Chemicals and materials / Diversified chemicals
Resonac: Standardizing AI agents built on the front line and rolling them out company-wide
- Company
- Resonac
- Country
- Japan
- Adoption stage
- In operation
- Source published
- In use from
- Date basis
- The date the source was published. It can differ from the date adoption started.
- How the source was checked
- Read the full source text
The work problem
According to a Resonac Holdings press release, Resonac aimed to go beyond increasing the use of generative AI and to change how work is done across the company with AI agents. The company singled out work that tends to depend on particular employees, such as safety activities and passing on specialist knowledge, as areas where the whole work process needed to be reviewed.
Technology and data
According to the press release, Resonac introduced generative AI in October 2025. The company repeats a cycle in which headquarters provides generative AI, employees at each site and department build AI agents themselves to solve their own work problems, and headquarters improves the quality of the most effective and widely applicable agents, standardizes them and rolls them out company-wide. Headquarters departments deploy AI agents for scheduling meetings, writing text and documents, communication, and learning and analysis to support employees' daily work and decisions. In hazard prediction (KY) activities, generative AI scores the safety of work sites and automatically generates feedback for improvement, and in chemical regulation work, the creation, delivery and scoring of training materials and tests have been automated. AI agents are also used for equipment installation, inventory management and research and development support.
Results
According to the press release, about half a year after the introduction, the generative AI utilization rate reached 95%, and use took hold widely across manufacturing, R&D and indirect departments. As a result, the company cut work equivalent to about 90,000 hours a month, worth about 470 million yen, and said it calculated these figures from the survey results of 2,405 employees and actual work records. The time needed for the hazard prediction work process fell to about 1/60, and training materials for chemical regulations can now be produced in about 1/5 of the previous time.
Limits and open questions
The press release does not disclose the AI models used, the number of agents built or how people review the outputs. The time and money saved are figures the company calculated from a survey and work records.
Sources
- 現場発のAIエージェントを標準化・全社展開する独自モデルを構築prtimes.jp, Accessed
Compiled from public sources. These are not results from ATF Works customers.