Live Alerts Connecting…
🔕
No alerts yet. Run a scan to see live results.
🔍

Running Compliance Scan

Analyzing transactions for AML risk patterns...
1
Loading demo transactions
500 transactions across 15 accounts
2
Running AML detection rules
Structuring · Rapid movement · Geographic risk · Velocity · Round amounts
3
Scoring risk & generating explanations
SHAP values calculated for each flag
4
Scan complete
Building results...
--
transactions flagged for AML review
--Critical
--High Risk
--Scanned
--Scan Time
Step 1 of 3

Transaction Overview

Your dashboard shows all monitored transactions across accounts, with risk scores and flag counts updated after every scan.

🎯 See AML detection in action

Run a live compliance scan on 500 sample transactions and watch Clerk catch structuring patterns, rapid movement, and geographic risk — with SHAP explanations for every flag.

1 Load 500 transactions
2 Run 5-rule AML engine
3 Review flagged alerts
No setup required · instant results

Compliance Score — 30 Day Progression

Good (80+)
Warning (60–80)
At Risk (<60)
Total Transactions
--
-- accounts
Open Flags
--
-- today
High Risk
--
avg risk: --
Total Volume
--
flagged: --
Escalated
--
SAR filed: --
% Reviewed
--
-- of -- flags

📥 Compliance Workbench

Your morning inbox — all pending actions, sorted by priority and SLA urgency

Flags Today
SARs Pending
Escalations Active
Health Score
Loading workbench...
7-Day Resolution Trend
Loading...
Opened Resolved Escalations

Executive Overview

Real-time compliance KPIs — auto-refreshes every 60s

--
Health
--
Compliance Program
Health Grade
Component Scores
Loading...
7-Day Trend
Last calculated: —
📁
Total Scanned
--
transactions
🚩
Flagged Rate
--
of total
🔍
Open Investigations
--
active cases
📋
Pending SARs
--
draft filings
Overdue Deadlines
--
require action
📈
Avg Risk Score
--
portfolio avg

Risk Distribution

-- flags
Low (0–30)
--
Medium (31–60)
--
High (61–80)
--
Critical (81–100)
--

Recent Activity

--
Time Action User

📄 Board Reports

No reports generated yet. Click Generate Board Report above to create one.

📦 Exam Packages

No exam packages generated yet. Click 📦 Exam Package above to create one.

🚨 Compliance Alert Rules

Threshold-based rules that fire when compliance KPIs exceed limits.

Active Alerts

Alert Rules

Rule Name Metric Condition Severity Enabled Last Triggered Actions

Flag Trend (30 Days)

Flag Types

Recent High-Risk Flags

0 critical
AccountFlag TypeRiskDescriptionAmountDateStatus

All Compliance Flags

IDAccountFlag TypeRiskDescriptionAmountDateStatusAction

Investigation Cases

Case #TransactionAccountRiskPriorityStatusAssignedOpenedAction

Transaction History

↓ Template
IDAccountTypeAmountRisk ScoreCounterpartyDateDescription

⚡ Webhook Integrations

Webhook endpoints let external systems push transactions in real-time via HMAC-signed HTTP requests. Each endpoint has its own secret key and rate limit.

Name Status Received Processed Flagged Last Activity Actions

Upload Transaction Data

Paste CSV data with columns: account_id, transaction_type, amount, transaction_date (optional: currency, counterparty_name, description)

🚨 Flag Review Queue

Loading…

Real-time alerts from the rule-based scan engine. Review flagged transactions and mark them as resolved.

Email Alert Settings

Loading…

Clerk proactively emails you when compliance scans detect high-risk flags — no need to check the dashboard. Competitors show dashboards. You get notified automatically.

1 — All flags High Risk 100 — Critical only

How Alerts Work

🔍
Scan Runs
After each compliance scan, Clerk checks for new flags above your threshold.
Instant Alert
If high-risk flags are found, an email is sent immediately with a full summary.
📋
Full Details
Email includes account, amount, flag type, risk score, and a direct link back here.

Human-in-the-Loop Review Queue

Critical flags (risk score > 75) requiring human review before action

MCB AIAUP compliance requires human review of all high-risk alerts. Reviewer actions (confirm, dismiss, escalate) are logged to the model audit trail with timestamps and override tracking.

Model Audit Log

⚡ Kill Switch

Loading…

Immediately disables ML predictions and falls back to rule-based detection only. Required by bank AI governance policy. Activating requires a documented reason.

