Compliance Case Management System Modernization

Context

eBay’s Compliance Case Management System (CCMS) handled sanctions, KYC/AML, and fraud workflows for a global retail marketplace — investigators and auditors routing and resolving cases across Risk, Compliance, and Regulatory teams. The existing system was slow to route cases, hard to extend, and gave compliance leadership no unified view into performance across the platform.

Approach & Key Decisions

I owned the roadmap for an AI-powered modernization of CCMS, built on Salesforce with a deliberate set of scope calls:

• Automate routing, not judgment. Machine learning handled case triage and anomaly/risk scoring; investigators kept final decision authority — a deliberate line to preserve auditability in a regulated environment.
• Integrate rather than replace. CRM and middleware integrations connected Salesforce to retail payments systems, fraud engines, and the Genesys contact center, so omnichannel support worked at high-volume e-commerce scale without a rip-and-replace of adjacent systems.
• Design for the investigator, not just the metric. I mapped case-review journeys for investigators and auditors directly, using that research to simplify review steps rather than only optimizing backend throughput.

Execution & Collaboration

Delivery spanned engineering, risk, KYC/AML, compliance, and regulatory stakeholders. I directed the implementation of ML-based anomaly detection models for fraud and risk scoring, and partnered with data teams to launch a unified fraud-risk analytics dashboard (Power BI + Databricks) giving compliance leadership executive-level visibility into performance for the first time.

Results

• 40% faster case routing and a 15% reduction in handling time through the automated routing protocol.
• 35% reduction in fraud losses, with a 28% improvement in risk-detection accuracy from the ML anomaly-detection models.
• 30% reduction in case review time for investigators and auditors, from the redesigned case-review journey.

Reflection

Automating routing without automating judgment turned out to be the right line to hold — it made the system easier to get buy-in for from compliance and legal stakeholders who needed to trust the audit trail, and it kept investigators as the accountable decision-makers the regulatory model requires. That’s the kind of call that only works if you own the outcome, not just the delivery date.