Exa Frontier Edge, an ExaWizards Group Company, Advances AI-Driven Reconstruction of Chugoku Electric Power’s Hydroelectric Power Planning System – Rebuilding the Verification System on a Cloud-Native Architecture Based on AI-Generated Specifications, Confirming Operational Consistency with the Existing System in About Two Months, and Applying the Results to Improve the Existing System First –

Exa Frontier Edge Inc. (Head Office: Minato-ku, Tokyo; Representative Director: Masafumi Fukuda; hereinafter “Exa Frontier Edge”), a group company of ExaWizards Inc. (Head Office: Minato-ku, Tokyo; President & CEO: Makoto Haruta; hereinafter “ExaWizards”), has been conducting preliminary verification of AI-driven system reconstruction since September 2026 for the hydroelectric power generation planning system (hereinafter “Hydroelectric Power Planning System”) used by The Chugoku Electric Power Co., Inc. (hereinafter “Chugoku Electric Power”) to formulate generation plans for its hydroelectric power plants.

This initiative uses the Living Spec generation service*1 provided by Exa Frontier Edge. AI analyzed the existing system’s source code and related documentation to produce specifications. Based on those specifications, we rebuilt a verification system with the same functions as the existing system on a new cloud-native*2 technology architecture. As a result, the specifications have been made visible without depending on the knowledge of specific individuals. As of the end of October 2026, about two months after the work began, we confirmed that the verification system matches the existing system’s behavior in the plan-formulation test cases for a reservoir-type hydroelectric power plant used for verification.

In addition, knowledge gained during the specification creation process is already being applied to the maintenance and improvement of the existing system. This creates business value before the system renewal is complete.

☑ Background and Challenges

In long-running business systems, specifications and implementation tend to diverge over time, making the systems dependent on specific individuals and turning them into black boxes. As a result, handling vulnerabilities, applying patches, and responding to audits require substantial manual effort, which increases maintenance burdens. At the same time, when renewing such systems, two major hurdles stand out: accurately understanding the functions and logic of the existing system, and ensuring that the rebuilt system behaves the same way. Business systems that support social infrastructure in particular demand a higher level of reliability and careful decisions about migration.

To address the challenge of accurately understanding existing systems, Exa Frontier Edge began offering the Living Spec generation service in August 2026. The service supports core and business systems built with traditional programming languages such as COBOL and VB.NET, and uses AI to restore and create specifications with high accuracy.

The initiative with Chugoku Electric Power goes beyond specification creation. It uses those specifications as a starting point for AI to rebuild a verification system on a cloud-native architecture, and verifies that the new system matches the existing system’s behavior, a process referred to as “current-new consistency.” Through AI-driven development, we are addressing the two major challenges of system renewal, “making specifications visible” and “ensuring compatibility and behavior,” in an integrated way.

☑ Overview of the Initiative

As part of its consideration of future system renewal options, Chugoku Electric Power is verifying the applicability of AI-driven development. Exa Frontier Edge is supporting this effort through the following process.

Verification Process

Specification creation: AI analyzes the source code and design documents of the existing system to generate high-precision specifications that serve as input for AI-driven development.
Rebuilding the verification system: Based on the generated specifications, AI is used to build a verification system with the same functions as the existing system on a cloud-native technology architecture. Rather than simply porting the existing code, the work involves renewing the programming language, reviewing the architecture, and containerizing the system for future cloud operation, all starting from the specifications.
Current-new comparison*3: Automated tests using identical input data are run to verify that calculation results and report outputs match. As of the end of October 2026, behavior was confirmed to match across all test cases set for verification.

Consistency with the Existing System as the Acceptance Criterion

The risk most important to avoid in a system renewal is that the rebuilt system unintentionally behaves differently from the original. For this reason, this initiative set consistency with the existing system as the acceptance criterion for the rebuilt verification system. Although the rebuilt verification system uses a completely different programming language and architecture, it was demonstrated to return results that are fully identical to the existing system for the same data. This objectively confirms that the specifications created by AI accurately cover and reproduce the business logic of the existing system.

