Development and Verification of Offline Disaster Countermeasure Support System Utilizing Small Language Models (SLM) Begins
Mitsui Fudosan and Hitachi are developing an offline disaster countermeasure support system using Small Language Models (SLM) to enhance crisis management center operations during communication failures. The AI will extract response details from manuals based on disaster input, providing prioritized initial actions to support decision-making regardless of staff experience. This initiative aims to contribute to sustainable and resilient town development.
📋 Article Processing Timeline
- 📰 Published: April 15, 2026 at 23:31
- 🔍 Collected: April 15, 2026 at 15:01
- 🤖 AI Analyzed: April 15, 2026 at 21:10 (6h 8m after Collected)
Key Points of This Release
■ Jointly develop an SLM-equipped disaster countermeasure support system that can operate even during communication failures, establishing a system where the crisis management center can function without interruption during disasters.
■ Depending on the different facilities and situations of each building, when disaster status is input, AI extracts response details from extensive manuals. It presents initial response actions with priority, supporting accurate judgment regardless of the proficiency of on-duty staff.
■ Fusion of Mitsui Fudosan's crisis management know-how with Hitachi's experience and AI technology in building solutions. Aiming for full-scale operation after verification, and considering deployment to other businesses, contributing to sustainable and resilient town development.
Mitsui Fudosan Co., Ltd. (hereinafter Mitsui Fudosan) and Hitachi, Ltd. (hereinafter Hitachi) have commenced the development and verification of an offline disaster countermeasure support system (hereinafter the System) utilizing SLM (Small Language Model), a type of generative AI, to further enhance disaster response capabilities at the 'Crisis Management Center' that oversees approximately 200 office buildings owned by Mitsui Fudosan nationwide.
To ensure accuracy that can withstand practical use even in an on-premises environment*1, specialized knowledge regarding building operations and disaster response has been incorporated into the AI. Through Hitachi's Generative AI Utilization Professional Services, the System realizes a disaster countermeasure support system that can be used in the field, leveraging business-specific language models*2 and AI agents*3.
Currently, both companies are proceeding with the demonstration of the System, aiming for early full-scale operation through technical verification and improvement. Furthermore, by aiming to offer the System as part of Hitachi's HMAX for Buildings: BuilMirai lineup, they will provide the results of this cutting-edge initiative, established by both companies, to other office building management operators, contributing to sustainable and resilient town development.
Screen of the offline disaster countermeasure support system currently being demonstrated at Mitsui Fudosan's Crisis Management Center
*1: A form of system implementation and utilization where companies use their own prepared facilities such as servers and software.
*2: Hitachi News Release August 29, 2024, 'Launch of 'Business-Specific LLM Construction and Operation Service' for Full-Scale Generative AI Business Utilization'.
*3: Hitachi News Release March 26, 2025, 'Hitachi Rapidly Provides Customer-Specific AI Agents Through Utilization Methods of OT Knowledge Acquired from Hundreds of Cases'.
Significance of This System Development
In Japan, natural disasters are intensifying and the working-age population is declining, making it an urgent task to balance 'ensuring safety and security' with 'efficiency and labor saving' in office building operations. Mitsui Fudosan's Crisis Management Center, as a permanent central hub, operates 24/7/365 with on-duty staff (two employees on rotation). When tremors of intensity 5 or stronger are observed, a Disaster Countermeasure Headquarters is established within the center, coordinating with regional headquarters of approximately 300 personnel, centrally managing nationwide information, and rapidly providing information to building users while ensuring continuity and recovery of functions.
However, in the event of a large-scale disaster, fixed and mobile phone lines and internet connections may be restricted, creating a need for a mechanism that does not rely on communication and enables initial response actions regardless of the proficiency of on-duty staff.
Therefore, Mitsui Fudosan and Hitachi have commenced the development and verification of an offline disaster countermeasure support system utilizing SLM, based on the belief that enabling on-duty staff at the Crisis Management Center to respond appropriately even during large-scale disasters with communication failures will lead to the safe and secure daily lives of users operating from buildings. This system provides accurate support that is not affected by communication environments.
Strengths and Roles of Both Companies in This System Development
Mitsui Fudosan provides on-site crisis management expertise cultivated through the operation of approximately 200 office buildings, extensive operational and disaster response manuals, and BCP/BCM governance, leading the requirements definition and evaluation. They ensure precise design aligned with the operational realities of the Crisis Management Center, guaranteeing individual optimization for each building and operational implementability.
