that sent it, or the pharmacy that asked. hellofax carries your half, and keeps proof it went.
Every outcome on a referral has a letter behind it, on your letterhead, faxed back the moment a person sets the status — appointment scheduled, unable to reach, patient declined, out of network, needs PCP authorization. Nobody stands at the machine, and for anything that list doesn’t name, you write the words and hellofax sets them on the letterhead and sends them.
Two details matter more than the automation. An appointment is recorded only once its confirmation letter has actually transmitted — a booked visit the referrer was never told about is worse than a blank row. And an automatic letter goes only to a fax number the referral document itself stated as one: if it named none, the letter is refused rather than guessed at, because a misdialled fax puts a patient’s name on a stranger’s machine and cannot be recalled.
When the carrier confirms delivery, the transmitted PDF is filed on that patient’s own thread — the exact page the other office received, openable long after the carrier has forgotten it.
Open the refill request, mark it up in the viewer — approval and denial stamps, checkboxes, a line of text — and fax the annotated page straight back to the pharmacy. The document is the reply; there is nothing to retype and no second system to open. A confirmed reply archives the request as dealt with, and a pharmacy whose machine was out of paper can be sent the same reply again.
When the answer is “we need to see them first”, the patient hears it without anyone dialling: marking the request needs appointment texts them to book. It re-reads the request a moment before sending, so correcting a misclick beats the message out the door — and it can go at most once per request, ever.
Every request carries the medication, the frequency and the patient’s number read off the fax, so the decision is made from the row rather than from the document.
A fax carrier remembers about 30 days. A medical record has to remember longer — so hellofax keeps its own copies.
Sent fax PDFs, voicemail recordings and the practice’s call history are stored by hellofax, in the practice’s own private storage — so the record survives the carrier’s ~30-day retention, its rate limits, and its outages. Proving a letter went out in March is a lookup, not an archaeology project.
The letters hellofax sent and the faxes staff sent by hand from the phone system’s own app, together on one page. AI reads who each one was about, so the list is searchable by patient name, and every entry opens as the exact PDF that transmitted — filed on the patient’s own thread.
The model does the reading; your team keeps the judgement. The line between the two is drawn deliberately.
Every classification carries a confidence score. Below the threshold, the fax goes to Needs review — a person decides, the machine doesn’t guess. When a reviewer refiles a fax, that correction becomes part of the classifier’s guidance, so recurring mistakes actually stop recurring. And a re-run of the classifier never overrides a decision a human already made.
Opening a fax, moving it, completing it, writing a note — all of it lands on the activity log, on an audit trail nobody can edit. Faxes are archived, never destroyed, so history survives. Search, filters and CSV export mean the answer to “what happened to that referral from March?” is a lookup, not an investigation.
The fax machine isn’t going anywhere. The hours it eats can.