Japan Data Science Consortium Co., Ltd. (Headquarters: Tokyo, Japan; CEO: Satoshi Erdos Kato, “JDSC”) has developed a solution to promote DX and work style innovation in the industry by using LLM to reduce the man-hours required to respond to inquiries specific to the maritime industry. This solution utilizes Claude 3 Opus / Sonnet /Haiku (Anthropic Claude) on Amazon Bedrock, an AWS generated AI service. The system can research and respond to approximately 10,000 specialized documents, including various related contracts, technical information, regulatory information, FAQs, and emails, reducing response time by approximately 97% of the time previously required by employees.

Japan is a resource-poor country surrounded by sea on all sides, making marine transportation crucial for supporting both the lives of its people and economic activities. However, recent years have seen various issues highlighted in international competition, such as delays in digitalization and growing demands for enhanced environmental performance, including greenhouse gas (GHG) reduction. In addition, alongside the 2024 logistics challenges under discussion, workforce decline and aging, as well as business continuity, emerge as significant concerns across the entire logistics cycle, including the maritime sector.

Aiming to upgrade the maritime industry, JDSC is promoting cross-industry DX support, including the launch of a joint venture, SEAWISE Co. As part of this effort, JDSC has been working with a general service company, including intermediaries in various maritime-related fields, since September 2023 to understand and analyze the volume and content of inquiries for each business scenario, collect and organize the data necessary for responses, and consider how to provide the data to customers, in order to improve the accuracy of AI to enable LLMs to respond accurately and quickly to inquiries. In addition, we have been working to address industry-specific description rules.

This system handles a wide range of ship operations, including inquiries about cargo loading, port-specific regulations, and cargo handling, crucial in ship operation management. By utilizing LLM for investigation and response, tasks that previously took up to one hour can now be completed in 1 to 2 minutes. Furthermore, it enables junior employees in their third year to handle highly specialized inquiries that previously required over 15 years of experience. It aims to facilitate of work style reforms, such as the workload equalization, enhanced of human resource utilization, succession planning for work methods, and knowledge transfer.

This solution leverages Amazon Bedrock to achieve low cost, speed enhancement, and improved accuracy. Haiku combines speed, balance, and accuracy, resulting in an approximate 30% improvement in response accuracy.

JDSC will continue to expand the use of solutions based on this technical achievement and will also continue to contribute to the upgrading of Japan by implementing cutting-edge technologies such as AI, machine learning, and mathematical optimization based on its accumulated knowledge of data science. leveraging its extensive data science expertise to address challenges not only for individual companies but across the entire industry.

■About Japan Data Science Consortium Co., Ltd.

JDSC delivers advanced AI solutions with high added value across various sectors, focusing on key industries like logistics optimization, demand forecasting, fraud detection, and education. JDSC develops and licenses algorithmic modules, develops and manages IT systems, and offers business management services in data science. Its goal is to upgrade Japanese industries by leveraging AI to address industry-wide challenges.

Three key features of the JDSC:

1. Profound expertise in AI algorithms.
2. Strong executive capability to support business operations, from AI solution demonstration to implementation.
3. A highly efficient business model integrating collaborative development with major corporations and horizontal industry expansion.