AI-powered Fraud Detection Solution to Improve Supply Chain Transaction Performance

About the Customer

The customer is a leading Pharma company based in New York, with a valuation of over $25 Billion and a headcount of over 20,000.

Business Challenge

The organization had been investing significant time addressing business disruptions caused by fraudulent vendor transactions. Fraudulent transactions impact the supply chain in the following ways:

• Overall increased costs
• Decreased transparency
• Reduced efficiency

Brillio’s Solution:

Brillio implemented an AI powered – Real Time Vendor Transaction Anomaly Detection solution that is able to identify potential fraudulent vendor transactions and profiles in near real-time, thereby enabling rapid intervention. This solution also quantified risks associated with adding new vendors, which tend to bring more risk exposure.

The four pillars that form the crux of this AI Solution are as follows: –

1. Audit Execution
   • Build Clustering models to identify potential anomalous transaction clusters.

   • Apply statistical techniques to these identified clusters to derive key indicators of fraud.

   • Leverage these key fraud indicators to produce the final fraud risk score.
2. Continuous Tracking
   • Continuous vendor tracking to provide early warning signals about new vendor activities.
3. Onboarding
   • Assign a risk score to new vendors based on statistical similarity with existing ones.
4. Automated Model Refresh and Maintenance
   • Periodical refresh of the Machine Learning (ML) models with human-feedback loops.

Business Benefits:

The AI-powered fraud detection solution enabled:

• Real-time monitorization and interpretation of rapidly changing vendor transactions

• The ability to immediately respond to potential breaches that might have an adverse effect on the system

• Intelligent vendor onboarding that helps to improve vendor quality through an ML-generated Fraud Risk Score

• Increased operational efficiencies by reducing the effort needed to identify and analyze potential fraudulent transactions by 50x