A leading insurance company achieves 40 times improvement in detection of fraudulent claims leveraging Wipro’s Artificial Intelligence powerhouse - HOLMES

Client Background

• Client: A leading insurance company
• Industry: Financial services and insurance
• Products or Services: Banking and Insurance  
• Areas of operation: Australasia
• Number of Employees: Over 14,000
  

Challenge

The client was under pressure to improve its profitability by reducing rising incidence of fraudulent claims. Detection of new fraud types was difficult in the insurer’s current fraud detection system and fraudsters could easily circumvent getting caught. High number of false positives resulted in significant investigation effort and delayed payments to genuine customers. The insurer was looking at reducing the cost of risks to improve operations performance and revenue.

Solution

Wipro deployed its Artificial Intelligence platform HOLMES to improve detection of suspicious claims. Wipro brought to the client deep domain expertise and technology competency along with machine learning capabilities of HOLMES.

With its self-learning algorithms and capabilities to understand patterns, HOLMES augmented the available rule-based models and identified new fraud schemes that had not been encountered before.

• HOLMES machine learning model studied the identified claims along with other data on policy, parties involved, insured assets, etc.
• Separate models were built for each fraud type. Using feedback, these models self-calibrated over time to adapt to new data and changes in user behavior
• Prediction capability in HOLMES provided an indicator of why a case was flagged; such as suspicious witness, incident occurrence on a weekend, etc.

Business Impact

Wipro HOLMES^TM enabled detection of fraudulent claims with high accuracy, helping investigating units to target the right cases and reduce false positives. The client achieved 40 times improvement in prediction of fraudulent claims across categories such as staged accident, staged theft and fraud at inception.

• 40 times improvement in detection of fraudulent claims
• Over 70% accuracy in prediction of fraud claims
• Negligible misses with over 98% detection of frauds
• Comprehensive coverage – all claims verified before payouts
• Targeted investigations only for flagged cases