Fujitsu Application Transform powered by Fujitsu Kozuchi cuts design document generation time by approximately 97% for COBOL systems.[1]

Why legacy COBOL code analysis needs urgent answers

Modernization projects routinely grind to a halt over one hurdle: understanding what the decades-old COBOL code actually does. According to TechRadar, Fujitsu’s AI service reduces dependence on specialist programmers by nearly erasing time spent deciphering legacy systems—slashing it by about 97%.[2] With financial, insurance, and public sector operations running on aging COBOL, this isn’t a theoretical problem. It defines budgets, risks, and even careers.

97%
reduction in design doc generation time (Fujitsu, 2026)

Fujitsu Application Transform is rewriting COBOL documentation workflows

Fujitsu Application Transform, powered by Fujitsu Kozuchi and available as a SaaS in Japan since March 30, 2026, is designed specifically for analyzing COBOL source code and generating design documents automatically.[1] The service applies Fujitsu’s proprietary Knowledge Graph-enhanced RAG technology to link large volumes of source code, preventing omissions and hallucinations. Most organizations struggle to keep design docs current, but this service generates high-quality documentation in minutes, not hours. That 97% time reduction isn’t just a productivity boast—it finally makes large-scale COBOL documentation feasible without armies of experts.

💡
Pro Tip: Don’t expect full automation. AI can generate, but strategic planning and critical review still need human expertise.

Documentation is 95% more comprehensive than with general AI

The data shows that compared to generic generative AI, Fujitsu’s approach increases documentation comprehensiveness by 95% and maintains consistent design information without omissions—even in complex COBOL projects.[1][2] That means fewer missed dependencies and less time spent hunting for hidden business logic. For anyone who’s reviewed legacy documentation, “consistent and comprehensive” sound like science fiction. Here, it’s backed by the numbers. If you’re responsible for migration or audit readiness, this level of detail changes the entire risk equation.

⚠️
Common Mistake: Assuming this service works equally well for modern languages. It’s optimized for COBOL and other legacy stacks.

Readability of design documents is up by 60% vs. old methods

Most people get this wrong: speed is useless if your documentation is indecipherable. Fujitsu’s AI service improves design document readability by 60% compared to conventional approaches.[1] Suddenly, design docs aren’t a cryptic wall of jargon—they’re actually usable for onboarding, code reviews, and modernization planning. This is what actually works. Not the fluffy advice you see everywhere about “collaboration tools” and “best practices.” If you’ve ever slogged through legacy specs, you’ll know a 60% boost is a game changer for both developers and auditors.

60%
improvement in doc readability (Fujitsu, 2026)

The service targets legacy-heavy sectors: banking, insurance, and government

Fujitsu’s AI service is aimed at organizations with entrenched COBOL systems—banking, insurance, and the public sector.[2] These industries face mounting pressure to modernize, but risk grinding to a halt when faced with missing or inconsistent documentation. The AI-generated documents enable a clear understanding of system specifications, supporting robust modernization and migration strategies.[1] You’ll notice the pattern: where documentation bottlenecks used to paralyze projects, now there’s a viable escape route—even if you’re dealing with 30-year-old code.

Human expertise still matters—even with 97% automation

The data shows automation is real, but not total. While Fujitsu Application Transform drastically reduces the need for COBOL expertise in documentation, it does not eliminate the need for human experts entirely.[1] Strategic decisions, exception handling, and system maintenance still require deep knowledge. Here’s the thing nobody tells you: trusting AI to fully replace skilled engineers is wishful thinking. Instead, treat automation as a force multiplier that lets your best people focus on actual modernization, not rote documentation.

“We see this announcement as an initiative that realistically advances the modernization of our legacy systems by combining Fujitsu’s deep system development expertise built up over many years with generative AI.” — Toshihiro Horiuchi, Managing Executive Officer, SMBC Nikko Securities Inc.[1]

How Fujitsu Application Transform and Kozuchi compare

ToolCOBOL SupportSaaS AvailabilityCore Feature
Fujitsu Application TransformYesJapan (2026)Design document automation (97% faster)
Fujitsu KozuchiYesJapan (2026)AI-powered code analysis
Fujitsu Knowledge Graph–Enhanced RAGYesIntegratedPrevents omissions/hallucinations

FAQ: Fujitsu AI Solutions for COBOL Code Analysis

How much faster is Fujitsu Application Transform at generating design documents for COBOL?
Fujitsu Application Transform reduces design document generation time by approximately 97% compared to traditional methods, even without expert knowledge.[1]
Does Fujitsu's AI service work for programming languages other than COBOL?
The service is specifically optimized for COBOL and other legacy programming languages. Its effectiveness may be limited for modern languages.[2]
Will Fujitsu's solution eliminate the need for COBOL experts?
While the solution reduces the need for COBOL expertise in documentation, human experts are still needed for maintenance, modernization planning, and complex problem-solving.[2]
How does the comprehensiveness of documentation compare to general AI tools?
Fujitsu’s service produces documentation that is 95% more comprehensive than that generated by general-purpose generative AI, especially for complex COBOL code.[1][2]

Where this all points for 2026

Legacy modernization isn’t a buzzword anymore. With design documentation bottlenecks falling by 97%, and readability up 60%, there’s finally breathing room for organizations buried in COBOL. But the real change comes from the combination—the right AI paired with focused human expertise. If you’re waiting for a world where AI handles every edge case, you’ll wait forever. But for everything else? The future just arrived in Japan.

Sources

  1. global.fujitsu/en-global/pr/news/2026/03/30-01
  2. techradar.com/pro/ibm-watch-out-fujitsu-uses-ai-to-understand-cobol-and-automatically…