Investigators see the bill.
They rarely see the chart.
MFCUs, HHS-OIG, and qui tam counsel run on claims-only signals — billing patterns, frequency outliers, peer-group deviation. The unstructured chart notes that determine whether a code is actually supported sit unreviewable at scale. The variance hides there.
Lost to fraud, waste, and abuse across Medicare and Medicaid every year.
What state MFCUs and federal partners actually recover annually.
Of fraud spend that claims-only screens never surface — hidden in the chart.
The fraud isn't in the claim. It's in the gap between the claim and the documentation that's supposed to support it. Read the chart, and 70–80% of the loss becomes visible — with citations.
Five capabilities.
One engine that thinks in citations.
Built on Vertex AI under signed BAA. The investigator surface is plain English. The outputs cite the regulation.
Variance Engine
Reads every chart note line-by-line, cross-references it against the billed code, and flags the gap. The investigator gets a recovery estimate and the documentation excerpt that proves it.
Provider Risk Score
A daily 0–100 score per provider that fuses claims signal, documentation quality, and network patterns. Surfaces problem providers before a claims-only screen would trigger.
Program Integrity Copilot
Ask in plain English: "Show me providers who billed therapy while the patient was inpatient." Get cross-referenced results in seconds — every row cited to source.
Audit Packet Generator
One click drafts the Notice of Intent to Audit, builds the evidence bundle, attaches regulatory citations, and computes the recovery estimate. Days of work collapse to seconds.
Citation Grounding
Every assertion traces to claim ID, document SHA-256 hash, and a specific CMS Internet-Only Manual citation. Court-defensible by design — no black-box outputs.
Built for the people who
actually recover the money.
HHS-OIG
Workplan-aligned variance detection across Medicare A/B/C/D. Findings ship as DOJ-referable evidence packets with citation chains intact.
State Medicaid Fraud Control Units
Closes the staff capacity gap — surfaces 10× more leads per investigator-hour. Behavioral health, PCA fraud, DME patterns, eligibility integrity.
Qui Tam Counsel
A force-multiplier for relator counsel preparing FCA filings under seal. Citation-defensible evidence with chain-of-custody hashes attached.
Commercial Payors
Enterprise SIU deployment for national and regional plans. Pre-payment variance signals on high-risk providers. Network-quality benchmarking.
Court-defensible
by design.
Every assertion traces to the underlying claim, the document hash that supports it, and the specific regulatory citation that governs it. Three examples — exactly what shows up in an investigator's packet.
Inside a finding,
from raw data to a signed packet.
A finding is only as defensible as the chain behind it. Here's exactly how one moves from a claims feed to an investigator's desk — every step, every citation, every signature accounted for.
Receive
Claims and clinical documentation arrive under signed BAA. Encrypted in transit and at rest. Private network — no public ingress, no shared infrastructure.
Read
The AI parses every chart note line-by-line — not just the structured fields, but the unstructured prose that determines whether a billed code is supported.
Cross-reference
Every billed code is matched against the documentation that’s supposed to support it. Variance is scored, peer-group benchmarked, and ranked by recovery potential.
Cite
Each assertion is tied to (a) the source claim ID, (b) a SHA-256 document hash, and (c) the specific CMS Internet-Only Manual citation that governs it. Court-defensible by design.
Deliver
A drafted Notice of Intent to Audit lands on the investigator’s desk with the evidence bundle, citation chain, and recovery estimate attached. They review, sign, and send.