AI-generated image for representation
COBOL may be decades old, but it still powers critical financial infrastructure and 95% of ATM transactions, according to IBM. Yet 66 years after its creation, financial institutions, businesses, and governments still use it widely. COBOL introduced object-oriented programming in 2002, while IBM has continued developing products around it. A new standard was also released in 2023. Computerworld highlighted the ageing COBOL workforce in 2008 and again in 2012.
Because it had become “uncool, in 2013, ZDNet reported that universities were avoiding teaching COBOL. Because financial systems can contain hundreds of millions of lines of code, moving away from COBOL is difficult. Now, as the engineers who maintain these systems age out and fewer young developers learn the language, the skills needed to keep this infrastructure running could become scarce. This could also create problems for data centers supporting financial services, where legacy systems still run on COBOL. Banks and businesses may struggle to maintain systems they can’t easily replace, and the shortage of skilled engineers has existed for years. It will only get worse as experienced COBOL programmers retire. COBOL was never designed to become such a deeply embedded part of modern infrastructure. The language has evolved over time. But modernising the language does not solve the shortage of people who can maintain decades-old systems. The problem is not new. Simply asking AI to translate that code into a newer language may not be enough.
In 2018, a major failure occurred when UK bank TSB shifted from one platform to another. Unable to access their accounts, the bank was then fined £48.65 million ($65.86 million) and affected millions of customers. Some jobs require several years of experience with the language and can pay between $100,000 and $150,000, with some positions offering more.
For banks and other organisations, the challenge may therefore be less about whether COBOL is old and more about whether enough people will remain who understand the systems built around it.
Calvin Smith, in a breakdown by YouTube channel OpenHands, said attempts to use AI agents for such work have not been particularly successful. IBM also pushed back against the idea that modernising COBOL systems is simply a matter of translating code, saying it involves “data architecture redesign, runtime replacement, transaction processing integrity, and hardware-accelerated performance. That complexity makes any transition, gradual or not, risky. The shrinking pool of COBOL engineers is also creating financial incentives for those with the expertise. The bigger question is what happens when even those engineers begin to disappear. Replacing decades of carefully maintained financial infrastructure with newer languages such as Rust would be expensive and carry significant risks. You use AI every day. Now get your AI Quotient. Take the AIQ test.