AI Commissions Fraud Detection Pilot

Designing an AI-driven fraud detection framework for enterprise healthcare

The Challenge

A national health insurer processes tens of millions in monthly commission payments, yet less than 1% of transactions were being monitored for fraud.
Manual audits and threshold-based approvals left major blind spots for bad actors taking advantage of collusion, rerouted payments, and rate manipulation.

The Goal

Design and scope an AI pilot to monitor 99% of transactions in real time and improve fraud detection by 15–20%.

The Approach

I led the design strategy and pilot scoping, connecting design, data science, and compliance to define how AI could safely scale inside a regulated enterprise.

My role included:

• Stakeholder discovery and workflow mapping

• Defining fraud detection logic and model inputs

• System architecture mapping and AI flow design

• Dashboard concepts for key stakeholders

• Pilot presentation and business case validation

System Design

To align teams, I created a visual systems map showing how data would move from ingestion to live monitoring.
It outlined how historical data, synthetic testing, and continuous feedback could train and refine a fraud-detection AI, illustrating the logic in a way that business and technical teams could both understand.

Business Case

The pilot presentation framed the financial risk and opportunity:

• $480M annual exposure

• 37K weekly transactions

• <1% currently monitored

The model demonstrated how improving monitoring to 99% could dramatically reduce undetected losses while integrating with HIPAA-compliant infrastructure.