Now Available · Enterprise & API

Fraud detection built for regulated environments

SENTINEL's eight-agent AI pipeline detects financial transactions, identity fraud, insurance claims, and insider threats simultaneously — with full explainability, human oversight, and an immutable audit trail built in by design.

1.2s

Avg detection time

Fraud domains

99.6%

SLA compliance

SR 11-7

Aligned governance

🛡️

$4.2M

Fraud prevented today

Live Pipeline · Case TXN-9821

TXN-9821 · Transaction

2 seconds ago

Velocity abuse + impossible travel — 4 merchants, 3 countries, 11 minutes. Card-not-present MCC 7995.

CRITICAL — HUMAN REVIEW REQUIRED

ORCH

TXN

EXPL

REV

LEARN

Explainability Agent Output

Pattern consistent with credential compromise followed by rapid liquidation. Velocity 6.0σ above baseline. Impossible travel confirmed. Recommend immediate card suspension.

The Platform

Four fraud domains.
One unified pipeline.

Most fraud platforms cover one domain well. SENTINEL covers all four with specialized agents that share signals, improving accuracy across every detection category.

💳

Financial Transactions

Detect payment fraud in real time using statistical signal extraction across velocity, geography, merchant risk, and behavioral baselines.

Velocity anomaly detection

Impossible travel analysis

Amount outlier scoring

Card-not-present MCC risk

🔐

Identity & ATO

Surface account takeover with device fingerprinting, behavioral biometrics, credential stuffing patterns, and session integrity monitoring.

Device fingerprint analysis

Keystroke dynamics

MFA fatigue detection

Tor / proxy identification

📋

Insurance Claims

Identify staged losses, duplicate claims, and provider network fraud using graph analysis across claimant, policy, and provider data.

Provider network graph

Policy inception analysis

Duplicate SSN detection

Cross-claim linkage

🔍

Insider Threat

Monitor privileged users for exfiltration patterns, access anomalies, and pre-departure behavior with UEBA integration and contextual HR risk signals.

Bulk export detection

UEBA score integration

Departure risk signals

DLP policy monitoring

How It Works

Eight agents. One governed pipeline.

Every case flows through a structured multi-agent pipeline with explainability and human review built in at the architecture level — not bolted on.

Ingest & Route

The Orchestrator Agent receives the case, validates it, classifies the fraud domain, and routes it to the appropriate specialist agent — all in under 100ms.

Orchestrator Agent

Signal Extraction

The domain agent runs six signal categories with statistical scoring, behavioral baselines, and network graph analysis to produce a calibrated risk score.

Domain Agents ×4

Explain & Audit

The Explainability Agent generates a structured, FCRA-sufficient rationale for every case — stored as an immutable audit record with ISO 8601 timestamps.

Explainability Agent

Human Review Gate

High-risk cases are hard-gated to human review before any consequential action. No auto-block, no auto-deny. People make the final call — with full context.

Review Agent

Why SENTINEL

Built for the exam,
not just the sale.

Most fraud vendors solve the detection problem. SENTINEL solves the governance problem — the one that shows up in your regulatory exam.

⚖️

SR 11-7 Model Governance — Included

A complete Model Governance Policy, Architecture Decision Record, independent validation framework, and Learning Agent change control pipeline ship with SENTINEL. Your model risk team has documentation from day one.

📜

Immutable Audit Trail — By Design

Every agent action, risk score, and analyst disposition is written to a tamper-evident, append-only audit log with 7-year retention. Regulators can reconstruct any decision in full. This is enforced at the architecture layer, not policy.

👤

Human Review Is Non-Negotiable

Cases above the risk threshold cannot be auto-closed, auto-blocked, or auto-reported. The system enforces human accountability — protecting you from automated adverse action liability under FCRA and ECOA.

🔬

Every Decision Is Explainable

The Explainability Agent generates structured, plain-language rationale for every case — with signal weights, confidence scores, and recommended actions. Suitable for adverse action notices without additional processing.

Live Performance Metrics

Detection Latency

p95 across all domains

1.2s

False Positive Rate

30-day rolling average

3.1%

False Negative Rate

Validated cases

0.4%

SLA Compliance

Human review within tier

99.6%

Audit Uptime

DynamoDB PITR enabled

99.99%

Domains Covered

Simultaneous detection

4 of 4

How SENTINEL compares

Against typical point solutions in each fraud category.

