ShadowTrace AI

Inspiration

Modern slavery and human trafficking create major financial crime exposure, but the signals are often hidden across connected accounts, shared entities, documents, and transaction behavior. AML teams receive thousands of alerts, yet many investigations are still handled manually and in isolation. We wanted to show how UiPath can turn that fragmented process into a guided, evidence-backed case lifecycle.

What it does

ShadowTraceAI / SentinelCase is an AI-orchestrated AML case management solution. It detects suspicious alerts, maps account networks, enriches evidence with specialist agents, creates a case intelligence brief, routes the decision through Action Center, prepares SAR and regulatory response packages, and closes the case with an audit-ready trail.

How we built it

We used UiPath Maestro as the case orchestration layer. RPA handles transaction monitoring, regulatory handoff simulation, counterparty outreach simulation, and audit package building. UiPath Coded Agents handle network mapping, research enrichment, document intelligence, risk scoring, case briefing, adaptive routing, SAR drafting, SAR readiness, and information-sharing assessment. Action Center and a Coded Action App provide the compliance officer review point.

Challenges we ran into

The hardest part was connecting many moving pieces into one coherent case lifecycle: agents, RPA jobs, Maestro stages, Action Center decisions, storage artifacts, and audit outputs. We also had to keep the flow realistic for AML compliance by ensuring the system prepares evidence and recommendations without automatically making regulated filing decisions.

Accomplishments that we're proud of

We built an end-to-end investigation flow from suspicious alert to audit package. The case can move through network mapping, evidence enrichment, risk scoring, intelligence briefing, adaptive routing, human review, SAR preparation, regulatory response simulation, and closure. We are especially proud of the explainable case brief and the human-in-the-loop decision model.

What we learned

We learned how powerful Maestro becomes when it is used as the backbone for agentic case management. Agents are most useful when each has a clear responsibility and produces structured outputs that downstream stages can trust. We also learned that auditability is just as important as automation in compliance workflows.

What's next

Next, we would improve the reviewer experience, add more realistic external data integrations, strengthen document intelligence, and extend the routing model across more AML typologies. We would also add post-response review dashboards so compliance teams can monitor outcomes, bottlenecks, and investigation quality.

Built With

• aml
• json-schema
• mermaid
• powershell
• python
• react
• synthetic
• transaction
• typescript
• uipath-action-center
• uipath-agents
• uipath-coded-apps
• uipath-maestro
• uipath-rpa
• uipath-storage-buckets
• vite