Organizations continue to grapple with fragmented systems, increasing operational complexity, and talent shortages. Although generative AI has fostered optimism, making progress in content creation, coding, and analysis, systems still lack the ability to act with foresight, autonomy, and coordination in multiple scenarios.
In this context, Fujitsu has just developed a collaboration technology between multiple AI agents to optimize supply chains. The idea is to be able to securely connect AI agents from multiple companies to respond quickly to changes in the environment and get them to act with greater coordination and sense.
Fujitsu plans to begin field trials in January 2026 to optimize the supply chain of Rohto Pharmaceutical Co., Ltd., in collaboration with the Tokyo Institute of Science (Science Tokyo).
This innovation not only streamlines daily supply chain operations, but also facilitates rapid recovery in emergency situations, such as sudden changes in demand or disasters that affect multiple points in the logistics network.
Fujitsu will continue to advance in demonstrations and technological improvement together with Science Tokyo and Rohto Pharmaceutical, with the aim of expanding this solution to various sectors, including manufacturing, where multiple simultaneous and highly sensitive processes are managed.
Likewise, it will develop technology for broader and more complex supply chains, with the goal of offering it, before the end of fiscal year 2026, through the Dynamic Supply Chain services of its Uvance business model. This will bring a new perspective to corporate supply chain strategies, strengthening resilience and enabling more sustainable business operations across multiple industries.
Under its Uvance business model, Fujitsu will leverage the technology developed in these trials to make secure data collaboration a reality by cooperating AI agents across borders and industries.
This will drive more resilient supply chains and sustainable industrial growth, ensuring reliability and governance in multi-supplier environments.
Katsuki Fujisawa, professor at the Digital Twin Research Unit of the Integrated Research Institute and the Department of Mathematical and Computational Sciences at Science Tokyo School of Computing, said: “Science Tokyo is actively promoting research in Cyber-Physical Systems (CPS) and working to improve the efficiency of the entire industrial value chain. Going forward, by collaborating with Fujitsu’s agentic AI technology to optimize the entire supply chain, we aim to contribute to the advancement of the industry and the resolution of societal challenges that span multiple sectors.
Collaboration technology
The new technology is based on the use of two key components:
• Global optimal control for multiple AI agents in environments with incomplete information
It allows effective collaboration between AI agents from different companies without the need to reveal sensitive data, something that is often essential in the coordination of AI agents between companies.
• Fujitsu secure gateway for agent interconnection
Designed based on AI distributed learning technologies and communication “guardrails” between AI agents, the gateway allows fluid and secure collaboration between agents from different companies and suppliers, while protecting confidential and private corporate information.
During the configuration phase, the technology allows AI agents to learn the characteristics of the supply chain and optimize their operations without directly sharing sensitive data, through the use of knowledge distillation, a deep learning technique that transfers knowledge from a “teacher model” to a “student model.”
In operation, and based on Fujitsu’s experience in language model guardrails (LLM), this technology detects malicious queries and prevents the inference of sensitive information.
During the configuration phase, the technology allows AI agents to learn the characteristics of the supply chain and optimize its operations
Additionally, it secures communications by repeatedly simulating the behavior and responses of AI agents, updating and providing the information in a secure, non-inferable format.
Field tests
By combining Science Tokyo’s AI agent technology with its own AI multi-agent collaboration technology, Fujitsu, in collaboration with Rohto Pharmaceutical, conducted initial tests in a virtual supply chain to optimize logistics routes and schedules, confirming a potential reduction of up to 30% in transportation costs.
Between January 2026 and March 2027, Fujitsu will conduct more practical, larger-scale tests, simulating real-world conditions based on Rohto Pharmaceutical’s supply chain.
