Source-level evidence
Every signal, score and recommendation links back to the underlying record — the email, the filing, the transaction, the document — with the source visible at a click.
Compass connects claims, policyholders, vehicles, providers, experts and witnesses to detect recurring fraud patterns that are invisible in isolated files.
The same vehicle, the same medical provider, the same witness, the same broker — reappearing across claims, across portfolios, across years. Each file in isolation looks legitimate. The pattern only appears in the network.
Ingest claims, policies, providers, vehicles, experts and external sources
Resolve entities and link claims through shared parties and assets
Detect recurring patterns, hubs and improbable coincidences
Score claims with relational fraud indicators and evidence
Recommend SIU review with full reasoning
Each claim passes individual triage. Compass surfaces a recurring medical expert appearing across all five, plus two shared addresses and one repeated witness — and recommends the cluster for organized-fraud investigation.
Every signal, score and recommendation links back to the underlying record — the email, the filing, the transaction, the document — with the source visible at a click.
No black-box outputs. Each priority states why it was promoted, which signals contributed, which were declassed, and which model version produced it.
Signals, model versions, reviewers and actions are logged and reconstructable later — ready for internal control, regulators and your own governance.
Compass recommends. Authorized people decide. Every consequential action requires a named reviewer who can confirm, override or annotate the reasoning.
Claim decisions, denials and SIU referrals remain with authorized adjusters and investigators. Compass surfaces patterns and supplies evidence — it does not adjudicate claims.
30 minutes. We'll wire Compass to a representative slice of your environment — and walk you through the relationships, signals and actions it would surface today.