- AI automates COBOL code exploration, maps dependencies and analyzes structural risks quickly
- Engineers can effectively prioritize modernization based on technical risk and business value
- Automated tests confirm that migrated COBOL components produce identical output to legacy systems
Modernizing legacy COBOL systems has long been an expensive and labor-intensive process requiring extensive human effort, as teams of consultants traditionally spent months or even years mapping workflows, documenting dependencies, and unraveling decades of accumulated business logic.
Hundreds of billions of lines of COBOL are still in production worldwide powering critical systems in banking, government and airlines, yet it has become increasingly difficult to find developers with the knowledge to interpret these systems.
Now, however, Anthropic is looking to replace this, with its Claude AI platform, which aims to take much of the heavy lifting away from human workloads.
How AI helps explore and analyze code
This scarcity of expertise has historically slowed modernization projects and increased costs – however, Anthropic now believes that AI can automate much of the exploration phase that once consumed most human effort.
“Modernizing a COBOL system once required armies of consultants who spent years mapping out workflows… AI is changing this,” the company said in a blog post.
Tools like Claude Code can map dependencies across thousands of COBOL lines, trace data flows between modules, and document workflows that current employees no longer actively remember.
These automated processes identify risks, isolate tightly coupled components, and flag duplicate or potentially fragile code.
By analyzing these structural and functional relationships, AI can prioritize which components should be modernized first based on technical risk, business value, and organizational priorities.
The best programming laptops allow engineers to efficiently integrate AI output while maintaining oversight of the modernization plan, and once components are prioritized, AI can generate preliminary functional tests to verify that migrated code produces identical outputs to the legacy system.
Human teams then decide whether these automated tests are sufficient, which scenarios require manual verification, and which performance benchmarks to maintain.
Implementation continues incrementally, with each module tested and validated before further changes are made.
AI tools can translate COBOL logic into modern languages, create API wrappers around legacy components, and build scaffolding that allows old and new code to work side by side.
This reduces the risk of large-scale failures and enables organizations to move forward with complex modernization projects.
AI also provides detailed insight into potential technical debt, isolated modules and high-risk areas, allowing teams to plan modernization strategically – as engineers can review these recommendations and sequence work to align with regulatory requirements, business priorities and operational constraints.
Automated documentation and analysis gives teams comprehensive situational awareness, but final decisions still rely on human judgment.
While this is a big win for many engineering teams, IBM, a major supplier of COBOL-powered mainframes and enterprise systems, will not be pleased.
The company saw its stock drop sharply after Anthropic announced that Claude Code could automate much of the labor-intensive modernization process.
AI’s ability to replace work traditionally done by human consultants threatens parts of IBM’s business model.
This shows that even well-established enterprise software vendors can face disruption as AI continues to reshape legacy system modernization.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, video unboxings, and get regular updates from us on WhatsApp also.
