"So Much Potential, Yet... AI Use in Insurance Still in Its Infancy"

A Gap Between AI's Theoretical and Actual Use in Insurance

Much That AI Can Do, but Little It Actually Does

Legal Barriers, Internal System Integration, and Human Verification Slow Down Adoption

It has been found that there is a gap between the technological potential of generative artificial intelligence (AI) in the insurance industry and its actual use at work. While a significant portion of insurance-related tasks could dramatically reduce processing time through the use of AI, actual adoption is largely limited to areas such as document drafting due to barriers such as integration with internal systems and regulatory and liability issues.

According to the financial sector on July 26, the insurance industry is regarded as a sector with high potential for improving work efficiency through AI, given the need for vast document processing and data analysis. However, despite this technological potential, actual use of AI in daily work remains limited so far.

The Korea Insurance Research Institute compared the AI utilization potential and actual use level for 104 insurance-related tasks listed on the U.S. Department of Labor's Occupational Information Network. In previous studies, AI experts assessed that of these tasks, 83 could cut processing time by more than half while maintaining the same quality if large language models (LLMs) were used independently or in combination with additional software and in-house systems. However, an analysis by Anthropic of actual Claude usage logs at a certain point last year showed that sufficient use was observed in only eight insurance-related tasks. This reveals a significant gap between tasks that could theoretically employ AI and those where it is actually being implemented.

By job category, insurance planners showed the highest level of AI utilization. This was due to the active use of AI for drafting guidance messages to customers, creating interview questions, and preparing tailored product descriptions. Rather than AI replacing direct meetings with customers or relationship-building itself, the findings indicate that AI is used primarily to prepare for and support face-to-face work.

On the other hand, actuarial professionals were rated as having the greatest potential for efficiency gains through AI, but actual usage was relatively low. This is because performing core tasks such as calculating premium rates, determining policy reserves, and analyzing financial soundness requires the use of sensitive internal data and systems, and the results must be rigorously verified for accuracy.

Constraints on AI adoption in the insurance industry include not only the reliability of the technology itself but also regulations, qualifications, and accountability issues. For example, tasks that must be performed or confirmed by a certified professional, such as loss adjusting, are difficult to delegate to AI even if they are technically automatable.

Song Yoona, a research fellow at the Korea Insurance Research Institute, commented, "Insurance-related occupations have a high proportion of tasks that could see reduced work hours through AI, but in reality, most of the application is focused on indirect exposure and limited to document work, among others. This shows that technological potential does not immediately translate into workplace adoption and productivity gains." She added, "Given the time lag observed during the initial diffusion phase of general-purpose technologies, the gap between potential and actual AI use will likely diminish over time. It is necessary to begin adopting AI for standardized, repetitive tasks and gradually build practical experience by supporting human work." She further emphasized, "In addition to reorganizing internal systems and data, it is essential to enhance employees' AI utilization capabilities and establish management frameworks to address errors and biases in AI."

This content was produced with the assistance of AI translation services.