⚖️ Fairness Assessment

Not run

Bias analysis across geographic segments, account types, and transaction size brackets. Flags segments where false positive rate exceeds 2× baseline (disparate impact ratio). Required for MCB AIAUP vendor compliance.

No fairness report yet. Click "Run Analysis" to generate one.

📋 Model Card

Detection Engine
Rule-Based (Phase 1)
5 detection rules: structuring, large transfers, velocity, round amounts, rapid movement
AI Risk Tier
Significant / High
BSA/AML compliance affects consequential decisions and regulatory reporting
Data Governance
Minimal (Default)
No training on bank data, zero prompt/output retention, US-only processing
HITL Coverage
All Critical Flags
Risk score > 75 requires human review before action. All decisions logged.
Kill Switch
Instantly disables AI predictions, falls back to rule-based only
Fairness Testing
Disparate impact ratio across geographic, account type, and size segments

Model Evaluation Metrics

Live performance data from compliance flag reviews and system telemetry
📋 View Public Model Card
Precision
Of human-reviewed flags
False Positive Rate
Dismissed by investigators
Total Flags
All time
Human Reviewed
Investigator evaluations
Flag Rate
% of all transactions
Avg Risk Score
Across all flags

Confusion Matrix

Human-evaluated flags only
Flagged
Not Flagged
Suspicious
TP
Confirmed
FN
?
Unknown
Legitimate
FP
Dismissed
TN
Not flagged
FN = unknown (no labeled negative ground truth). Recall and F1 require labeled dataset — Phase 2 roadmap.

Alert Volume by Type

All time

Feature Importance

SHAP-equivalent attribution
Score = 60% flag frequency + 40% avg risk. Higher = stronger AML signal.

Precision & FPR Trend

Last 4 weeks
Single data point shown if <2 weeks of review history.

Risk Score Distribution

Full Model Card →
📋
Recall & F1 — Phase 2 Roadmap
Recall requires a labeled ground-truth dataset with known negatives. Our current synthetic data doesn't provide ground-truth negative labels. Phase 2 will introduce a held-out labeled evaluation set enabling full Recall, F1, and AUC-ROC reporting per SR 11-7 model validation standards.

Scan History

Daily compliance scan results — click any row to see flag breakdown by severity.

Daily Breakdown

Date Transactions Scanned Flags Avg Risk Compliance Score Breakdown

⚡ Automated Scanning

Loading…

Enable scheduled scans so Clerk monitors your transactions 24/7 — no manual clicks required. Email alerts fire automatically when critical flags are detected.

Last Scan
Next Scheduled Scan

🔗 Webhook Notifications

Send real-time compliance events to Slack, Teams, PagerDuty, or any HTTP endpoint. Payloads are HMAC-SHA256 signed for verification.

📋 Scan History

📦 Export Audit Package

Exam-ready compliance bundle

Generate an exam-ready compliance audit package for OCC/FDIC regulatory reviews. Export includes transaction summary, flag details, audit trail, and system configuration.

Model Distribution Status
Loading drift data…
Data freshness:
Stable (PSI < 0.10)
Warning (0.10–0.25)
Critical > 0.25 — review/retrain needed

PSI Feature Heatmap

Last 7 days vs. training baseline
Model Decisions (30d)
compliance flags generated
Human Override Rate
of reviewed flags
Avg Confidence
average risk score (0–100)
Last Retrain Date
baseline computed

Model Decision Volume

Average Confidence Trend

PSI History (30 days)

Your API Keys

Manage API keys for secure bank integrations. All keys start with ck_live_. The full key is only shown once at creation.

Loading API keys...

Role Management

Assign roles to API keys to control what each integration can read, write, or export.

Loading roles...

Model Decision Audit Trail

⬇ Export CSV
Complete log of all model decisions (compliance flags) with transaction context, confidence scores, SHAP-style explanations, and human review outcomes. Use CSV export for regulatory examination preparation.

Activity Audit Log

⬇ Export CSV
System-wide activity trail: who triggered scans, reviewed flags, changed settings, created API keys, generated reports, and downloaded exports. Required by OCC/FDIC examiners — answers "who did what, when, and why?"

Scheduled Report Delivery

Automatically generate and email compliance summaries on a weekly or monthly cadence. Arrives Monday morning — board-ready.

Generated Reports

Generate Compliance Report

Generate regulatory-grade reports for SAR narratives, risk assessments, and exam preparation — formatted for BSA/AML compliance teams.

