Skip to content
Book Call
Back to Projects
Banking • Active

Corridor: Exchange Houses

Two AI agents carrying the heaviest AML workload for exchange houses — scoring and grouping transaction-monitor alerts with pre-assembled evidence, then drafting STR narratives for goAML filing under the 20-day deadline.

Live system: Two-step AML automation for exchange houses. A triage layer scores and groups transaction-monitor alerts, assembling KYC, history, and counterparty evidence wit…

60–80% faster
False-Positive Review
seconds/alert
Alert Triage
< 1 day
STR Turnaround
→ 0
Deadline Misses
Free 20-min discovery + 48h ROI audit Production-grade, not a demo3 weeks to first value
✓
✓
✓
Alert triage in seconds with grouped evidence • Officer signs every STR — the human stays accoun • Borderline flagged
Why this vs status quo
Do nothing
60–80% faster lost
Manual Banking
Build in-house
6–12 months
Hire + maintain
Corridor
seconds/alert
Weeks, not months
Alert Scoring & Grouping • Evidence Pre-assembly
Save ~10×

What ships

  • Alert Scoring & Grouping — ML scores every transaction-monitor alert and groups related…
  • Evidence Pre-assembly — Pulls customer KYC, transaction history, counterparties, and…
  • One-Paragraph Rationale — Drafts a review-ready rationale per alert so the analyst clo…
  • STR Narrative Drafting — Generates the Suspicious Transaction Report narrative in the…
  • goAML Formatting — Outputs draft STRs structured for the goAML portal with the …
  • Deadline Tracking — Tracks the 20-calendar-day clock from alert to MLRO recommen…

Human signs every critical step • Audit trail included

Sample output & architecture

Corridor — Production Blueprint
Dated 18 Aug 2026 • #COR-2026-041
False-Positive Review: 60–80% fasterAlert Triage: seconds/alertSTR Turnaround: < 1 dayBanking
Agents
Triage Agent • Evidence Agent • Rationale Agent
Stack
Azure AI Foundry, Azure AI Search, Python, FastAPI
Workflow preview
1. Alert Intake2. Alert Scoring3. Alert Grouping4. Evidence Assembly
Data flow: Alert Intake → Scoring → Grouping → Evidence Assembly → Rationale → Analyst Decision → STR Drafting → MLRO Verify & Sign → goAML Filing
Azure AI Foundry + LangChain • Azure AI Foundry provides the regulated model hosting exchanv1.0 • Audit trail

Why Azure AI Foundry + LangChain wins

Azure AI Foundry provides the regulated model hosting exchange houses need, while LangChain orchestrates the triage-to-STR agent workflow with an immutable audit trail aligned to CBUAE expectations.

Alternatives: AWS Bedrock + LangGraph, Google Vertex AI + CrewAI

How Corridor works — < 1 minute to first value

Alert Intake → Alert Scoring → Alert Grouping → delivery. Full 8-step trail collapses below.

01 • < 1 minute
Alert Intake
Alerts arrive from the transaction monitor — rule-based or ML — and are deduplicated.
02 • < 5 seconds
Alert Scoring
Every alert gets an ML risk score from pattern, velocity, and corridor features.
03 • < 30 seconds
Alert Grouping
Related alerts — same customer, corridor, or counterparty — are grouped into one case.
Full 8-step audit trail
1
Alert Intake • < 1 minute
Alerts arrive from the transaction monitor — rule-based or ML — and are deduplicated.
2
Alert Scoring • < 5 seconds
Every alert gets an ML risk score from pattern, velocity, and corridor features.
3
Alert Grouping • < 30 seconds
Related alerts — same customer, corridor, or counterparty — are grouped into one case.
4
Evidence Assembly • < 1 minute
KYC file, transaction history, counterparties, and watchlist hits are pulled per case.
5
Rationale Generation • < 10 seconds
AI writes a one-paragraph rationale with the indicators matched and a recommendation.
6
Analyst Decision • seconds
Analyst closes the case as benign with rationale, or escalates it for investigation.
7
STR Drafting • < 10 minutes
Escalated cases produce a goAML-formatted STR narrative with the full case package.
8
Filing & Deadline Tracking • < 1 day
MLRO reviews, verifies, and signs the STR before goAML filing, with the 20-day clock tracked.

Which Banking use case is yours?

Tap your model — see the exact hinge we test.

Exchange House Alert Triage

Clear thousands of daily transaction-monitor alerts with grouped evidence and instant rationales.

Do nothing
60–80% faster wasted
  • • Manual Banking process
  • • No audit trail
  • • No evidence for inspector/investor
Build in-house
6–12 months • Hire team
  • • Hire + train + maintain
  • • No re-run if product changes
  • • Hard to show investor quickly
Corridor — seconds/alert
8 steps • Agents • Evidence
  • Alert Scoring & Grouping
  • Evidence Pre-assembly
  • Daily watch + re-run

Multi-agent mesh — orchestrated

Live system presentation

Corridor · Agent Mesh

ORCHESTRATORBankingTriageACTIVEEvidenceNODE 02RationaleNODE 03STR DrafterNODE 04Compliance Offi…NODE 0512345678
Step 1< 1 minute

Alert Intake

Alerts arrive from the transaction monitor — rule-based or ML — and are deduplicated.

Alert ingestionDeduplicationCorrelation keysPriority seeding

Objections, answered

If it’s not answered, you’ll hesitate. So we answer it here.

Two AI agents carrying the heaviest AML workload for exchange houses — scoring and grouping transaction-monitor alerts with pre-assembled evidence, then drafting STR narratives for goAML filing under the 20-day deadline. Delivered as 5 specialized agents + 8-step workflow. You get Alert Scoring & Grouping + Evidence Pre-assembly and a dated evidence trail.
Production in 3 weeks • Audit trail

Get Corridor live — without the Banking drag.

Two AI agents carrying the heaviest AML workload for exchange houses — scoring and grouping transaction-monitor alerts with pre-assembled ev… Implementation: Alert Integration → Triage Engine → STR Drafting → Deadline & Audit.

Questions? Talk to a human • Responses in < 24h • Human signs every critical step

Evidence + audit trail•Azure AI Foundry + LangChain•Deterministic — LLM only for rationale