The same narrative, in forty unrelated patients' charts.
Provider fraud medical record review surfaces cloned note text, templated treatment protocols, and billed-versus-documented gaps across an entire provider's claim set, not one file at a time. Duplicates are free here — so a 200,000-page ring investigation costs what a single case costs elsewhere.
The rulebook is published. So is the referee.
We do not build for a case type until we can name the standard that defines a correct output and the person who grades the file against it. In SIU provider-fraud investigation work, both are already in statute.
Three steps between the claim set and the referral.
Send the claim set
Every file in the provider's claim set, in any format, at whatever volume it exists — a single suspicious file or the full 200,000-page ring investigation. In the demo case, one file: 342 pages across 2 packets, with a wrong-patient page quarantined before the read starts.
The platform flags the pattern, every line cited
Cloned note text, templated treatment protocols, and billed-versus-documented gaps are surfaced across the entire claim set, not read one file at a time. Duplicates are free, so a 200,000-page ring investigation costs what a single case costs elsewhere. Every match links to its source page in every file it touches.
SIU investigates and decides
The output is a page-cited signal, not a fraud determination. The investigator, not the platform, decides whether it becomes a referral to the DOI fraud bureau, NICB, or law enforcement.
A pattern match that knows it isn't an accusation.
We surface cross-claim narrative matches and cite every page they come from. We do not accuse a provider, score fraud risk, or make a referral. That decision, and the professional judgment behind it, stays with the SIU investigator.
In the demo case, page 140 turned out to belong to a different patient. The scan flags it and quarantines it in its own line, instead of folding it silently into the read — the same discipline it applies when a note template repeats across forty unrelated charts.
The rules the platform never breaks.
EVERY LINE CITED
HIPAA, under a signed BAA
Every file is handled under our Business Associate Agreement, from the first byte.
Never trains a model
Your records are never used to train any AI model — ours or anyone else's.
Every line cited
If we can't cite it, we don't say it. Every sentence links to its source page.
Deleted after delivery
Files are deleted 30 days after delivery, with a full audit log of every access.
Four capabilities behind every fraud signal.
The cross-claim scan is one deliverable of the same platform that reads, sorts, and cites the whole record.
Medical Records OCR
Every provider production and claim file read in full — PDFs, scans, handwritten notes — nothing skipped, nothing guessed, across however many files the claim set runs.
Medical Chronology
The treatment timeline builds itself from each provider's file, synced to every source page — so a templated protocol shows up as the same shape, visit after visit, chart after chart.
Verifiable AI Citations
If we can't cite it, we don't say it. Every cloned-note match and billed-versus-documented gap links to the exact page it was pulled from, in every file it touches.
Medical Summary Reports
The page-cited findings format into a referral-ready packet — your template, your letterhead — for the SIU investigator to take to the DOI fraud bureau or NICB.
Provider fraud record review, answered.
Send one file. We'll tell you what we can read.
No obligation. If the case type is buildable we'll scope it; if it isn't, we'll say so.