Hitachi leverages its expertise gained from building facilities, including elevators, and the operation of control centers, as well as its development capabilities for next-generation solution groups like HMAX for Buildings that create new value by combining data with domain knowledge and advanced AI. They designed and implemented a lightweight and highly reliable offline AI system. With an architecture rooted in strict security requirements and mission-critical operations, cultivated through the Digital System & Service Sector's experience in the financial sector, they met Mitsui Fudosan's advanced demands.
Features of This System
(1) System Configuration Utilizing Generative AI in an Offline Environment
To address network failures and external server failures anticipated during large-scale disasters, a mechanism operating in a local environment has been adopted. This builds a system infrastructure that does not depend on cloud connectivity and keeps the Crisis Management Center's decision-making operational. Generative AI (SLM) on this infrastructure organizes information necessary for initial response decision-making and presents response plans, supporting the center staff's judgment.
(2) Automatic Presentation of 'Initial Response Guidance' Specialized for Disaster Countermeasure Operations
When disaster status at each building is input from devices such as smartphones, the generative AI comprehensively searches and interprets extensive disaster response manuals, presenting the center staff with prioritized response and work items. It provides support tailored to the individual requirements of buildings with different locations and facility configurations.
(3) Inheriting Experienced Personnel's Knowledge as 'Organizational Strength'
To prepare for on-site events that exceed manual assumptions, the knowledge and past response know-how of experienced personnel are learned by the generative AI in advance. This enables rapid and accurate initial responses that are not dependent on the proficiency of on-duty staff.
(4) Utilization of VLM and Fine-tuning for High-Level Responses
Achieves response accuracy and speed comparable to general-purpose large language models even during large-scale disasters. When responding, it also presents the reference manual information, providing high explanatory power that allows judgment while confirming the basis for the response content. To improve the accuracy of this system, a VLM (Vision Language Model) capable of handling diagram and table information within manuals has been applied. Furthermore, by utilizing manuals and past question-and-answer data, the model has been fine-tuned to train the AI on specialized disaster response knowledge and the response formats required in the field, resulting in response performance suitable for the Crisis Management Center's operations.
The SLM adopted in this system features a lightweight model design compared to general-purpose LLMs (Large Language Models), allowing it to operate on common PCs and smart devices used in offices. This simultaneously achieves: 1. Continued operation during communication failures (no cloud required), 2. Protection of intellectual property and confidential information (sensitive manuals and operational know-how are not placed externally), and 3. High-precision optimization for specific tasks (focused learning on Mitsui Fudosan's specialized operations).
Future Initiatives
Mitsui Fudosan aims to achieve both increased productivity and added value through the utilization of generative AI, and has been deploying ChatGPT Enterprise licenses to approximately 2,000 employees since October 1, 2025. As societal and customer needs diversify and become more sophisticated, they plan to gradually expand the application of generative AI to areas such as advanced management decision-making, addressing labor shortages in the field, and integration with data analysis to create higher levels of added value.
Hitachi's CI (Connective Industries) Sector Urban Solution & Service Business Unit is focusing on 'HMAX for Buildings,' a group of next-generation solutions that combine domain knowledge and advanced AI with data from a rich installed base of products (digitized assets). Through the provision of 'Facilitity Services' centered on these, they aim to maximize customers' lifetime value and transform industries globally, contributing to the realization of a prosperous society.
Following the demonstration, Mitsui Fudosan and Hitachi aim for full-scale operation at Mitsui Fudosan's Crisis Management Center, and will also provide the knowledge gained from this initiative to other office building management operators, contributing to the development of sustainable and resilient cities.
About Mitsui Fudosan
Under the philosophy of 'Creating New Value Through Coexistence, Co-harmony, and Co-creation,' Mitsui Fudosan promotes town development utilizing diverse assets such as office buildings, commercial facilities, residences, hotels, and logistics facilities. Centered on urban development, they address social issues such as improving industrial competitiveness, coexisting with the environment, and enhancing well-being through building operations, area management, and digital utilization. Furthermore, by leveraging their extensive assets, including office buildings nationwide, and their customer base, they promote initiatives that foster inter-company exchange and the creation of new value, contributing to the realization of sustainable and attractive cities.
About Hitachi, Ltd.
Hitachi contributes to the realization of a Harmonized Society where environment, well-being, and economic growth are in balance through its Social Innovation Business (SIB), which utilizes IT, OT (Operational Technology), and products. Operating globally with a business structure comprising four sectors—Digital System & Services, Energy, Mobility, and Connective Industries—along with a Strategic SIB Business Unit for creating new growth businesses, Hitachi solves customer and societal challenges by creating value from data with Lumada at its core. In FY2024 (ending March 2025), revenue was 9,783.3 billion yen, with 618 consolidated subsidiaries and approximately 280,000 employees worldwide as of the end of March 2025. For more information, please visit www.hitachi.co.jp.
