Misumi Deploys Nine AI Agents for Product Data

Misumi Group, which sells products such as mold components, began operating an AI system in October 2026 to generate and manage product information for tens of thousands of items on its e-commerce site. Fujitsu provided the system platform and a Forward Deployed Engineer (an engineer who works on development at a customer’s site), and announced the system on October 22, 2026. The system handles information collection and preprocessing, AI-generated data and evaluation, and final review by staff as one workflow.

As Misumi expanded its product range, the amount of category and specification information registered on its e-commerce site grew. Registering new products and series required staff to check multiple systems and external services, delaying product listings. The company faced the task of improving efficiency while maintaining product information quality.

The system was built on Fujitsu Data Intelligence PaaS, which provides a data platform and an AI execution environment. Misumi incorporated business knowledge, including criteria for judging product information and review methods, into AI agents. The workflow also includes staff review and approval.

The workflow uses nine AI agents with different roles, including generating series names, suggesting category candidates, extracting specifications, checking prohibited characters, and assessing whether candidates are appropriate. The agents evaluate one another’s suggestions as well as providing candidates. Combined with staff review and approval, this process is intended to limit inconsistent wording and categorization.

Misumi plans to expand the use of its data model, AI processing platform, and workflow in stages. It will first apply them to product information management, from importing product data received from suppliers through managing the product master, and will also pursue links with surrounding systems and use in other operations. In the future, the company aims to automate the entire business process, including staff’s final review and approval. The summary does not report the system’s processing accuracy at launch or how much it shortened listing times. The planned expansion includes supplier data imports, product master management, and integration with surrounding systems.

Misumi Group, which sells products such as mold components, began operating an AI system in October 2026 to generate and manage product information for tens of thousands of items on its e-commerce site. Fujitsu provided the system platform and a Forward Deployed Engineer (an engineer who works on development at a customer’s site), and announced the system on October 22, 2026. The system handles information collection and preprocessing, AI-generated data and evaluation, and final review by staff as one workflow.

As Misumi expanded its product range, the amount of category and specification information registered on its e-commerce site grew. Registering new products and series required staff to check multiple systems and external services, delaying product listings. The company faced the task of improving efficiency while maintaining product information quality.

The system was built on Fujitsu Data Intelligence PaaS, which provides a data platform and an AI execution environment. Misumi incorporated business knowledge, including criteria for judging product information and review methods, into AI agents. The workflow also includes staff review and approval.

The workflow uses nine AI agents with different roles, including generating series names, suggesting category candidates, extracting specifications, checking prohibited characters, and assessing whether candidates are appropriate. The agents evaluate one another’s suggestions as well as providing candidates. Combined with staff review and approval, this process is intended to limit inconsistent wording and categorization.

Misumi plans to expand the use of its data model, AI processing platform, and workflow in stages. It will first apply them to product information management, from importing product data received from suppliers through managing the product master, and will also pursue links with surrounding systems and use in other operations. In the future, the company aims to automate the entire business process, including staff’s final review and approval. The summary does not report the system’s processing accuracy at launch or how much it shortened listing times. The planned expansion includes supplier data imports, product master management, and integration with surrounding systems.