The AI Race to Modernize Mainframes Has BegunThe AI Race to Modernize Mainframes Has BegunThe AI Race to Modernize Mainframes Has Begun

By Omer Rosenbaum, Swimm

When COBOL was created in 1959, it was revolutionary — offering a standardized computer language perfectly suited for business applications. Fast-forward over six decades, and while the world has been to the moon and back (literally), COBOL remains firmly grounded as the backbone of global financial systems.

Every day, this seemingly outdated language powers 3ドル trillion in financial transactions and runs 95% of all ATM operations. COBOL has stood the test of time, seamlessly handling the critical operations of organizations like banks, insurers, and governments.

But even the toughest legacies eventually show their cracks; today, COBOL is a technological bottleneck for enterprises undergoing digital transformation .

The Legacy Dilemma

Enterprises that rely on COBOL systems face a harsh reality. The language might be durable, but it exists in a world unrecognizable from the one in which it was born. Modern businesses demand agility, integration, and innovation — all of which are hindered by mainframe applications that haven't evolved in decades.

Why hasn't newer coding replaced COBOL? Enterprises operating on mainframes have massive sunk costs in these systems, ones that reliably run payroll, credit card processing, airline bookings — the list goes on. Many organizations are carrying out mainframe modernization projects, but modernizing all mainframe systems would have an astronomical cost, not just financially but also operationally. Accenture estimates the cost of fixing accumulated technical debt in the U.S. alone could reach 1ドル.52 trillion.

Related:How to Eliminate Software Development Bottlenecks

It's a catch-22.

While CIOs are funneling up to 20% of their budgets into maintaining legacy systems, those same systems are holding them back from true modernization.

Enter AI (but There's a Catch)

Artificial intelligence revolutionizes everything it touches, and programmers might feel it the most. According to Google, AI already writes 25% of all code , enabling developers to significantly accelerate time-to-market. AI-powered coding assistants like GitHub Copilot have become invaluable tools for modern languages such as Python, Java, and JavaScript.

There's just one small problem. Ask AI to code in COBOL, and its performance plummets.

Why? AI thrives on vast datasets, and modern programming languages dominate the internet, with an abundance of documentation, sample code, and active online communities. COBOL, meanwhile, exists largely outside the digital space. Its documentation is tucked away in decades-old manuals or proprietary archives, and real-world COBOL code — hidden within mainframes — rarely makes its way online.

Related:Who Pays for Open Source Software? Uncovering Hidden Costs and Funding Models

This knowledge gap creates a significant barrier for AI systems attempting to interact with or generate COBOL code. Modern enterprises may wish to leverage AI to help refactor or modernize COBOL, but the technology simply isn't there yet, leaving businesses scrambling for solutions.

The Stakes Are Really High

For those tasked with modernization, resolving the COBOL bottleneck isn't just a technical consideration — it could shape the future of their organizations. Companies stuck in reactive maintenance cycles on legacy systems risk falling behind in their industries, burdened by spiraling technical debt .

But the stakes go beyond individual enterprises. These mainframe systems underpin global economic stability, from financial transactions to governmental operations. Modernizing them isn't just a business imperative — it's an economic priority.

The AI vs. COBOL Revolution

Recognizing the scale of the problem, some of the world's biggest tech players are stepping into the ring. Amazon, Google, and Microsoft have launched initiatives aimed at AI-driven modernization of mainframe systems.

Startups are also entering the fray, offering specialized solutions to extract, document, and refactor COBOL using new AI techniques. Some companies are developing platforms that "translate" COBOL into modern languages like Java or Python , bridging the gap between legacy and contemporary systems. Others are creating tools to help explain "ancient" COBOL code, significantly reducing the months of research required for modernization initiatives while simultaneously preserving technical knowledge for the next generation of engineers and developers.

Related:Citizen Development: The Wrong Strategy for the Right Problem

Cracking the COBOL Barrier

The big question is clear: Will AI win the race to finally make COBOL systems accessible, efficient, and modern? The stakes couldn't be higher. Solving the COBOL conundrum could unlock billions — if not trillions — of dollars in global enterprise potential.

This race offers both risks and opportunities. Those who adopt forward-thinking modernization strategies (and learn to partner with AI effectively) are poised to lead their industries into the future. Meanwhile, those who cling to outdated practices could find themselves stuck in a never-ending cycle of technical debt.

The message is simple: Legacy systems don't have to be a hurdle. With AI on the rise, the tools to modernize and transform are closer than you think. The winners of this race won't just update their systems, they'll define the future of enterprise technology.

So, What's Next?

If your organization is grappling with legacy applications written in COBOL, now is the time to act. Explore emerging AI-driven solutions, assess the feasibility of modernization within your specific context, and begin prioritizing systems that deliver the highest ROI when updated.

The COBOL bottleneck is one of today's most pressing IT challenges. While solving it requires effort, the payoff — greater efficiency, agility, and innovation — is more than worth it.

About the author:

Omer Rosenbaum isCTO and co-founder of Swimm , an Israeli startup that raised 33ドル.3 million to build the next generation of AI-driven software documentation tools. Its product is used by 12,000 teams to automatically document 300 million lines of code. Omer previously founded the Check Point Security Academy and was Deputy Commander of the IDF Cybersecurity Research Team.

• ITPro Today’s 2024 State of DevOps Report

  Dec 16, 2024
  |

  2 Min Read

• BCDR Basics: A Quick Reference Guide for Business Continuity & Disaster Recovery

  Oct 10, 2024
  |

  1 Min Read

• ITPro Today’s 2024 IT Priorities Report

  Sep 25, 2024
  |

  1 Min Read

• Tech Careers: Quick Reference Guide to IT Job Titles

  Sep 13, 2024
  |

  1 Min Read