Japan's insurance industry is venturing into uncharted territories, embracing cutting-edge technologies to reshape its landscapes. A prime example of this innovation surge is Tokio Marine & Nichido Fire Insurance's recent efforts in leveraging generative AI, specifically utilizing GPT-4o, to streamline its system development processes. This initiative has not only demonstrated significant reductions in programming labor but also highlights broader market opportunities within Japan's insurance sector that remain largely unexplored by foreign enterprises.
Generative AI: Transforming System Development
In a move to optimize their core system development, Tokio Marine began experimenting with generative AI to automate code generation. This innovative approach resulted in a remarkable 44% reduction in programming effort for new application development and system specification modifications. As a testament to AI's growing influence, the company plans to integrate these AI-powered solutions into their operations starting October 2024.
Yamashita Hiroki, attached to the IT Infrastructure Services division, emphasized that this initiative allows personnel to focus on strategic tasks that require human insight, thus optimizing resource allocation. By choosing generative AI for programming tasks—areas where AI aligns naturally—the company positions itself at the forefront of digital transformation within the insurance industry.
Collaborative Innovation: A Japan-IBM Partnership
Beginning in September 2023, Tokio Marine, in collaboration with IBM Japan, embarked on a pilot project using AI for code generation. The focus was on enhancing the claims investigation system via the Guidewire ClaimCenter platform, utilizing the Gosu programming language—a Java-like language that's relatively niche but integral to the project.
The project involved translating design documents into code through AI, using prompts and translation processes to navigate challenges with language precision. To achieve high-fidelity code generation, they adapted prompting techniques such as "Few-shot Prompting," refining the instructional inputs to the AI to ensure accuracy.
One noteworthy issue encountered was the omission of null checks in AI-generated code. The solution lay in a verification step, checking and correcting code for such omissions. The adaptability of AI models remains crucial; as new AI models emerge rapidly, maintaining the flexibility to switch frameworks will ensure continued efficiency and relevance.
The Path Ahead: Scaling and Training for Future Development
Looking ahead to October 2024 and beyond, Tokio Marine aims for a large-scale rollout with an increase in input prompts, necessitating strategies to manage this growth. Steps like segmenting design documents for input are among the strategies being considered.
The vision extends beyond deployment—training internal AI specialists to fine-tune prompts and manage the AI systems will be vital, as will expanding AI's role across other areas such as document creation and test automation. This commitment to AI-driven development underscores the vast opportunities awaiting other enterprises willing to explore Japan's dynamic insurance market.
Are you ready to explore and capitalize on the burgeoning potential within Japan's insurance sector? Our startup specializes in guiding foreign companies through entry and growth in the Japanese market. Discover how our tailored services can help you navigate and thrive in this evolving industry landscape. Learn More About Our Market Entry Solutions
Comments