The AI Architect: Fujitsu Kozuchi’s Blueprint for Modernizing Legacy COBOL
Dillip Chowdary
March 30, 2026 • 12 min read
Fujitsu’s "Application Transform SaaS" is leveraging specialized GenAI to automate COBOL analysis and design documentation with 97% efficiency, addressing a multi-trillion dollar documentation debt crisis.
The global financial and governmental infrastructure still runs on hundreds of billions of lines of **COBOL**—a language born in 1959. The primary obstacle to modernizing these mainframe systems isn't just the difficulty of writing new code; it's the total lack of documentation for the old code. Fujitsu has stepped into this vacuum with its **Kozuchi AI platform**, launching a new service that can automate up to **97% of the analysis phase** of modernization projects.
Solving the Documentation Debt
For many enterprises, the engineers who wrote their core banking systems retired decades ago. What remains is "spaghetti code" that is too risky to touch. Fujitsu’s **Application Transform SaaS** uses Generative AI to reverse-engineer this logic. Instead of a general-purpose LLM, Fujitsu uses a specialized model trained on millions of lines of proprietary and open-source mainframe code, capable of understanding the nuances of **JCL (Job Control Language)** and complex data structures.
The system doesn't just translate code; it creates **Functional Design Documents** that explain *what* the business logic is doing, not just *how* it's written. This transition from "code-to-code" to "code-to-design" is the breakthrough required for safe migrations to cloud-native microservices.
Technical Breakdown: The AST Approach
The core of Kozuchi's capability lies in its use of **Abstract Syntax Trees (AST)**. By parsing COBOL source code into a structured tree, the AI can trace data lineage and identify dependencies that traditional text-based analysis would miss. The GenAI layer then synthesizes these structures into human-readable documentation, including data flow diagrams and interface specifications.
Fujitsu claims this reduces the time required for the initial "discovery" phase of a project from months to days. This is critical for the "2025 Cliff"—a term used to describe the aging out of the last remaining COBOL experts.
Modernize Your Architecture Documentation with ByteNotes
Whether you're migrating from COBOL or building a new AI platform, **ByteNotes** is your central source of truth. Manage your design docs, API specs, and migration roadmaps in a secure, unified workspace.
The Path to Cloud-Native
Once the design documentation is generated, Fujitsu's platform provides a roadmap for refactoring. The AI identifies candidate functions for **microservices encapsulation**, allowing organizations to move to the cloud incrementally rather than attempting a risky "big bang" migration. This "AI-augmented refactoring" ensures that business rules are preserved while the underlying infrastructure is modernized to support 2026-era performance requirements.
Conclusion: Preservation Through Intelligence
Fujitsu’s Kozuchi is more than just a developer tool; it is a cultural preservation engine for the digital age. By translating the arcane logic of the 20th century into the documented architecture of the 21st, Fujitsu is ensuring that the world's most critical systems remain resilient. For enterprise architects, the message is clear: the legacy bottleneck is finally opening, and GenAI is the key.