December 7, 2015
AI’s Huge Potential for Underwriting
Artificial intelligence (AI) has the potential to transform underwriting and risk management, using a series of simple smartphone apps.
For decades, the insurance industry has led the world in predictive analysis and risk assessment. And today, with the treasure trove of big data available from historical processes, IoT and social media, insurance companies have the opportunity to take this discipline to a whole new level of accuracy, consistency and customer experience.
The actuarial models that were once driven solely by large databases can now be fueled with tremendous quantities of unstructured data from social media, online research and news, weather and traffic reports, real-time securities feeds and other valuable information sources as well as by “tribal knowledge” such as internal reports, policies and regulations, presentations, emails, memos and evaluations. In fact, it is estimated that 90% of global data has been created in the past two years, and 80% of that data is unstructured.
A large portion of this data now comes from the Internet of Things — computers, smart phones and wearables, GPS-enabled devices, transportation telematics, sensors, energy controls and medical devices. Even with the advancement of big data analytics, the integration of all this structured and unstructured data would appear to be a monumental achievement with traditional database management tools. Even if we could somehow blend this data, would we then need thousands of canned reports, or a highly trained data analytics expert in every operating department to make use of it? The answer to this dilemma may be as close as our smartphones.
Apps that Unleash the Power
As consumers, we are no stranger to the union of the structured and unstructured datasets. A commuter, for example, used to rely on Google Maps to get from his office to his home. But with the advent of apps like Waze, not only can he get directions and arrival times based on mileage and speed data, but can also combine this intelligence with feeds from social media and crowd-sourced opinions on traffic. Significant advances in the power of in-memory processing, machine learning, artificial intelligence and natural language processing have the potential to blend millions of data points from operational systems, tribal knowledge and the Internet of Things — using apps no more complicated than Google Maps.
Using apps that harness the power of artificial intelligence and machine learning can provide far superior predictive analysis simply by typing in a question, such as: What are the chances of a terrorist act in Omaha during the month of December? Where is the most likely place a power blackout will occur in August? How many passenger train accidents will occur in the Northeast corridor over the next six months? What will be the effect on my fixed income portfolio if the Federal Reserve raises short term interest rates by .25 percentage point?
Using a gamified interface, these apps can use game theory such as Monte Carlo simulations simply by moving and overlaying graphical objects on your computer screen or tablet. As an example, you could calculate the likely dollar damages to policyholders caused by an impending hurricane simply by moving symbols for wind, rain and time duration over a map image. Here are some typical applications for AI app technology in insurance:
Catastrophe Risk and Damage Analysis
Incorporate historical weather patterns, news, research reports and social media into calculations of risk from potential catastrophes to price coverage or determine prudent levels of reinsurance.
Targeted Risk Analysis (Single view of customers)
With the wealth of individual information available on people and organizations, it is now possible to apply AI and machine learning principles to provide risk profiles targeted down to an individual. For example, a Facebook profile of a mountain climbing enthusiast would indicate a propensity for risk taking that might warrant a different profile than a golfer. Machine learning agents can now parse through LinkedIn profiles, Facebook posts, tweets and blogs to provide the underwriter with a targeted set of metrics to accurately assess the risk index of an individual.
Each individual assessor has his own predilection to assessing risks. By some estimates, insurance companies could lose hundreds of millions of dollars either through inaccurate risk profiling or through lost customers because of overpricing. AI apps provide the mechanics to capture “tribal knowledge,” thereby providing a uniform assessment metric across the entire underwriting process.
By unifying unstructured data across historical claims, it is possible to establish ground rules (or quantitative metrics) across fuzzy baselines that were previously not possible. Claims notes from customer service representatives that would previously fall through the cracks are now caught, processed and flagged for better claims expediting and improved customer satisfaction. By incorporating personnel records when a major casualty event occurs, such as a severe storm or flood, you can now dispatch the most experienced claims personnel to areas with the highest-value property.
Integrate social media into the claims review process. For example, it would be very suspect if someone who just put in a workers’ compensation claim for a severe back injury was bragging about his performance at his weekend rugby match on Facebook.
A Powerful Value Proposition
The value proposition of artificial intelligence apps for better insurance industry underwriting and risk management is too big to ignore. Apps have been transformational in the way we intelligently manage our lives, and App Orchid predicts they will be just as transformational in the way insurance companies manage their operations.