en.Wedoany.com Reported - At the AWS Summit Japan 2026, Amazon Web Services (AWS) Japan's annual event held on June 26, 2026, Yoshiaki Umezu, General Manager of the AI Service Business Division at Ricoh's Corporate Planning Headquarters, delivered a speech detailing the company's initiatives to combine generative AI with physical AI to advance the assetization of tacit corporate knowledge.

Umezu pointed out that tacit knowledge accumulated within companies faces a common dilemma: intuition and know-how cultivated through experience cannot be documented and often disappear permanently when employees leave. How to archive and utilize such hard-to-transfer knowledge has become a shared challenge for many enterprises. He believes that the know-how embedded in existing internal documents, such as manuals and specifications, has not been fully tapped, and AI-ifying the know-how in these documents has been a core focus of Ricoh's efforts in recent years.

To this end, Ricoh developed its own AI platform, "Hi.DEEN (ひでん)." This platform not only reads text but also analyzes charts, drawings, internal terminology, and relationships between technologies within documents, successfully converting know-how into AI. With Ricoh's cloud service "Digital Buddy (デジタルバディ)," which is powered by Hi.DEEN, users simply input questions into a chat field, and the AI instantly searches relevant internal documents and displays answers on the screen. For example, when asking "I want to know the transaction conditions applicable under the Subcontracting Act," the AI reads a chart combining transaction content and capital-based classifications, generates an answer, and provides links to the supporting internal documents. Umezu emphasized that the platform can return answers based on chart content, enabling intuitive and comprehensive searches for specialized and internal knowledge.

Beyond document utilization, Ricoh is also advancing "operations" initiatives where business tasks are delegated to AI agents. Using the low-code tool "Dify," the company has built a mechanism for AI agents to replace fixed-process tasks, with the number of self-developed AI agents reaching approximately 9,500. For non-routine tasks that vary based on circumstances, Ricoh employs "self-improving AI agents" that autonomously learn through repeated interactions with humans. Specific examples include a "persona distillation" approach, where through multiple conversations with an individual, the AI learns that person's thought patterns and speech habits, creating a "digital clone" that simulates them. In business negotiation scenarios, AI characters sit in on conversations and record dialogues; as the number of sessions increases, the counterparty's values and decision-making criteria are automatically profiled. Umezu's own digital clone was created through internal one-on-one meetings and project consultation dialogues. He believes that preserving human judgment axes and thinking habits as memories will become a key information asset for companies in the future.