Trademark Notice
Company names and product names mentioned herein are trademarks or registered trademarks of their respective companies.
■ Jointly develop an SLM-equipped disaster countermeasure support system that can operate even during communication failures, establishing a system where the crisis management center can function without interruption during disasters.
■ Depending on the different facilities and situations of each building, when disaster status is input, AI extracts response details from extensive manuals. It presents initial response actions with priority, supporting accurate judgment regardless of the proficiency of on-duty staff.
■ Fusion of Mitsui Fudosan's crisis management know-how with Hitachi's experience and AI technology in building solutions. Aiming for full-scale operation after verification, and considering deployment to other businesses, contributing to sustainable and resilient town development.
Mitsui Fudosan Co., Ltd. (hereinafter Mitsui Fudosan) and Hitachi, Ltd. (hereinafter Hitachi) have commenced the development and verification of an offline disaster countermeasure support system (hereinafter the System) utilizing SLM (Small Language Model), a type of generative AI, to further enhance disaster response capabilities at the 'Crisis Management Center' that oversees approximately 200 office buildings owned by Mitsui Fudosan nationwide.
To ensure accuracy that can withstand practical use even in an on-premises environment*1, specialized knowledge regarding building operations and disaster response has been incorporated into the AI. Through Hitachi's Generative AI Utilization Professional Services, the System realizes a disaster countermeasure support system that can be used in the field, leveraging business-specific language models*2 and AI agents*3.
Currently, both companies are proceeding with the demonstration of the System, aiming for early full-scale operation through technical verification and improvement. Furthermore, by aiming to offer the System as part of Hitachi's HMAX for Buildings: BuilMirai lineup, they will provide the results of this cutting-edge initiative, established by both companies, to other office building management operators, contributing to sustainable and resilient town development.
Screen of the offline disaster countermeasure support system currently being demonstrated at Mitsui Fudosan's Crisis Management Center
*1: A form of system implementation and utilization where companies use their own prepared facilities such as servers and software.
*2: Hitachi News Release August 29, 2024, 'Launch of 'Business-Specific LLM Construction and Operation Service' for Full-Scale Generative AI Business Utilization'.
*3: Hitachi News Release March 26, 2025, 'Hitachi Rapidly Provides Customer-Specific AI Agents Through Utilization Methods of OT Knowledge Acquired from Hundreds of Cases'.
Significance of This System Development
In Japan, natural disasters are intensifying and the working-age population is declining, making it an urgent task to balance 'ensuring safety and security' with 'efficiency and labor saving' in office building operations. Mitsui Fudosan's Crisis Management Center, as a permanent central hub, operates 24/7/365 with on-duty staff (two employees on rotation). When tremors of intensity 5 or stronger are observed, a Disaster Countermeasure Headquarters is established within the center, coordinating with regional headquarters of approximately 300 personnel, centrally managing nationwide information, and rapidly providing information to building users while ensuring continuity and recovery of functions.
However, in the event of a large-scale disaster, fixed and mobile phone lines and internet connections may be restricted, creating a need for a mechanism that does not rely on communication and enables initial response actions regardless of the proficiency of on-duty staff.
Therefore, Mitsui Fudosan and Hitachi have commenced the development and verification of an offline disaster countermeasure support system utilizing SLM, based on the belief that enabling on-duty staff at the Crisis Management Center to respond appropriately even during large-scale disasters with communication failures will lead to the safe and secure daily lives of users operating from buildings. This system provides accurate support that is not affected by communication environments.
Strengths and Roles of Both Companies in This System Development
Mitsui Fudosan provides on-site crisis management expertise cultivated through the operation of approximately 200 office buildings, extensive operational and disaster response manuals, and BCP/BCM governance, leading the requirements definition and evaluation. They ensure precise design aligned with the operational realities of the Crisis Management Center, guaranteeing individual optimization for each building and operational implementability.
Hitachi leverages its expertise gained from building facilities, including elevators, and the operation of control centers, as well as its development capabilities for next-generation solution groups like HMAX for Buildings that create new value by combining data with domain knowledge and advanced AI. They designed and implemented a lightweight and highly reliable offline AI system. With an architecture rooted in strict security requirements and mission-critical operations, cultivated through the Digital System & Service Sector's experience in the financial sector, they met Mitsui Fudosan's advanced demands.