Capability	SENTINEL	Legacy Rule Engines	Single-Domain AI	SIEM / DLP Tools
All 4 fraud domains	✓ Unified	Partial	1 domain only	Insider only
Real-time detection (<2s)	✓ 1.2s avg	Batch / delayed	✓ Varies	Alert lag
Explainability per case	✓ Every case	Rule match only	Score only	Log correlation
SR 11-7 model governance	✓ Included	Not applicable	Custom build	Not applicable
Immutable audit trail	✓ 7-year retention	Limited	Varies	Log files only
Human review enforcement	✓ Architecture-level	Manual process	Optional	Analyst discretion
SaaS + API delivery	✓ Both	On-premise	✓ Usually	On-premise

What Customers Say

Trusted by fraud and risk teams.

"The governance documentation alone saved us six months of internal work. Our model risk team had a complete SR 11-7 framework before we even finished onboarding."

Head of Model Risk

Tier 2 Regional Bank

"We evaluated six fraud vendors. SENTINEL was the only one that could answer our compliance team's questions about the Learning Agent change control process out of the box."

Chief Product Officer

Embedded Finance Fintech

"The insider threat domain pays for the entire platform. We caught a pre-departure exfiltration event in the first two weeks that our SIEM had missed for three months."

CISO

Large Insurance Carrier

Pricing

Simple, transparent pricing.

All plans include the full eight-agent pipeline, governance documentation, and SLA-backed uptime. No per-agent fees. No hidden compliance add-ons.

Starter

$4,500

per month · billed annually

For fintech companies and growing businesses beginning their fraud detection program. Covers two fraud domains with full API access.

✓2 fraud domains (choose any)

✓Up to 50,000 cases/month

✓Full REST API access

✓Explainability on every case

✓90-day audit log retention

✓Human review queue

✓Email support (48hr SLA)

✓Standard onboarding

Common questions.

Is SENTINEL a black-box AI system?

No. The Explainability Agent generates a structured, plain-language rationale for every case — including signal weights, confidence scores, and the recommended action. Every decision is fully auditable and retrievable. This is enforced at the architecture level, not as an optional report.

How does SENTINEL satisfy SR 11-7 requirements?

SENTINEL ships with a complete Model Governance Policy aligned to SR 11-7 / OCC 2011-12, including pre-production validation requirements, performance monitoring thresholds, Learning Agent change control, and model retirement criteria. Your model risk team has a full governance framework from day one.

Can SENTINEL automatically block or deny transactions?

Cases above the risk threshold cannot be auto-blocked, auto-denied, or auto-reported without human analyst review. This is enforced at the architecture layer — it is an architectural decision, not a configuration setting. This design protects your organization from automated adverse action liability under FCRA and ECOA.

How is the audit log protected from tampering?

The audit log is written to AWS DynamoDB with an IAM condition expression that prevents any overwrite or deletion. Point-in-time recovery is enabled. The application layer uses Pydantic frozen models that raise an exception if mutation is attempted. Immutability is enforced at three independent layers.

What happens when the Learning Agent receives analyst feedback?

Feedback passes through four integrity controls before entering the pipeline: HR investigation quarantine, a 15% per-analyst volume cap, systematic bias detection, and conflict-of-interest screening. Material changes to detection patterns require Model Risk Committee approval before any production update is applied. Nothing deploys automatically.

Is there an API for integration with existing systems?

Yes. SENTINEL exposes a full REST API with auto-generated OpenAPI documentation. All four fraud domains are accessible via a single POST /cases endpoint. Webhook support for real-time disposition callbacks is available on Professional and Enterprise plans.

What cloud infrastructure does SENTINEL run on?

SENTINEL runs on AWS using ECS Fargate, DynamoDB, SQS, and Secrets Manager — all deployed within a private VPC. No public IPs on application containers. Enterprise customers can request private cloud or on-premise deployment with full Terraform infrastructure-as-code provided.

How long does implementation take?

Standard SaaS onboarding takes 5–10 business days including API integration, domain configuration, and analyst training. Enterprise deployments with custom signal development and private cloud setup typically take 4–6 weeks. Governance documentation review and model validation are included in both timelines.