Vigil Health AI applies the same data analytics stack used by CMS to detect, prevent, and recover fraudulent Medicare payments — now available to commercial payers, state Medicaid agencies, and PBMs.
The U.S. healthcare system loses an estimated $68B+ annually to fraud, waste, and abuse (FWA). Schemes are increasingly run by organized crime, scaled with AI-generated claims, and designed to disappear before investigations begin.
A provider rapidly bills for never-delivered services, then vanishes. The 2023-2024 urinary catheter scheme tried to bill $4 billion in 60 days. CMS stopped 99% of the payments — by the time most payers react, the money is gone.
Fraud rings abroad recruit U.S. "front owners" to enroll as providers, route funds overseas, and disappear. GAO found these rings now target Medicare as a primary revenue source.
Fraudsters are now using generative AI to manufacture fake beneficiary records and submit thousands of plausible-looking claims per hour. Rules-based detection cannot keep up.
From the GAO, April 2026:
"Medicare uses data analytics to identify and prevent fraud... However, supplemental payers — private plans and state Medicaid agencies — paid at least $196,000 in cost-sharing on suspended-provider claims in 2023-2024 alone, because payment-suspension information was not shared."
— GAO-26-107799, Medicare: CMS's Use of Data Analytics to Identify and Prevent Fraud
Vigil Health AI operationalizes the seven model types the U.S. government uses to prevent $11.9B in fraudulent Medicare payments — packaged as a turnkey platform for commercial payers, state Medicaid, PBMs, and provider-sponsored plans.
Plug into existing claims, provider, and beneficiary feeds. No rip-and-replace. Our pipeline handles CMS, EDI 837/835, FHIR, and proprietary payer formats.
Every claim, provider, and beneficiary is scored in real time against the seven model types below. No black box — investigators see exactly which rule fired and why.
Recommend a precise administrative action: prepayment review, payment suspension, or referral to law enforcement. Case files generated automatically for SIU teams.
Every model is explainable, tunable, and mapped to the CMS Fraud Prevention System (FPS) taxonomy. Investigators see model output in plain English — not a confidence score.
Identifies sudden increases in specific services or codes that may indicate emerging fraud schemes.
Compares provider billing to same-specialty and same-geography peers to surface outliers.
Composite fraud-likelihood score per provider, updated nightly, with full feature attribution.
Historical fraud patterns applied to current claims to surface high-risk submissions in real time.
Analyzes clinical notes, medical records, and other unstructured data for signs of fraud.
Deterministic rules for known patterns — e.g., provider billing 24+ hours of care in a single day.
Detects anomalies between provider locations and beneficiary geographies — a key signal for fraud rings.
Visualizes referring↔billing provider relationships to detect collusion. Vigil Health AI exclusive.
Vigil Health AI's detection engine is built on the public CMS Medicare Provider Utilization dataset. Below is a live sample of what the engine surfaces on the actual federal data — no synthetic injection, no marketing claims.
Data: CMS Medicare Provider Utilization and Payment Data (public), 2022 release. Vigil scores are illustrative outputs of the prototype described in the linked business case; not a CMS product.
30-day pilot. We deploy our detection engine against a sample of your historical claims and produce a report showing the fraud, waste, and abuse we found — before you sign anything.