Anton Engine

Five layers of intelligence behind every transaction.

The Anton Engine is a proprietary AI system built into the core of Anton Payments. It doesn't bolt on after the fact — it processes every transaction through five distinct intelligence layers before a single dollar moves.

How It Works

Every transaction. Five layers. One decision.

TRS

Transaction Risk Scoring

Real-time severity scoring on every transaction

ORS

Onboarding Risk Scoring

Intelligent merchant and payee vetting at signup

TM

Transaction Monitoring

Rolling-window behavioral analysis in real time

AD

Anomaly Detection

Pattern recognition for emerging threats

GI

Graph Intelligence

Network mapping and relationship analysis

Deep Dive

Inside each layer.

01TRS

Real-time risk assessment on every transaction

The TRS layer evaluates each transaction against a proportional severity model, scoring from 0 (no risk) to 1000 (critical). Two-tier reason codes provide both machine-readable classification and human-readable explanations for every decision.

Proportional scoring 0-1000Two-tier reason codesSub-second evaluation

Transaction Analysis

Transactionpout_abc123 — $12,450.00
Risk Score142 / 1000
Reason CodeRC-TRS-001
ExplanationVelocity pattern within normal range
TRSORSTMADGI
02ORS

Intelligent vetting before a single dollar moves

The ORS layer evaluates every merchant and payee during onboarding. Identity verification, business validation, and risk profiling happen inline — not as a separate step. High-risk onboarding triggers enhanced due diligence automatically.

Inline with onboardingPersona KYB/KYC integrationAutomated enhanced due diligence

Onboarding Assessment

MerchantAcme Marketplace Inc.
ORS Score218 / 1000
KYB StatusVerified
Risk ProfileStandard — no enhanced review required
TRSORSTMADGI
03TM

Behavioral analysis that catches what static rules miss

The TM layer uses rolling-window analysis to track behavioral patterns over time. Instead of simple threshold rules, it builds behavioral baselines for each payee and flags deviations. Redis-backed state management ensures real-time performance at scale.

Rolling-window analysisBehavioral baselinesRedis-backed state

Behavioral Monitor

Payeepayee_tok_xyz789
Window24h rolling — 47 transactions
BaselineAvg $2,340 / txn — current within 1.2 SD
StatusNormal behavioral pattern
TRSORSTMADGI
04AD

Pattern recognition that evolves

The AD layer identifies emerging threats that no predefined rule would catch. It learns what normal looks like for each merchant's transaction patterns and flags statistical outliers. Critical anomalies trigger automatic re-evaluation through the ORS layer.

Adaptive learningStatistical outlier detectionAutomatic ORS re-evaluation on CRITICAL

Anomaly Flagged

Transactionpout_jkl012 — $47,200.00
AD Score847 / 1000
Reason CodeRC-AD-003
ExplanationAnomalous amount deviation — 4.7 SD above merchant baseline
TRSORSTMADGI
05GI

Network mapping that reveals hidden relationships

The GI layer maps relationships between payees, merchants, and transaction patterns. It identifies coordinated risk — multiple apparently unrelated entities that share hidden connections. This is the layer that catches what individual transaction analysis cannot.

Payee network mappingCoordinated risk detectionRelationship analysis

Network Analysis

ClusterGI-CLU-0042 — 7 connected entities
Shared Attributes3 payees share IP range, 2 share bank details
Risk AssessmentElevated — coordinated pattern detected
RecommendationManual review recommended for cluster
TRSORSTMADGI

Scoring

Proportional severity.
Not binary decisions.

Most fraud detection systems make binary decisions: approve or block. The Anton Engine scores every transaction on a 0-1000 scale, giving platforms the granularity to set their own risk thresholds.

142
05001000

0 — 300

Low Risk

Auto-approve. Standard processing path with full audit trail.

301 ��� 700

Elevated

Enhanced monitoring active. Optional manual review at platform discretion.

701 — 1000

Critical

Auto-block or mandatory review. ORS re-evaluation triggered automatically.

Two-tier reason codes

Every score comes with a machine-readable classification code and a human-readable explanation. Your team sees context. Your systems see structure.

RC-TRS-001Velocity pattern within normal range
RC-TRS-012Elevated amount for new payee corridor
RC-AD-003Anomalous amount deviation detected

See the Anton Engine in action.

Anton Payments is currently in early access. Request an invitation to be among the first platforms to integrate.