Cinnamon AI Fully Provides Next-Generation KY Suggestion System Automatically Supporting 'Hidden Dangers in Tasks' with High-Precision Generative AI 'Super RAG™' to Mitsubishi Gas Chemical
Cinnamon AI will begin full-scale development in April 2026 to integrate its generative AI 'Super RAG' into the Danger Prediction (KY) system at Mitsubishi Gas Chemical's 5 plants. It automatically extracts risk factors from unstructured data to support safety management.
📋 Article Processing Timeline
- 📰 Published: April 23, 2026 at 22:00
- 🔍 Collected: April 23, 2026 at 13:31
- 🤖 AI Analyzed: April 24, 2026 at 00:19 (10h 47m after Collected)
Cinnamon Inc. (Headquarters: Chiyoda-ku, Tokyo, Representative Director and CEO: Mirai Hirano, hereinafter 'Cinnamon AI'), which provides AI (Artificial Intelligence) solutions to major domestic companies, will begin full-scale development in April 2026 to incorporate its 'Super RAG™', which realizes highly accurate internal knowledge search and answer generation, into the KY (Danger Prediction) suggestion system (hereinafter 'MGC-KYAS') currently in operation at five domestic plants of Mitsubishi Gas Chemical Company, Inc. (Headquarters: Chiyoda-ku, Tokyo, Representative Director and President: Masanori Isahaya, hereinafter 'Mitsubishi Gas Chemical'). The new system is scheduled to be introduced sequentially within the year.
MGC-KYAS is a system that effectively supports pre-work KY activities by extracting cases and information related to on-site work from a massive case database obtained through past near-miss activities (*1). Based on Cinnamon AI's advanced AI solutions as a technological foundation, this system has been introduced to Mitsubishi Gas Chemical's five domestic plants since around June 2023.
With the introduction of 'Super RAG', it becomes possible to read vast amounts of documents saved in various formats across different factories and departments with high precision. This knowledge can be utilized for multifaceted purposes such as immediate reference to past similar disasters, optimization of work procedures, and creation of educational materials for safety training. Additionally, by utilizing AI in KY activities, it prevents the bias in perspectives and the rut that humans tend to fall into alone, drastically improving the comprehensiveness of danger prediction. Amidst a worsening labor shortage, a mechanism will be built to securely pass down the valuable know-how of danger information held by veterans via AI, further solidifying the effects of preventing human errors and raising the sensitivity of workers.
*1 Initiatives to report and share incidents where employees felt 'chills' or were 'startled' but which did not lead to accidents or disasters, connecting them to systematic disaster prevention activities.
■ Background: The Challenge of Passing Down Veteran Knowledge and the Formalization of KY Activities
At manufacturing sites, 'KY (Danger Prediction) activities', which nip the buds of accidents and disasters before work begins, are the cornerstone of safety management. However, in recent years, while equipment automation has progressed, opportunities for employees to directly handle safety and disaster prevention tasks on-site have decreased, raising concerns about a decline in danger sensitivity on the floor. Furthermore, although a massive amount of near-miss cases and occupational accident data has been accumulated, it is difficult to utilize effectively because it is saved in different formats for each site. This has led to challenges such as activities becoming a mere formality or requiring a significant amount of time to extract necessary information.
To solve these challenges, Cinnamon AI jointly constructed the KY suggestion system 'MGC-KYAS' with Mitsubishi Gas Chemical, which accurately and automatically suggests 'hidden dangers in tasks' from vast amounts of past knowledge, and has promoted its introduction to 5 domestic factories.
■ Main Achievements and Features of Introducing 'Super RAG'
By implementing Cinnamon AI's proprietary RAG (Retrieval-Augmented Generation) technology, 'Super RAG', into 'MGC-KYAS', the following three points will be realized.
1. Improvement of KY Activity Quality through Rapid Information Extraction
Previously, investigating past similar cases required searching through massive files. With this system, simply entering the equipment number or task name allows the AI to instantly extract highly relevant cases. By automatically presenting multiple instances of 'hidden dangers in tasks'—the first step in danger prediction training—it minimizes the burden of information gathering on-site and significantly improves the quality of KY activities.
2. Reduction of Operational Burden through Direct Utilization of Unstructured Data
Through the advanced analytical capabilities of 'Super RAG', it is possible to generate highly accurate answers directly from 'original files' like PDFs without processing them into 'structured data' for Excel, etc. This achieves high convenience by drastically reducing the man-hours required for data registration on-site while making buried knowledge immediately usable.
