← 事例

この事例は現在、英語でのみご覧いただけます。

日本製造2023 – 継続中

Toyota Systems

Generative AI for core-system code updates (with Fujitsu)

Toyota Systems halved the time to update core-system code with generative AI in a 15,000-file trial

進行中企業自身の発表公式資料 ↗

課題

Toyota Systems develops and runs the Toyota group's core systems for production, logistics and sales. Updating their operating systems and programming languages meant finding and fixing every incompatible part of the code by hand, a large and recurring maintenance job.

取り組み

From October 2023 the company tested Fujitsu's Kozuchi Generative AI service on about 15,000 Java and SQLJ files. The AI identified code likely to break under the update and rewrote it, and engineers checked the changes. The companies said the technology would move to live operations in January 2025 and be extended to other languages and to testing.

主な数値

-50%work time for code updates versus manual work, in the trial企業自身の発表[1][2]
~15,000Java and SQLJ files covered by the trial企業自身の発表[1][2]

示唆

The 50% saving covered a narrow, repetitive maintenance task with clear correctness checks, leaving open-ended coding outside the trial. The customer and vendor jointly reported the trial result.

分類

規模
大企業
業種
製造
業務機能
IT・研究開発
イノベーションの類型
プロセス・イノベーション
AI技術
生成AI
根拠
企業自身の発表

出典

出典の要約

[1]Fujitsu and Toyota Systems2024年10月24日

Toyota Systems and Fujitsu said a trial on about 15,000 Java and SQLJ files cut code-update work time by about 50% compared with manual work. They said Toyota Systems would apply the technology in live operations from January 2025.

[2]ITmedia NEWS2024年10月25日

ITmedia reported the trial that began in October 2023 on core systems for production, logistics and sales. It describes the AI finding incompatible code and modifying it, with a reported 50% cut in work time over 15,000 files. It adds plans to cover other languages and test work and Fujitsu's aim to sell the capability as a service.

詳細は原典でご確認いただけます。

掲載日 2026年10月25日 · 更新日 2026年10月25日
事例ページを開く ↗