LetterAgent
Mail physical letters by just asking your AI assistant.
Home > Guides > Fax-to-EHR intake
Small clinics route incoming faxes through a cloud fax platform instead of a desk machine. Software classifies each page, reads it with OCR, matches it to existing patient records, and holds anything uncertain in a human review queue. LetterAgent only sends faxes, so clinics cannot receive faxes through it.
| Stage | What happens | Why it matters |
|---|---|---|
| Receive | Faxes arrive at a cloud fax number, never a paper tray | Pages cannot sit on a machine where anyone can read them |
| Classify | Software decides the document type: referral, lab result, refill request, insurance form | Refills and prior-authorization requests look identical on a pile, so sorting by hand is slow and error-prone |
| Extract | OCR pulls out named fields: patient name, date of birth, referral type, diagnosis codes, insurance details | Typed demographics and checkboxes extract cleanly; handwriting is less reliable |
| Match | The system finds the patient in the EHR, usually by name plus date of birth with fuzzy matching for typos | One fax with two plausible patients is worse than no fax at all |
| Review | Anything low-confidence lands in a queue for a coordinator to confirm or correct | Medication lists, allergies, and handwriting stay human-reviewed because the stakes are clinical |
| File | Confirmed data and the original fax image land in the patient chart | The image is kept too, since it is the legal source document |
The easy wins are demographics and insurance: full name, date of birth, health card or member ID, and the referring provider. Referral platforms also pull document-specific fields, like the test codes on a lab requisition or the diagnosis codes on an authorization request. The hard cases are exactly what you would expect: medication names that differ by a few letters, allergy lists jotted beside the printed lines, and anything handwritten. Those get extracted as suggestions, not facts, and a staff member verifies them. This is why a good intake setup reports confidence per field rather than pretending the read is perfect.
Because some mistakes are too costly to automate away. Matching the wrong patient to a lab result can lead to treatment based on someone else's bloodwork. A misread dose can land in the chart and get acted on months later. In practice, clinics tune the system so it auto-files the clean, high-confidence documents (legible forms for known patients) and parks everything else. Most vendors in this space describe the same split: a small fraction of documents take nearly all of the human attention, and the rest flows through. The coordinator still reads the uncertain pile the way they used to read everything.
Any vendor handling this pipeline for a US clinic signs a Business Associate Agreement, the contract that makes them legally responsible for protecting patient data. The intake platforms also log every step: when the fax arrived, what the software extracted, and who confirmed or corrected each field, so there is an audit trail from receipt to chart. One thing worth knowing as a patient: a fax sent to the wrong number by mistake is treated as a privacy breach by regulators, and some of the largest fax-related fines in healthcare came from misdirected faxes. Confirming the number before sending is the cheapest privacy control there is.
Ready to fax it? Fax with LetterAgent: describe it to your AI assistant, approve the exact quoted price, and it is sent.
LetterAgent is a working name. Prices include print, postage, and our service fee. US prices in USD, Canadian prices in CAD (Canadian prices exclude applicable sales tax, added at checkout). Currently in private testing.
Need help? Ask your assistant: it can check a job's status any time. If it ever can't handle something: hello@getletteragent.com