Case Study: Global Bank cuts check fraud by 50% and forecasts $20M annual savings with Cognizant

Advanced AI/ML Solution Detects Check Fraud for a Global Bank

Cognizant Digital Business partnered with a leading global bank that still processes millions of handwritten checks each month and was incurring significant losses to counterfeiters. Because handwritten checks require one-by-one human verification, the bank faced high manual workloads, slow turnarounds and ongoing fraud risk.

Cognizant built a TensorFlow-based neural network that analyzes scanned check images—including payee, amounts, account/routing numbers and signatures—against a growing database of historical and known-fraud checks, producing a confidence score in under 70 milliseconds. The self-learning solution flags anomalies for review, scales to millions of checks per day (up to 1,200/sec), demonstrated a 50% reduction in fraud in tests and is forecast to cut fraudulent losses by about $20 million annually while reducing manual effort and costs.