AI document filing for medical practices

From scanner to chart. Automatically.

doc_autofile reads scanned documents, identifies the patient, classifies the document type, and files it into the correct folder in your EMR — in about 20 seconds, with no manual searching or sorting.

~20sscan to chart 34eCW categories 0identifiers sent off-network

The problem

The last mile between your scanner and your EMR is still manual.

Every fax, consult note, outside lab, and insurance card that arrives at a practice has to be opened, read, matched to a patient, categorized, and filed by hand. For a busy front office, that's hours of repetitive work each week — and every manual step is a chance to file a document to the wrong chart.

5 stepsopen, read, match, categorize, file — for every single document
Hours / weekof repetitive front-office work spent on manual import
1 wrong clickis all it takes to file a document to the wrong chart

How it works

Scan. Press a button. Done.

Watch one document make the trip — the same pipeline every page goes through, end to end, in about twenty seconds. Scroll to follow it.

Step 1
Scan

Staff scan documents as usual. Files land in a watched folder — no new habits to learn.

Step 2
Read

OCR extracts the text; AI identifies the document type, date, and patient details.

Step 3
Match

Identity is confirmed against your own records with fuzzy matching and confidence scoring. Below threshold? Human review — never auto-filed.

Step 4
File

High-confidence documents are filed into the correct patient folder and EMR category, with a PHI-free filename.

Features

Built like a filing cabinet, not a black box.

Every decision the engine makes is scored, logged, and explainable — and anything it isn't sure about goes to a human.

34 categories

Standard eCW document categories

Lab Documents, Consult Notes, Insurance Card, X-Ray Documents, Referral Notes, and more — filed straight into the structure your staff already use.

≥ 85% or a human

Confidence-based routing

Auto-file only above a configurable threshold. Everything else goes to manual review with a plain-English explanation of why it was flagged.

sha-256 hashing

Duplicate detection

Every document is fingerprinted before processing. The same page is never filed twice — even if it's scanned twice.

DOCTYPE_DATE_ID.pdf

PHI-free filenames

Patient identity lives in the chart, not the filename. Filed documents carry only a type, a date, and a unique ID.

3 logs, 1 trail

Audit logging

A staff-readable daily ops log, an automatic end-of-day summary, and a separate anonymized, hash-based audit trail designed for HIPAA environments.

on-prem

Runs inside your network

Lightweight enough to run on a single small device on-premises — no servers to rack, no patient database in someone else's cloud.

Privacy & security

Designed for HIPAA environments.

Patient identification never leaves your network. All patient matching happens locally against the practice's own records. Only document content is sent for AI classification.

  • No patient demographics sent off-network
  • No chart data, no identifiers, no MRNs
  • Staff initiate every processing run — a human is always in the loop
  • Anonymized, hash-based audit trail for every document
Your practice network
Patient records & matching
names · DOBs · MRNs · charts
stays local
Filed documents & folders
patient folders · 34 eCW categories
stays local
Audit & ops logs
hashed · anonymized · staff-readable
stays local
only this crosses the line → document content, for classification

About

Built inside a working practice. Running in production today.

doc_autofile was built inside a working medical practice by the IT staff who lived the problem — not designed in a boardroom. It runs in production today, filing real documents for a real clinic.

We're currently developing direct EMR integration for eClinicalWorks v12.

Interested?

Whether you're a practice drowning in faxes or an EMR partner, we'd like to hear from you.

Get in touch

replies within 1–2 business days