It’s a little-known fact that many of the world’s financial and government systems still rely heavily on ancient code. This legacy software is often confusing, hard to understand, and extremely costly to replace. Now, artificial intelligence is starting to change that, offering a shortcut to modernize these old systems. The market responded strongly to this shift, especially after a notable AI startup claimed it could speed up COBOL modernization, which has traditionally been a slow and expensive process.

Why COBOL Still Matters More Than People Think

Many don’t realize just how much COBOL code is still in use today. Hundreds of billions of lines of this old language run critical systems daily, especially in finance and government sectors. For example, in the US alone, COBOL handles about 95% of ATM transactions. Despite its importance, the real challenge isn’t the code itself but finding people who understand it. Most of the original developers have retired, and the pool of COBOL experts is shrinking fast.

This talent shortage has made modernizing COBOL systems a costly and time-consuming task. Big consulting firms like IBM, Accenture, and Cognizant built profitable practices around these projects, often taking years and millions of dollars. It’s a complex process that involves mapping workflows, understanding dependencies, and rewriting large parts of the system without disrupting daily operations.

AI Offers a New Approach to Legacy Code

Recently, AI startups like Anthropic have announced tools that could change the game. Their Claude Code software can analyze thousands of lines of COBOL code, identify dependencies, document workflows, and spot risks faster than human analysts. This means teams can now get a detailed understanding of their systems much quicker, paving the way for faster modernization.

Anthropic claims that with tools like Claude Code, companies can modernize COBOL systems in a matter of months instead of years. This is a huge shift, as it could significantly cut costs and reduce the time needed to upgrade critical legacy systems. The technology works by automating parts of the process that used to require large teams of consultants, making modernization more accessible and less expensive.

It’s worth noting that IBM has been advocating similar ideas for years. About three years ago, IBM suggested using AI to rewrite COBOL code into Java and developed a product called “watsonx Code Assistant for Z” to help with this. Recently, IBM’s CEO confirmed that many of their mainframe customers are already using AI tools to understand and decide what to modernize. IBM argues that their systems can deliver the same performance and security regardless of the programming language used, emphasizing that modernization doesn’t mean sacrificing reliability.

What’s clear is that AI is starting to reshape how legacy systems are handled. While the market reacted strongly to Anthropic’s announcement, IBM’s long-standing position shows that this is not a sudden shift but an ongoing evolution. The real question now is how quickly organizations will adopt these new AI tools and what impact that will have on the future of enterprise IT.

Inspired by

• https://www.artificialintelligence-news.com/news/cobol-modernization-ai-claude-ibm/

Sources

• bloomberg.com
• claude.com
• pcmag.com
• theregister.com
• ibm.com