Fujitsu Unveils Self-Evolving Multi-AI Agent Technology for Business Operations

Fujitsu Limited today announced a significant advancement in artificial intelligence with the development of a self-evolving multi-AI agent technology. This new system empowers multiple AI agents to collaborate as a cohesive team, continuously learning and adapting to the ever-changing landscape of business operations. The core innovation lies in the agents' ability to safely and autonomously learn from their daily execution results, human feedback, policy revisions, and evolving specifications. Traditional AI agents, while capable of high-level processing for defined instructions, have faced challenges in independently diagnosing the causes of operational failures and integrating these insights into subsequent tasks. This often necessitated continuous intervention from human experts to adjust prompts, refine search methodologies, update evaluation criteria, and modify operational rules to keep pace with dynamic business environments. Fujitsu's breakthrough technology overcomes these limitations by enabling AI agents to verify their operational experiences autonomously. A key feature is the agents' capacity to pinpoint the reasons behind both successes and failures, subsequently extracting actionable knowledge and operational insights. Crucially, this goes beyond mere storage of improvement proposals; the agents actively implement these learnings. This capability allows the AI system to assume responsibilities such as prompt adjustments and evaluation criteria updates, tasks previously managed by human specialists. The technology is poised to revolutionize corporate operations where legal revisions, system changes, and specification updates are constant. For instance, in tasks involving extensive documentation and design specifications, determining relevant information and prioritizing judgment criteria has historically relied on the tacit knowledge of skilled professionals. With this new technology, AI agents can learn from past search results, failure scenarios, and human corrections to autonomously enhance their search range expansion and document extraction strategies, thereby boosting accuracy and reducing expert effort. Fujitsu intends to apply these insights to its AI-Driven Software Development Platform, aiming to streamline the entire design and development process. The company envisions deploying these continuously learning AI teams not only in cloud environments but also in highly confidential on-premises and edge settings, fostering "sovereign AI" that evolves with the workplace and addresses critical societal challenges like specialized personnel shortages and knowledge succession.