Stockmark Selected for GENIAC Phase 4 and Launches AI-Ready Project for Japanese Companies’ Tacit Knowledge and Internal Data

Stockmark Inc., a provider of domestically developed generative AI infrastructure and generative AI services, announced that it has been selected for the “Research and Development on Making Manufacturing Data and Other Data AI-Ready” program under Phase 4 of “GENIAC,” a project run by Japan’s Ministry of Economy, Trade and Industry and NEDO to strengthen domestic generative AI development capabilities. Following the selection, Stockmark will collaborate with 16 leading Japanese companies to launch the “AI-Ready Project for Japanese Companies’ Tacit Knowledge and Internal Data.” The project will define specific use cases with each participating company, promote AI readiness at each organization, and ultimately publish the AI-ready know-how obtained through the project to accelerate the social implementation of generative AI across Japanese business. The background is that, in business use of generative AI, general-purpose AI trained mainly on public internet data is considered insufficient in terms of expertise and reliability. The next phase urgently requires the use of non-public data held within companies. However, in core Japanese industries such as manufacturing, much information is managed as unstructured data such as drawings and manuals, while expert know-how often remains tacit and undocumented. Completing AI utilization internally requires advanced expertise and substantial labor. Stockmark will therefore work with 16 top-running Japanese companies to conduct proof-of-concept projects that convert confidential data and expert knowledge into formats that AI can learn from and use, a process referred to as making data AI-ready. Stockmark has supported many companies through generative AI solutions including its proprietary domestic generative AI infrastructure, the manufacturing AI agent “Aconnect,” and the autonomous AI operations platform “SAT Agent Cockpit.” In this project, Stockmark will apply the knowledge gained through these efforts and collaborate with major Japanese companies to establish AI-ready know-how in specific domains such as manufacturing, where general-purpose AI struggles. By making large volumes of internal non-public data usable, the project aims to overcome the limits of AI learning that depends on public data, raise Japanese companies’ ability to use generative AI, and promote real-world deployment. In this project, AI readiness is defined as transforming non-public data into a state in which AI can learn from and use it. Based on METI guidelines, the project will develop three areas: data structuring, which converts drawings, forms, and tacit knowledge into logical structures that AI can semantically understand; quality and governance, which builds a management foundation that balances confidentiality protection with accuracy verification; and learning cycles, which implement mechanisms for continuously updating AI based on frontline feedback. The project will focus on two key themes. The first is making industry-specific data AI-ready and creating Agent-Ready manuals through Skills. Industry-specific drawings, forms, manuals, chat histories, and business procedure documents will be converted into Skills that AI agents can understand, while best practices for structuring data by industry will be established. Traditionally, improving AI accuracy required dedicated engineers. With this method, frontline staff can continuously improve AI accuracy simply by updating Skills in natural language, offering a practical path for AI adoption at many Japanese companies that cannot assign dedicated AI operations personnel. The second theme is building industry- and business-specific data platforms and AI agents. Japan’s manufacturing sector holds world-class experimental data, formulation recipes, design drawings, and business-specific data across industries, but much of it has remained unused due to confidentiality. The project will implement federated learning and synthetic data technologies that allow companies to use confidential data without moving it outside the organization, while building industry- and business-specific data platforms and highly specialized AI agents. According to the announcement, this is Japan’s first full-scale demonstration of growing industry- and business-specific AI assets while fully maintaining data sovereignty. In addition to these two themes, the project will also advance other topics necessary for each company’s AI readiness, including verification of guardrail functions such as permission management and masking, and secure processing infrastructure in private cloud environments. The 16 participating companies are Ajinomoto, ITOCHU Corporation, NGK, Kobe Steel, JTEKT, Suzuki, Sumitomo Chemical, TAIYO YUDEN, Teijin, Tokyo Electric Power Company Holdings, JGC Holdings, Sumitomo Mitsui Banking Corporation, Mitsubishi Chemical, Yanmar Holdings, Lion, and LIXIL. By combining the world-class technologies, domain expertise, and vast non-public data accumulated by these companies over many years, the project aims to go beyond individual corporate digital transformation and build a powerful cross-industry AI ecosystem that strengthens Japan’s industrial competitiveness and drives the social implementation of generative AI. Stockmark CEO Satoru Hayashi said the company is honored to be selected for GENIAC Phase 4. In Japanese manufacturing and other business fields, the rapid decline in the working population and the aging of skilled engineers have made the transfer of tacit knowledge, a source of competitiveness, a serious social issue. The key to the next stage of innovation lies in using non-public data inside companies, but converting it into a form that AI can learn from and use is extremely difficult for individual companies to pursue alone. Through collaboration with 16 leading Japanese companies, Stockmark will build a cross-industry AI ecosystem. Under its mission to “reinvent the mechanism of value creation and advance humanity,” the company aims to enable AI to autonomously handle repetitive tasks that create stagnation, allowing people to focus on value creation and contributing to the competitiveness of Japanese companies as a whole. GENIAC is a project run by METI and NEDO to strengthen Japan’s domestic foundation model development capabilities and encourage ingenuity among companies and other organizations. Its activities include providing computing resources, supporting matching with user companies and data holders, assisting collaboration with global technology companies, holding community events, and evaluating the performance of foundation models under development.