According to the FBI, the total cost of insurance fraud (non-health insurance) is estimated to be more than $40 billion per year, which costs average U.S. family between $400 and $700 per year in increased premiums. Insurance claims processing is a tedious work which leads to errors. Some fraud types are well known and can be spotted using rules-based systems. However, new or more nuanced fraud is missed by rules-based systems until that fraud becomes well documented.

AI is ideally suited to fraud detection for insurance claims. Machine learning models can be used to automate claims assessment and routing based on existing fraud patterns. This process flags potentially fraudulent claims for further review, but also has the added benefit of automatically identifying good transactions and streamlining their approval and payment. More advanced anomaly detection systems can be deployed to find new patterns and to flag those for review, which leads to prompt investigation of new fraud types. AI systems can also provide clear reason codes for investigators, so they can quickly see the key factors that led the AI to indicate fraud which streamlines their investigation. With AI based fraud detection, fraudulent claims can be evaluated and flagged before they are paid, which reduces costs for insurance providers and helps reduce costs for consumers.

The mission at H2O.ai is to democratize AI for all so that more people across industries can use the power of AI to solve business and social challenges. The insurance industry is a key focus for the company with leading insurance companies including Progressive, TransAmerica, Zurich Insurance helping to drive significant product innovation. H2O Driverless AI is an award-winning platform for automatic machine learning that empowers data science teams to scale by dramatically increasing the speed to develop highly accurate predictive models. Driverless AI includes innovative features of particular interest to insurance companies including machine learning interpretability (MLI), reason codes for individual predictions, and automatic time series modeling.