Tokushima University-born AI Company ILU Launches New Service "Manufacturing Site Knowledge AI" to Support Skills Transfer and Knowledge Sharing in Manufacturing

ILU Inc. (Tokushima City, Tokushima Prefecture), an AI company born from Tokushima University with 24 years of history and part of the Sansan Group, will begin offering its new service "Manufacturing Site Knowledge AI" on July 1st. This service utilizes generative AI to make electronic manuals easier to use, supporting skills transfer and knowledge sharing in the manufacturing industry.

In Japan's manufacturing sector, which supports the nation's "monozukuri" (manufacturing prowess), the aging of skilled workers and labor shortages have become significant challenges in recent years, making the transfer of technology and know-how to the next generation a crucial theme.1 As a measure to address these issues, efforts have been made to visualize the knowledge and experience cultivated by skilled workers and to digitize manuals.

However, to effectively utilize digitized information in the field, it is essential to create an environment that allows appropriate access to necessary knowledge. Particularly with search systems utilizing generative AI, if the document structure and information relationships are not organized, or if different terms are used at different sites, it may not be possible to search for information appropriately.

To address these specific challenges in manufacturing sites, ILU is launching "Manufacturing Site Knowledge AI," a new service that utilizes its proprietary technology (natural language processing technology)3 accumulated over approximately 40 years since its research began at Tokushima University. This technology organizes and structures human knowledge and know-how into a form that computers can understand and utilize.

This service structures information such as design documents, procedure manuals, and inspection records written in PDF, Excel, charts, etc., based on their meaning and relationships2, and converts it into data that generative AI can easily use. This provides an environment where all workers can find the necessary information without hesitation, regardless of differences in expression.

Furthermore, ILU aims to provide this service to 6 companies annually, promote skills transfer in manufacturing sites, and achieve sustainable development.

1 For details, see "Challenges Facing Japan's Manufacturing Sites." 2 For details, see "What is Information Structuring." 3 For details, see "What is Natural Language Processing Technology."

Overview of the New Service "Manufacturing Site Knowledge AI"

This service organizes and structures electronic data such as design documents, procedure manuals, and inspection records accumulated in manufacturing sites, and is a generative AI utilization support service for searching and utilizing knowledge that has previously depended on the know-how of skilled workers, using natural language.

It can be introduced in a way that leverages existing management systems and current site operations, allowing for its use while minimizing the burden on the site, such as learning to operate new tools.

Image of "Manufacturing Site Knowledge AI"

Detailed Functions of "Manufacturing Site Knowledge AI"

1 Retains Document Structure

Documents such as manuals and procedure manuals are organized and structured while maintaining the relationships between headings, items, and tables. By digitizing the information while preserving its connections, generative AI can better understand the content, creating an environment where necessary information can be accurately searched and utilized.

2 Countermeasures for Notation Variations (Synonym Map Technology)

Terms used with different names or expressions depending on the site or person are treated as words with the same meaning. For example, even if the same equipment has different names or abbreviations used at different sites, by recognizing them as having the same meaning, necessary information can be searched for regardless of differences in expression.

3 Design of Keyword Tags and Category Classification (Annotation Technology)

Keywords such as "voltage" or "part replacement" are tagged to manuals and troubleshooting history, and categories such as "inspection" or "repair" are classified. This creates an environment where necessary information is easy to find.

4 Attaches Grounding Information to Answers Using RAG (RAG: Retrieval-Augmented Generation)4

Relevant information is retrieved from internal data such as registered manuals, and the AI generates answers based on that content. Since the source manual name and page information are presented along with the answer, field workers can check the original text as needed.

5 Pre-verification to Maintain Answer Accuracy

By preparing anticipated questions and answers for the field in advance, the AI's answer results are continuously checked and evaluated. This allows for confirmation of the validity of search results and answers before implementation.

4 For details, see "What is RAG (Retrieval-Augmented Generation)."

・ Challenges Facing Japan's Manufacturing Sites

According to the Ministry of Internal Affairs and Communications' "Labor Force Survey5," the working population in Japan is on the rise, driven by increased female employment rates, expanded employment of the elderly, and an increase in foreign workers. While the utilization of diverse human resources is progressing, there are many situations where it is difficult to smoothly share necessary knowledge and know-how due to differences in years of experience, changes in responsibilities, returns from childcare/parental leave, and language differences.

This situation is also reflected in the skills transfer challenges faced by the manufacturing industry as a whole. The "2026 Manufacturing White Paper6," released on May 29, Reiwa 8 (2026), indicates that many manufacturing companies recognize skills transfer as an important management issue. In particular, many companies cite "the ability to respond to troubles and unexpected events" as skills to be transferred, suggesting that manufacturing sites are still supported by the experience and judgment of skilled workers.

Furthermore, many companies cite "aging and retirement of skilled workers" as a future concern, and efforts such as re-employment and extended working hours for elderly employees, and visualization of skills (manualization and digitization) are being promoted.

5 Ministry of Internal Affairs and Communications Statistics Bureau website, "Labor Force Survey (Basic Tabulation) April 2026 (Reiwa 8)", May 29, Reiwa 8 (2026). 6 Ministry of Economy, Trade and Industry website, "2026 Manufacturing White Paper (Annual Report Based on Article 8 of the Basic Act for the Promotion of Manufacturing Industries)", May 29, Reiwa 8 (2026).

・ What is Information Structuring

This refers to organizing the relationships between items and terms within data to make information easier to use, arranging it in a form that computers can understand the meaning of. By structuring information, generative AI can more easily treat data as meaningful information rather than just strings of text, enabling more appropriate searches and responses.

・ What is Natural Language Processing Technology

Technology that enables computers to process, understand, and utilize the language that people use in daily life (natural language) is called "natural language processing." Natural language processing includes various methods, such as processing methods based on rules defined by humans (rule-based) and generative AI using large language models (LLMs).

ILU's strength lies in rule-based Japanese processing. Based on development and research results accumulated over approximately 40 years since its time as a Tokushima University research laboratory, it has built one of Japan's largest Japanese databases that systematizes the meaning, grammar, and expressions of the Japanese language. Utilizing this linguistic asset, it is engaged in high-precision Japanese processing that captures the meaning, context, and differences in expression of words.

・ What is RAG (Retrieval-Augmented Generation)

RAG is a technology where generative AI creates answers by searching and referencing information accumulated within a company, such as internal manuals and business documents, in addition to the knowledge it has learned. Since it can utilize information not publicly available, it makes it easier to obtain answers tailored to the specific site.

Company Profile of ILU Inc.

With the mission "To maximize the value of words and continue to exceed customer expectations," ILU engages in research and development of natural language processing technology. In particular, it excels in rule-based technology that processes words based on human-defined rules. Leveraging knowledge accumulated over 40 years since its time as a Tokushima University research laboratory and one of Japan's largest Japanese databases, it achieves high-precision and reproducible Japanese processing that accounts for the ambiguity and differences in expression unique to the Japanese language. Established: January 28, 2002 URL: https://www.ilu.co.jp/ Location: 1-32-1 Nakajyo-cho, Tokushima City, 770-00813, Japan Capital: 58 million yen (as of June 2026) Business Activities: Creating operational efficiencies and advanced value-added services utilizing large-scale knowledge related to Japanese language processing.

[Display Regarding Trademarks, etc.]

* Company names and product/service names mentioned are trademarks or registered trademarks of their respective companies.

<Inquiries from the Press> ILU Inc. General Affairs Department, Public Relations Representative: Mai Yoshida Phone Number: 088-657-7624

MAIL: info@ilu.co.jp