Features of This System
(1) System Configuration Utilizing Generative AI in an Offline Environment
To address network failures and external server failures anticipated during large-scale disasters, a mechanism operating in a local environment has been adopted. This builds a system infrastructure that does not depend on cloud connectivity and keeps the Crisis Management Center's decision-making operational. Generative AI (SLM) on this infrastructure organizes information necessary for initial response decision-making and presents response plans, supporting the center staff's judgment.
(2) Automatic Presentation of 'Initial Response Guidance' Specialized for Disaster Countermeasure Operations
When disaster status at each building is input from devices such as smartphones, the generative AI comprehensively searches and interprets extensive disaster response manuals, presenting the center staff with prioritized response and work items. It provides support tailored to the individual requirements of buildings with different locations and facility configurations.
(3) Inheriting Experienced Personnel's Knowledge as 'Organizational Strength'
To prepare for on-site events that exceed manual assumptions, the knowledge and past response know-how of experienced personnel are learned by the generative AI in advance. This enables rapid and accurate initial responses that are not dependent on the proficiency of on-duty staff.
(4) Utilization of VLM and Fine-tuning for High-Level Responses
Achieves response accuracy and speed comparable to general-purpose large language models even during large-scale disasters. When responding, it also presents the reference manual information, providing high explanatory power that allows judgment while confirming the basis for the response content. To improve the accuracy of this system, a VLM (Vision Language Model) capable of handling diagram and table information within manuals has been applied. Furthermore, by utilizing manuals and past question-and-answer data, the model has been fine-tuned to train the AI on specialized disaster response knowledge and the response formats required in the field, resulting in response performance suitable for the Crisis Management Center's operations.
The SLM adopted in this system features a lightweight model design compared to general-purpose LLMs (Large Language Models), allowing it to operate on common PCs and smart devices used in offices. This simultaneously achieves: 1. Continued operation during communication failures (no cloud required), 2. Protection of intellectual property and confidential information (sensitive manuals and operational know-how are not placed externally), and 3. High-precision optimization for specific tasks (focused learning on Mitsui Fudosan's specialized operations).
Future Initiatives
Mitsui Fudosan aims to achieve both increased productivity and added value through the utilization of generative AI, and has been deploying ChatGPT Enterprise licenses to approximately 2,000 employees since October 1, 2025. As societal and customer needs diversify and become more sophisticated, they plan to gradually expand the application of generative AI to areas such as advanced management decision-making, addressing labor shortages in the field, and integration with data analysis to create higher levels of added value.
Hitachi's CI (Connective Industries) Sector Urban Solution & Service Business Unit is focusing on 'HMAX for Buildings,' a group of next-generation solutions that combine domain knowledge and advanced AI with data from a rich installed base of products (digitized assets). Through the provision of 'Facilitity Services' centered on these, they aim to maximize customers' lifetime value and transform industries globally, contributing to the realization of a prosperous society.
Following the demonstration, Mitsui Fudosan and Hitachi aim for full-scale operation at Mitsui Fudosan's Crisis Management Center, and will also provide the knowledge gained from this initiative to other office building management operators, contributing to the development of sustainable and resilient cities.
About Mitsui Fudosan
Under the philosophy of 'Creating New Value Through Coexistence, Co-harmony, and Co-creation,' Mitsui Fudosan promotes town development utilizing diverse assets such as office buildings, commercial facilities, residences, hotels, and logistics facilities. Centered on urban development, they address social issues such as improving industrial competitiveness, coexisting with the environment, and enhancing well-being through building operations, area management, and digital utilization. Furthermore, by leveraging their extensive assets, including office buildings nationwide, and their customer base, they promote initiatives that foster inter-company exchange and the creation of new value, contributing to the realization of sustainable and attractive cities.
About Hitachi, Ltd.
Hitachi contributes to the realization of a Harmonized Society where environment, well-being, and economic growth are in balance through its Social Innovation Business (SIB), which utilizes IT, OT (Operational Technology), and products. Operating globally with a business structure comprising four sectors—Digital System & Services, Energy, Mobility, and Connective Industries—along with a Strategic SIB Business Unit for creating new growth businesses, Hitachi solves customer and societal challenges by creating value from data with Lumada at its core. In FY2024 (ending March 2025), revenue was 9,783.3 billion yen, with 618 consolidated subsidiaries and approximately 280,000 employees worldwide as of the end of March 2025. For more information, please visit www.hitachi.co.jp.
Trademark Notice
Company names and product names mentioned herein are trademarks or registered trademarks of their respective companies.