Designed for confidential environments, the system enables visual workflow design, controlled agent coordination, and secure data exchange across enterprise operations.
Enterprises are increasingly exploring agentic AI to automate complex workflows, but adoption remains limited due to challenges in handling confidential data, maintaining workflows, and extending AI beyond digital environments into physical operations. Existing systems often struggle to securely coordinate multiple AI agents or connect digital decision making with real world actions.
To meet these challenges, Fujitsu Limited announced the development of Fujitsu Kozuchi Physical AI 1.0, a technology aimed at integrating physical AI and agentic AI for enterprise applications..The technology combines NVIDIA’s software stack with Fujitsu’s proprietary AI technologies. It introduces a multi AI agent framework designed to automate highly confidential enterprise workflows and a set of specialized AI agents based on Fujitsu’s Takane large language model to support procurement operations.
According to the company, the multi AI agent framework enables secure workflow construction using a visual interface. It combines multiple AI agents through the company’s Composite AI and supports secure data exchange using a secure inter agent gateway. The framework is designed to maintain version control and update management through integration with NVIDIA NIM microservices.
The company also developed three specialized AI agents for procurement processes. These include a document comprehension agent for structuring complex documents, a procurement regulation analysis agent to generate compliance checks and a compliance checking agent to verify regulatory requirements before sending quotation requests to external suppliers.
Key features include :
- Multi AI agent framework with visual workflow design
- Secure inter agent communication using Fujitsu secure inter agent gateway
- Specialized Takane based agents for procurement automation
- Reduced workflow complexity through coordinated agent execution
- Compatibility with NVIDIA NIM microservices
The company’s purchasing department showed a reduction of approximately 50% in order confirmation workload, with improved processing speed for internal compliance checks as they note, “By integrating agentic and physical AI, we aim to enable secure automation across complex enterprise workflows and extend AI capabilities into real world operations.”
