[Financial Insider] "Increasing Car Insurance Enrollment and Decreasing Loss Ratio with Machine Learning, a Win-Win"

KB Sonbo, Industry's First to Apply Machine Learning in Auto Insurance Underwriting
TF Led by Gong Hyunseong, Deputy of Automobile Operations Part

Editor's NoteFinance simply means 'to provide funds.' Although it sounds straightforward, it involves complex tasks and requires the expertise and efforts of various professionals. We take a closer look at the hidden workers who support and drive large corporations.

"If traditional underwriting was about filtering out bad risks, now the goal has shifted to discovering good risks."

Recently, Gong Hyun-seong, Assistant Manager of the Auto Insurance Department at KB Insurance, emphasized this point in an interview with Asia Economy. KB Insurance, the first in the industry to introduce artificial intelligence (AI) machine learning techniques into auto insurance underwriting, is changing the concept of underwriting. Until now, underwriting served as a barrier to exclude high-risk cases. It focused on applying the practical experience of underwriters to various data to eliminate risky cases. However, with the introduction of AI machine learning, the situation has changed. The focus has shifted to identifying high-quality cases even among those previously filtered out as bad risks.

Photo by Getty Images Bank

For example, the number of accidents is an important indicator in underwriting. In the past, if a customer had four or more accidents within three years, insurance applications were often rejected, with each company setting its own standards. However, AI now allows for comprehensive consideration of factors such as driving habits, tendencies, and age to predict the probability of accidents. Assistant Manager Gong explained, "Even if there are multiple accidents within three years, if they are concentrated in the first year or if a policyholder owns multiple vehicles with one accident each, the AI can take these into account to predict accident probability. Additionally, it can assess risk by considering changes in the condition of frequently used roads, fault ratios after accidents, and accident times."

In particular, the use of machine learning techniques that maximize AI's computational power has been effective. Instead of setting rules to exclude cases such as four accidents, the AI learns from vast amounts of data to uncover correlations between various factors and traffic accident occurrences. This has allowed for more precise refinement and advancement of complex patterns that are difficult for humans to detect. This is similar to how banks discover 'thin filers'?customers with limited financial transaction history but creditworthy?using various non-financial data. Gong said, "Overseas, underwriting professionals are usually older and predict risk based on extensive experience. Machine learning compresses the time needed to accumulate such experience by leveraging its powerful learning capabilities."

This project began in May of last year. At the corporate level, KB Insurance partnered with LG CNS to develop an accident prediction model aimed at enhancing contract review. Assistant Manager Gong is the sole KB Insurance member on the task force driving this project, collaborating closely with LG CNS developers. Despite his assistant manager rank, he was chosen as the ideal candidate due to his background in applied statistics at university, hands-on underwriting experience, and previous involvement in AI model development.

Since its initial implementation in November last year, results have been emerging. KB Insurance's auto insurance loss ratio stood at the mid-70% range for the cumulative period from January to April this year. This is comparable to the mid-70% range during the same period last year when mobility was significantly reduced due to COVID-19. Typically, the auto insurance breakeven loss ratio in the industry is considered to be between 78% and 82%. Even though customers who would have been previously rejected were newly accepted, the loss ratio (the ratio of claims paid to premiums received) remained stable. By expanding the range of high-quality customers, the market was further broadened.

In the future, the project plans to expand to commercial vehicles and include analyses based on subscription channels such as offline or direct sales. Assistant Manager Gong said, "It was very daunting to enter uncharted territory, but now that we are achieving results, I have gained confidence. I will continue to realize and expand on the early lesson that underwriting is not about avoiding risk but about how much risk can be assumed."

Kong Hyun-seong, Deputy Manager of the Automobile Business Planning Team at KB Insurance, is being interviewed by Asia Economy at the KB Insurance headquarters in Gangnam-gu, Seoul on the 31st of last month.