IBM, watch out! Fujitsu uses AI to understand COBOL and automatically generate design documents “without expert knowledge” in minutes rather than hours
- Fujitsu automates COBOL documentation, cutting analysis time by 97%
- The system generates design documents without requiring deep programming expertise
- Knowledge Graph retrieval reduces hallucinations and improves documentation completeness
Fujitsu has introduced a new generative AI service which analyzes COBOL and other legacy source code, automatically producing design documents in minutes rather than hours.
The system can operate and automatically produce design documents without expert knowledge, reducing reliance on specialized human programmers.
The company says Fujitsu Application Transform powered by Fujitsu Kozuchi eliminates the need for extensive manual review, cutting the time required to understand complex source code by approximately 97%.
Article continues below
Accuracy and readability improvements
The service aims to provide a clear understanding of existing system specifications, offering organizations a practical tool to support modernization strategies efficiently.
Compared with analysis conducted solely by general generative AI tools, Fujitsu’s proprietary solution enhances the quality of generated documentation.
This system links large volumes of source code through a Knowledge Graph — Enhanced retrieval system.
With this, it prevents omissions and hallucinations, ensuring the generated design documents reflect all relevant system details.
This approach improves comprehensiveness by 95% and boosts readability by 60%, producing documentation that is easier for teams to interpret and act upon.
These improvements are particularly critical when dealing with COBOL, a programming language designed by Dr. Grace Hopper in 1959 that has an estimated 850 billion lines of code.
Although over 65 years old, COBOL remains widely used, and businesses remain heavily reliant on this venerable programming system.
It often powers the core transaction systems of banks, insurers, and government agencies without public visibility.
During the pandemic, the United States raised the prospect of a shortage of COBOL programmers to help manage critical systems, showing its importance.
Before the Fujitsu system, Anthropic noted that AI could help keep COBOL running for a long time.
Fujitsu plans to provide support services to guide organizations in deploying the system effectively – this year, the company intends to introduce features for rebuilding existing source code for future use.
The system will automatically rewrite code and support ongoing operation and maintenance.
This sequential development ensures that companies can not only understand legacy systems but also adapt and maintain them with minimal manual intervention.
“We see this announcement as an initiative that realistically advances the modernization of our legacy systems… we have come to recognize the potential of this technology,” said Toshihiro Horiuchi, Managing Executive Officer, SMBC Nikko Securities Inc.
For many engineering teams, this announcement represents a clear advantage, but IBM, a major vendor of COBOL-powered mainframes and enterprise systems, needs to revise its strategy or risk being outpaced in automated legacy system modernization.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
Fujitsu automates COBOL documentation, cutting analysis time by 97% The system generates design documents without requiring deep programming expertise Knowledge Graph retrieval reduces hallucinations and improves documentation completeness Fujitsu has introduced a new generative AI service which analyzes COBOL and other legacy source code, automatically producing design documents in minutes…
