Re-engineering anti-fraud processes using AI & big data

Arresting fraud in the European insurance market

Fraud is a widespread problem within the European insurance market. Fraudulent motor claims are particularly serious – Italy has a nationwide average of 19.3% per year, which increases the cost of business for the entire industry. Faced with a growing fraud challenge, this leading Italian general insurer wanted a smart technical solution that would increase prevention and detection. Specific issues included:

• Inconsistent data across markets hindered fraud detection
• Static fraud detection models were slow and becoming ineffective
• As fraud became more sophisticated there was a need for a smarter solution that could learn heuristically from behaviour patterns

Implement a solution combining real-time data, rules and predictive modelling, and real-time scoring

GFT provided the insurance company with an accelerator to rapidly implement an enterprise-wide anti-fraud solution that harnessed the power of artificial intelligence and big data. The focus of the solution was online fraud prevention so it was necessary to combine and analyse structured and unstructured from internal and external sources. The anti-fraud solution implemented included:

• Real-time technology to enable instant analysis and a swift response
• Improved scoring effectiveness by using heuristic rules and predictive models to  minimise false positives and discovery of false negatives to improve effectiveness
• Extended data and search capabilities across structured and unstructured sources
• User-friendly investigation instruments that allow the antifraud team to access and query structured and unstructured documents and perform social network analysis
• Clear reporting that presents actionable information that simplifies the underlying complex data and analysis

Because of the broad scope of the project, GFT developed a modular framework comprising several specialised components, each of which performs a unique role in detecting fraud and investigating suspicious items.

In only a few months, this general insurance company improved the accuracy of suspicious claims 

• 30% increase in fraud detection and 7% decrease of false positive
• 40% increase in daily real-time reporting
• With real-time detection, the anti-fraud team can act immediately, extract relevant information and examine suspicious positions without the involvement of the IT department
• The antifraud team can insert and amend heuristic rules and predictive models manually and evaluate the impact of modifications on historical datasets, eliminating the need for expensive simulations or extractions