3. Passing Down Veteran Know-How with AI, Breaking the Rut of KY Activities
By having AI instantly draw out the tacit knowledge of veterans, safety know-how that tends to rely on specific individuals is standardized, realizing reliable technology transfer even amidst labor shortages. Through a feature where AI clarifies the basis of its answers (reference documents), young employees can directly learn the judgment criteria and past lessons of skilled workers, leading to improved danger sensitivity.
Additionally, as AI presents 'hidden dangers in tasks' from new perspectives, it breaks the routine thinking that often plagues daily work, and continuously updates the company-wide safety culture without letting past major accidents be forgotten.
■ Future Outlook
Cinnamon AI will strengthen its collaboration with Mitsubishi Gas Chemical to consider expanding the application scope of 'MGC-KYAS'. Furthermore, to group companies and
MGC-KYAS is a system that effectively supports pre-work KY activities by extracting cases and information related to on-site work from a massive case database obtained through past near-miss activities (*1). Based on Cinnamon AI's advanced AI solutions as a technological foundation, this system has been introduced to Mitsubishi Gas Chemical's five domestic plants since around June 2023.
With the introduction of 'Super RAG', it becomes possible to read vast amounts of documents saved in various formats across different factories and departments with high precision. This knowledge can be utilized for multifaceted purposes such as immediate reference to past similar disasters, optimization of work procedures, and creation of educational materials for safety training. Additionally, by utilizing AI in KY activities, it prevents the bias in perspectives and the rut that humans tend to fall into alone, drastically improving the comprehensiveness of danger prediction. Amidst a worsening labor shortage, a mechanism will be built to securely pass down the valuable know-how of danger information held by veterans via AI, further solidifying the effects of preventing human errors and raising the sensitivity of workers.
*1 Initiatives to report and share incidents where employees felt 'chills' or were 'startled' but which did not lead to accidents or disasters, connecting them to systematic disaster prevention activities.
■ Background: The Challenge of Passing Down Veteran Knowledge and the Formalization of KY Activities
At manufacturing sites, 'KY (Danger Prediction) activities', which nip the buds of accidents and disasters before work begins, are the cornerstone of safety management. However, in recent years, while equipment automation has progressed, opportunities for employees to directly handle safety and disaster prevention tasks on-site have decreased, raising concerns about a decline in danger sensitivity on the floor. Furthermore, although a massive amount of near-miss cases and occupational accident data has been accumulated, it is difficult to utilize effectively because it is saved in different formats for each site. This has led to challenges such as activities becoming a mere formality or requiring a significant amount of time to extract necessary information.
To solve these challenges, Cinnamon AI jointly constructed the KY suggestion system 'MGC-KYAS' with Mitsubishi Gas Chemical, which accurately and automatically suggests 'hidden dangers in tasks' from vast amounts of past knowledge, and has promoted its introduction to 5 domestic factories.
■ Main Achievements and Features of Introducing 'Super RAG'
By implementing Cinnamon AI's proprietary RAG (Retrieval-Augmented Generation) technology, 'Super RAG', into 'MGC-KYAS', the following three points will be realized.
1. Improvement of KY Activity Quality through Rapid Information Extraction
Previously, investigating past similar cases required searching through massive files. With this system, simply entering the equipment number or task name allows the AI to instantly extract highly relevant cases. By automatically presenting multiple instances of 'hidden dangers in tasks'—the first step in danger prediction training—it minimizes the burden of information gathering on-site and significantly improves the quality of KY activities.
2. Reduction of Operational Burden through Direct Utilization of Unstructured Data
Through the advanced analytical capabilities of 'Super RAG', it is possible to generate highly accurate answers directly from 'original files' like PDFs without processing them into 'structured data' for Excel, etc. This achieves high convenience by drastically reducing the man-hours required for data registration on-site while making buried knowledge immediately usable.
3. Passing Down Veteran Know-How with AI, Breaking the Rut of KY Activities
By having AI instantly draw out the tacit knowledge of veterans, safety know-how that tends to rely on specific individuals is standardized, realizing reliable technology transfer even amidst labor shortages. Through a feature where AI clarifies the basis of its answers (reference documents), young employees can directly learn the judgment criteria and past lessons of skilled workers, leading to improved danger sensitivity.
Additionally, as AI presents 'hidden dangers in tasks' from new perspectives, it breaks the routine thinking that often plagues daily work, and continuously updates the company-wide safety culture without letting past major accidents be forgotten.
■ Future Outlook
Cinnamon AI will strengthen its collaboration with Mitsubishi Gas Chemical to consider expanding the application scope of 'MGC-KYAS'. Furthermore, to group companies and