☑ Results Obtained Through the Preliminary Verification

The results of this initiative are already emerging in current operations, without waiting for a full system overhaul. During the specification creation process, AI comprehensively identified exception handling and edge cases, revealing latent issues and improvement opportunities that were difficult to notice in day-to-day operations. Chugoku Electric Power is already applying these insights to modifications of the existing system and to improving maintenance quality.

In this way, the specification creation process itself leads to improvements in the existing system. This makes it possible to prepare for future renewal safely and in stages while continuing to use the existing system, rather than committing to a high-risk, all-at-once “big bang” migration.

▼ For inquiries about the Living Spec generation service, please visit:
https://exawizards.com/efe/

☑ Comment

Masafumi Fukuda, Representative Director, Exa Frontier Edge Inc.
“Conventional system renewals not only require long development periods but are also accompanied by concerns about compatibility, namely whether the rebuilt system will continue to operate as before. This has been a major reason why organizations hesitate to renew their systems. The concern becomes even greater for systems that are critical to business operations, such as power generation planning. In this initiative, we rebuilt a verification system on a new cloud-native technology architecture based on specifications created by AI. In only about two months from the start of the work, we confirmed that it matches the existing system’s behavior in the plan-formulation test cases for a reservoir-type hydroelectric power plant used for verification. I believe that by using AI, the time required for system renewal can be significantly shortened while maintaining quality and safety. It is also highly meaningful that the insights discovered during specification creation are being returned to improve the existing system. Together with Chugoku Electric Power, we will continue to advance this approach, which creates value immediately without waiting for renewal to be completed.”

☑ Future Developments

Based on the knowledge gained from this initiative, Exa Frontier Edge will enhance the Living Spec generation service and provide stepwise support, from automating maintenance based on specifications to full system renewal. Under the banner of “Bringing AI-driven development to organizational strength,” we will continue to contribute to the transformation of corporate system development and operations through the development and provision of AI-native services.

*1 The Living Spec generation service is a service in which AI analyzes existing system source code and related documentation to create specifications. “Living Spec” refers to specifications that are restored from source code and kept up to date as the implementation changes. (Related announcement: “ExaWizards Group’s Exa Frontier Edge Begins Offering Living Spec Generation Service, Which Restores Specifications from Source Code with AI” dated August 12, 2026.)
*2 Cloud-native: A technology architecture designed and built on the premise of operating in a cloud environment.
*3 Current-new comparison: A method of supplying the same data to the existing system and a newly built system, comparing the processing results, and confirming that their behavior matches.

The Chugoku Electric Power Co., Inc. Profile
Company name: The Chugoku Electric Power Co., Inc.
Location: 4-33 Komachi, Naka-ku, Hiroshima City, Hiroshima Prefecture
Established: May 1, 1951 Representative: Kengo Nakagawa, President and Representative Executive Officer
Business: Integrated energy business, primarily in the Chugoku region
URL: https://www.energia.co.jp/e/index.html

Exa Frontier Edge Profile
Company name: Exa Frontier Edge Inc.
Location :4-2-8 Shibaura, Minato-ku, Tokyo, 5F
Established: April 2026
Representative: Masafumi Fukuda, Representative Director
Business: Providing development and operational frameworks predicated on AI technologies, including coding agents, and building/deploying AI-driven development platforms for enterprises.

ExaWizards Company Profile
Company name: ExaWizards Inc.
Location: 4-2-8 Shibaura, Minato-ku, Tokyo, 5F
Established: February 2016
Representative: Makoto Haruta, Representative Director, President & CEO
Business: Industrial innovation and resolution of social issues via AI service development
URL: https://exawizards.com/en/

＜Media Contact＞
ExaWizards Inc. Public Relations:publicrelations@exwzd.com