AI for medical file exchange
Every medical file has always moved blind. Faxes, scans, referrals and images travel from one system to the next while people open them, read them, sort them and type them in. MedFile puts AI inside the transfer. It reads each file in motion, knows what it is and who it belongs to, protects what's sensitive, and delivers it to the right place, ready to use.
HIPAA BAA included · People approve what AI isn't sure about · Your data never trains shared models
87%
filed with no human touch
99.2%
patient-match accuracy
< 30 sec
from arrival to the right chart
The problem
Fax lines, secure email and SFTP servers deliver files, and that's all they do. They can't tell a referral from a denial letter, notice a missing signature, or see that a lab result went to the wrong practice. So every file lands on a person's desk to open, read, rename, file and re-key. The work is slow, expensive and easy to get wrong, and the patient waits while it happens.
Staff spend their day indexing faxes and attachments instead of helping patients.
A misfiled record is a safety risk and a privacy breach waiting to be reported.
Missing pages and signatures surface days later, as a denied claim or a delayed procedure.
A clinician gets 200 pages of outside records and minutes to find what matters.
The opportunity
35%
of documents sent to healthcare facilities still arrive by fax
Documo, 2025
88%
of practitioners say fax delays hurt patient care
Documo, 2025
~50%
of specialty referrals are never completed
2017 study, via Fierce Healthcare
$21.8B
healthcare document AI market by 2034, up from $4.2B
Market Intelo
Before and after
TODAY
WITH MEDFILE
Every step is done in seconds while the file is in transit, and every step is logged.
How the AI works
MedFile's AI works at every stage of the file's life, in transit and at rest. It sits on the path between your systems and your partners, so no file needs a person to touch it just to get where it's going.
Faxes, scans, handwriting, PDFs, HL7, FHIR, DICOM headers and images. The AI turns each one into text and data it can work with.
It classifies the document, matches it to the right patient and provider, and pulls out the diagnoses, medications, orders and dates inside.
It finds protected health information, strips out what a recipient doesn't need, and stops a send that looks misdirected or unusual before it leaves.
It converts the file to the format the receiving system expects and files it to the right chart, queue or person, with missing pieces flagged.
Stored files stay searchable in plain language, summarized for clinicians, retained on schedule, and watched for unusual access.
The story
Maria's primary care doctor faxes a cardiology referral. Here's what MedFile's AI does before anyone at the cardiology practice opens it.
Six pages with handwritten notes. The AI reads every page, including the margins.
Classified as a cardiology referral, matched to her existing record, and the ECG separated out as its own document.
The missing signature gets flagged, and a request goes back to the sender automatically, days before it could become a denial.
A short summary sits on top: reason for referral, current medications, relevant history. Each line links back to its source page.
The referral lands complete in the scheduling queue. Every AI step is recorded in the audit log.
The fax machine on the other end didn't change. What happened to the file in between did.
AI capabilities
Reads faxes, scans and handwriting, then sorts each document into referrals, results, orders, authorizations, records requests and more.
Matches every file to the right patient using several identifiers, and sends uncertain matches to a person instead of guessing.
Turns unstructured pages into structured data, such as problems, medications, allergies and orders, delivered as FHIR resources your EHR can use.
Finds protected health information in every file and removes what a recipient doesn't need, so each send shares only the minimum necessary.
Knows what each payer and specialist requires. It catches missing pages, signatures and fields before a packet goes out, not after a denial comes back.
Condenses long outside records into a brief for the clinician, with every statement linked to the page it came from.
Learns your normal patterns of who sends what to whom. It holds a send that looks wrong and spots unusual bulk downloads or ransomware-like activity.
Chat with a whole patient file in plain language, like "When was the last echo?" or "Which referrals are unsigned?" Every answer points to the exact source page, within each user's permissions.
Builds a timeline of a patient's history from every file received, with visits, results, procedures and medications in order from oldest to newest.
Finds duplicate pages and repeated records across faxes, scans and outside charts and sets them aside, so reviewers read each page once.
Explains dense clinical notes, pathology results and radiology reports in plain terms for patients, families and generalists, without losing the clinical source.
Spots missed follow-ups, overdue screenings and orders that never got results, and flags them for the care team.
Benefits
FOR STAFF
The sorting, indexing and re-keying are done before files reach a desk. People handle the exceptions instead of the whole pile.
FOR CLINICIANS
Outside records arrive summarized and sourced, so the visit starts with what matters.
FOR PATIENTS
Referrals and authorizations move in hours instead of days, and patients stop repeating their history at every stop.
FOR COMPLIANCE
Misdirected sends are caught before they leave, only the minimum necessary is shared, and every action is on record.
FOR REVENUE CYCLE
Claims attachments and prior authorizations go out complete the first time, and cash comes in sooner.
FOR MEDICAL-LEGAL
Attorneys and record reviewers get organized, de-duplicated timelines with every fact linked to its page, instead of boxes of unsorted records.
FOR INSURANCE AND CLAIMS
Claims adjusters and utilization reviewers see summaries, gaps and source pages together, so every determination is backed by the record.
FOR IT
One platform replaces fax servers, SFTP boxes and hand-built interfaces, and it watches its own traffic for threats.
87%
of inbound documents filed with no human touch
120 hrs
of staff time returned per practice, per month
4 days
faster from referral to scheduled visit, on average
32%
fewer denials from incomplete packets
Responsible AI
In healthcare, trust isn't optional. MedFile's AI shows its work, knows its limits, and keeps protected health information protected.
Every decision carries a confidence score. Anything below your threshold goes to a person for review.
See why a file was classified, matched or routed, down to the text the AI relied on.
Patient data is never used to train shared models and is never sent to public, consumer-grade AI tools. Processing stays inside MedFile's HIPAA environment.
AI actions are logged like human ones: what it did, to which file, when, and with what confidence.
Signed BAA, AES-256 at rest and TLS 1.2+ in transit, with SSO, MFA and role-based access.
Independent third-party assessments in progress, aligned to the HITRUST CSF.
For developers
Every file that passes through the MedFile API comes back with what the AI learned: document type, patient match, extracted data, PHI found and routing decision. Each comes with a confidence score, ready for your own apps and AI agents.
{
"file": "inbound-fax-0412.pdf",
"type": "referral.cardiology",
"confidence": 0.97,
"patient": { "match": "verified" },
"extracted": ["medications", "reason", "ecg"],
"phi_redacted_for": ["payer"],
"issues": ["signature_missing:p2"],
"routed_to": "cardiology.intake"
}
Pricing
$299/month
AI fax and document intake with classification, patient matching and secure sharing.
Start a pilot$899/month
Adds data extraction, completeness checks, clinical summaries, PHI redaction and imaging transfer.
Start a pilotCustom
Health systems, labs and payers: custom AI models for your documents, threat detection, data residency and a dedicated team.
Talk to salesIn a pilot, MedFile's AI runs on your own documents, under a BAA, so you see the results on your files and not a demo deck.