Home

Fujitsu develops multi-AI agent collaboration technology to optimize supply chains, launches joint trials

Securely connecting AI agents across multiple companies for rapid response to changing circumstances

Challenges in Inter-Company AI Agent Collaboration

It is difficult to achieve an optimal overall state when the information each company can share is limited.
Each company independently builds AI agents, making it uncertain whether their AI agents can collaborate securely.
It is challenging to evaluate the impact across companies in advance.

By using our global optimal control for AI agents under incomplete information, the traits of multiple parties can be estimated from limited information, achieving an optimal overall state.
Fujitsu secure inter-agent gateway prevents the leakage of confidential information during AI model training and during dialogues between AI agents.
By employing supply chain digital rehearsal technology, responses to various environmental changes surrounding companies are verified in advance.

Collaborating only with the information each company can share enables flexible and rapid response as a whole.	For example, in supply chain management (SCM), the exchange of information between companies not only improves operational efficiency during normal times but also allows for rapid recovery during emergencies such as sudden demand changes, accidents, or disasters.
Establishing and operating secure and safe AI agent collaboration across companies.	AI agents belonging to different companies and developed by different vendors collaborate seamlessly and securely while protecting each company's confidential and privacy information.
Proposing the best practices through rehearsing future scenarios.	For resilient supply chains, the management and DX departments can broadly anticipate future events likely to occur in the medium to long term and can direct strategic orientation.

Difficulty in coordinating the entire supply chain to swiftly respond to disasters or sudden increases in demand

When AI agents of different companies collaborating on procurement, manufacturing, and delivery, the following issues exist:
- Information obtained from other companies' AI agents is limited, making it difficult to coordinate for overall optimization across multiple companies.
- Secure data management across AI agents spanning different companies is challenging.
Additionally, the occurrence scenarios are diverse, including climate change, price increases, and manpower shortages, making pre-validation difficult.

Our multi-AI agent collaboration technology consists of the following two core technologies to address the issue of AI agents between companies not being able to collaborate due to concerns of information sharing:
- Global optimal control for AI agents under incomplete information:
  - Proposal-side estimates characteristics of multiple counterpart AI agents from the negotiation interactions of suggestion and answer between AI agents.
  - Based on the estimated characteristics, the proposal-side AI agent identifies the most optimal state rapidly.
- Fujitsu secure inter-agent gateway:
  - Distributed AI learning: Improves AI performance across the industry by learning between specialized AI models of companies without sharing confidential information.
  - AI agent communication guardrail technology: Provides secure AI agent communication by detecting sophisticated and malicious requests, and processing information so confidential data cannot be inferred from conversation history.
Furthermore, for scenarios involving climate change, price increases, manpower shortages, etc., understand the impact on your company through future rehearsal and optimize the supply chain structure.
- Supply Chain Digital Rehearsal technology provides the following functions:
  - Evaluates the impact on the entire supply chain structure.
  - Proposes the best measures.

Global optimal control for AI agents under incomplete information:
- Capable of rapidly identifying the optimal state using limited information shared by each AI agent.
Fujitsu secure inter-agent gateway:
- From the development phase to the operational phase of cross-company AI agent collaboration, we can provide a one-stop solution that ensures secure AI agent communication by protecting confidential information, enhancing AI agent performance, and leveraging the strengths of our LLM guardrails.
Supply Chain Digital Rehearsal technology:
- When various future events occur, extracts causal chains including intermediate factors to propose appropriate improvement measures across multiple perspectives such as cost, delivery dates, inventory levels, etc.

End Users:
- Collaborate through AI agents with different companies such as suppliers and delivery agents within the supply chain. Achieve cost reduction through efficient operation during normal times and rapid recovery during emergencies such as sudden demand increases or accidents/disasters.
- At the management level, engage in pre-validation with Supply Chain Digital Rehearsal for responding to environmental changes surrounding the company's supply chain including risks of climate change, geopolitical risks, logistics disruptions, material price surges, and changes in market demographics due to urban development.

About the field trials
- Combining AI agent technology developed by the Institute of Science Tokyo with our multi-AI agent collaboration technology, Fujitsu, in collaboration with Rohto Pharmaceutical Co., Ltd., conducted initial field trials on a virtual supply chain to optimize logistics routes and schedules, confirming a potential reduction of up to 30% in transportation costs.
- From January 2026 to March 2027, Fujitsu will conduct more practical and large-scale trials, simulating real-world conditions using Rohto Pharmaceutical's supply chain.

Demo Video
- Multi-AI Agent Collaboration (In Preparation)
- Supply Chain Digital Rehearsal (1.Robustizing Supply Chains、2.Risk scenario analysis、3.Scenario forecasting)