Past Reports

Active Rules

Status Name Type Severity Description Actions
Loading rules...

Active Webhooks

Loading webhooks…
SAR Reference Type Status Subject / Account Activity Type Amount 30-Day Deadline Actions

📋 SAR Auto-Draft Generator

FinCEN-format pre-fill from case data — review and finalize before submission

Drafts
--
Under Review
--
Approved
--
Filed This Month
--
Overdue
--
Subject Filing Type Status Amount Deadline Actions
Loading...
Status Title Type Priority Due Date Countdown Assigned To Actions

📋 AI Governance Repository

SOC II & ISO/IEC 42001 readiness — redacted policy library

Total Documents
--
policy library
Categories
--
domain areas
Due for Review
--
next 30 days
Source Documents
--
ingested & redacted

🗂 Document Library


Loading documents…

📥 Source Document Intake Log

Source Document Ingested Redaction Integrated Into
Loading…

🏆 Certification Tracker

SOC II Type II
Gap Assessment: Planned
Service Organization Controls report covering Security, Availability, Confidentiality, and Privacy Trust Service Criteria. AI governance program documentation is a prerequisite for auditor scoping.
ISO/IEC 42001
Gap Assessment: Planned
International standard for AI Management Systems. Requires documented AI governance policies, risk assessment procedures, and continuous monitoring — all addressed by this repository.
AIUC-1
Gap Assessment: Planned
AI Use Case attestation framework covering risk triage, control adequacy, and validation requirements. Use case intake documentation feeds directly into this certification track.

🔔 Notification Preferences

Configure which alert severities trigger email notifications and for whom.

Add / Update Subscriber

Subscribers
Loading…

📨 Recent Notifications

Last 50 alert emails sent from this account.

Recipient Severity Subject Sent At
Loading…

🔮 Digital Twin Platform G2 In Review

Governance · Evaluation Harness · Champion-Challenger · AI Compliance

⚠️ All regulatory filing activities remain out of scope for FCC pilot
Loading…
Loading FCC domain…

🤖 SAR Investigation Case Build — AI-Assembled Evidence AGENT ACTIVE

Agent assembles multi-source evidence index · Investigator directs all decisions

Kill Switch:
Open
AI Assembled
Inv. Reviewing
SAR Drafting
Filed
Closed
Case Queue
Case IDCustomerSourcePriorityStatusAmountOpenedActions
Loading…

🤖 Fraud Investigation Narratives — AI-Assisted Drafting AGENT ACTIVE

Agent drafts investigation narratives from case data · Investigator retains authorship

Kill Switch:
Open
AI Drafted
Inv. Reviewed
Recovery
Closed
Fraud Case Queue
Case IDCustomerFraud TypeStatusClaim AmountReg EOpenedActions
Loading…

TM Alert L1 Review — AI-Assisted Triage

AML.2.1.1 · 3,200 alerts/mo · 22% AML Ops FTE · 95% false positive rate
● Active
Pending
AI Reviewed
In Review
Disposed FP
Disposed TP
Escalated
AI Accuracy
analyst agreement rate
False Positive Rate
of disposed alerts
Avg Disposition Time
hours (AI-assisted)
Volume (30d)
Alerts by Scenario
Alert ID Customer Scenario Score AI Assessment Status Triggered Actions
Loading…
Loading ICO domain…

9-Dimension Champion vs. Challenger

— — Champion (BAU) —— Challenger (Twin)
Dimension
Champion (BAU)
Challenger (Twin)
Δ Delta
Status
Select an agent above
Overall Score: / 100
Recommendation:

Evaluation History

Loading…

🚨 Alert Escalation Workflow

SLA-enforced escalation chains — critical/high alerts auto-escalate through defined roles

Critical Active
--
High Active
--
Pending
--
Acknowledged
--
Resolved
--
Overdue SLA
--

Active Escalations

0 escalations
Flag ID Severity Level Assigned To SLA Status Actions
Loading escalations…

⚙️ Escalation Policy Configuration

Edit SLA times per severity — changes apply to new escalations
Loading policies…

🔔 Real-Time Notifications

Email & webhook dispatch for critical compliance events — officers get immediate alerts

Total Sent
--
Delivered
--
Failed
--
Delivery Rate
--
Last 24h
--
Critical Alerts
--
📡 Notification Channels
Loading channels…
⚡ Routing Rules
Loading rules…
📋 Delivery Log
Event Channel Status Time Details
Loading…