The Risk Graph: AI-Powered Network Intelligence

TrustRelay's Risk Graph combines machine learning with real payment outcomes across customers. As more teams run payments through TrustRelay, predictive models sharpen, fraud patterns surface earlier, and every customer benefits from collective intelligence—with full transparency into how decisions are made.

What is the Risk Graph?

The Risk Graph is TrustRelay's proprietary network of anonymized supplier profiles, payment outcomes, and behavioral signals. It connects data points across customers to identify high-risk vendors, fraud patterns, and compliance issues. All without exposing sensitive customer information.

🕸️

Cross-Customer Learning

Machine learning models train on anonymized payment outcomes across all TrustRelay customers. Every return, fraud flag, and successful payout improves predictive accuracy for everyone.

🔗

Intelligent Entity Resolution

ML-powered identity resolution matches suppliers across customers using EIN, bank accounts, addresses, and behavioral signals. Detects the same entity operating under multiple names or receiving payments from multiple buyers.

📊

Predictive Risk Scoring

AI models analyze vendor history, payment patterns, and network-wide signals to generate predictive risk scores. Suppliers with suspicious patterns receive higher risk scores, triggering policy holds with explainable reasoning.

🔄

Continuous Learning Loop

Every decision outcome flows back into the Risk Graph. Models continuously improve as they learn from real-world results—creating network effects that benefit all customers.

How Network Effects Strengthen Over Time

1

Initial Customer

First customer makes payouts. TrustRelay captures outcomes (ACH returns, wire failures, fraud reports) and builds initial supplier profiles.

2

Supplier Portability

Second customer encounters the same supplier (matched by EIN or bank details). TrustRelay surfaces historical risk signals from first customer's experience.

3

Pattern Detection

As more customers join, patterns emerge: certain suppliers consistently return payments, change bank accounts frequently, or operate under aliases.

4

Flywheel Effect

Each new customer and payment outcome strengthens the graph. Risk scores become more accurate. Fraud is detected faster. All customers benefit from collective intelligence.

As more customers use TrustRelay, the Risk Graph becomes more accurate and reliable for everyone. Each transaction adds real-world signals that improve risk assessment over time delivering insight that's difficult to reproduce without the same depth of shared payment outcomes.

Risk Graph Use Cases

Fraud Prevention

Detecting Supplier Impersonation & BEC

The Risk Graph identifies suppliers who have changed bank accounts multiple times across customers—a common indicator of Business Email Compromise (BEC) or account takeover fraud. When a supplier requests a bank account change, TrustRelay surfaces historical patterns and triggers additional verification workflows.

Example: Supplier "Acme Logistics" changes bank details at Customer A. Risk Graph shows the same EIN changed accounts 3 times in the past 90 days across other customers. TrustRelay holds the payout and alerts the AP team.
Return Prediction

Avoiding Returned Payments

Suppliers with histories of ACH returns (invalid accounts, closed accounts, incorrect routing numbers) receive higher scrutiny. The Risk Graph flags these suppliers before payout, reducing return rates and reconciliation overhead.

Example: New supplier onboards at Customer B. Risk Graph matches EIN to prior returns at Customer A. TrustRelay requires micro-deposit verification before first payout.
Sanctions & Compliance

Cross-Customer Sanctions Alerts

When a supplier is flagged for sanctions exposure at one customer, the Risk Graph propagates alerts to all customers transacting with the same entity. This ensures compliance even when suppliers use different legal names or addresses.

Example: Customer A detects sanctions hit on supplier X. Risk Graph matches supplier X to alias "Company Y" at Customer B. TrustRelay proactively alerts Customer B before next payout.
Risk Tiering

Dynamic Risk Scoring for New Suppliers

New suppliers with no payment history at a given customer can still receive risk scores based on cross-customer telemetry. This enables faster onboarding for trusted suppliers and stricter controls for high-risk entities.

Example: Customer C onboards a new supplier. Risk Graph shows 500+ successful payouts across other customers with zero fraud or returns. TrustRelay assigns low risk score, expediting first payment.

Privacy & Anonymization

TrustRelay operates the Risk Graph with strict privacy controls. Customer-specific data (invoice amounts, payment dates, internal identifiers) is never shared across customers. Only anonymized risk signals and outcome telemetry contribute to the graph.

What We Share

  • Anonymized supplier identifiers (hashed EINs, bank hashes)
  • Payment outcome telemetry (return, success, fraud flag)
  • Behavioral risk signals (account change frequency, verification failures)
  • Sanctions and compliance alerts (no financial details)

What We Never Share

  • Payment amounts or invoice details
  • Customer identities or business relationships
  • Internal purchase orders or approval workflows
  • Contract terms or pricing information

Risk Graph Roadmap

v1

Foundation

  • Basic supplier identity resolution (EIN, bank account)
  • Outcome telemetry capture (returns, fraud flags)
  • Simple risk scoring (return rate, account change frequency)
  • Manual fraud-pattern investigation workflows
v2

Network Effects

  • Advanced entity resolution (address matching, behavioral fingerprinting)
  • Automated fraud-pattern detection (clustering, anomaly detection)
  • Cross-customer alert propagation for sanctions and fraud
  • Risk score APIs for real-time decisioning
v3

Intelligence Layer

  • Predictive risk modeling (ML-powered fraud and return prediction)
  • Supplier reputation scores visible to all customers (opt-in)
  • Industry-specific risk benchmarks (e.g., 3PL fraud patterns)
  • Integration with external fraud consortiums and data providers

Learn How the Risk Graph Works

Schedule a technical deep dive with our product team to see Risk Graph use cases, anonymization practices, and network effects in action.

Schedule Technical Deep Dive