The world depends on a 60-year-old code that no one knows about anymore

The world processes $3 trillion in financial transactions every day using the 64-year-old COBOL programming language, which almost no one studies anymore. Although most educational institutions stopped teaching it decades ago, COBOL remains one of the leading mainframe programming languages, widely used in banking, automotive, insurance, government, healthcare, and finance.

According to the International Journal of Advanced Research in Science, Communication and Technology, 43% of all banking systems still use COBOL, which processes $3 trillion of these transactions per day, including 95% of all US ATM transactions and 80% of all credit card transactions.

The main problem is that few people are interested in learning COBOL these days. Programming in it is considered cumbersome and inefficient, and the code reads like an English lesson. Moreover, the encoding format requires care and is not flexible, and compilation takes much longer compared to competitors. Because of this, people who know how to work with and support COBOL are becoming increasingly rare.

IBM offers a solution to this problem using AI. The company has developed an AI assistant for coding ( watsonx ), which helps translate older COBOL code into more modern languages, thereby saving programmers many hours of reprogramming. This process can be simplified to converting English essays to Esperanto using ChatGPT. It allows programmers to take a piece of COBOL code and convert it to Java using watsonx. However, in practice everything is not so simple.

Skyla Loomis, vice president of IBM Z software at IBM, says, “AI can deliver up to 80 to 90 percent of the required output, but additional adjustments are required. It is a means of increasing productivity, not a complete replacement for developer work.”

According to report Gartner 2023, by 2028, a combination of human and AI assistants could reduce the time to complete coding tasks by 30%, and 80% of programmers will use AI in some form.

Now, says Gartner analyst Arun Chandrakekara, all we can do is “wait and see.”