Modernizing their tech stacks and COBOL coding language is a major task for financial institutions. 

Created in the 1950s, Common Business Oriented Language —  COBOL — became the bedrock coding language for many industries in the right through the 1980s, Jeana Bolanos, chief executive of AI implementation and consulting firm Sales E, told FinAi News. 

The language processes $3 trillion a day for companies globally. Nearly 95% of ATMs are built with it, and it supports more than 1.5 million transactions per second with an uptime of 99%, Bolanos said.  

“These are systems that we can’t afford to have go down,” she said.

Unlike other coding languages, COBOL is highly reliable and the main reason it hasn’t been replaced until now, and the language is “highly deterministic,” Bolanos added. 

“When you’re talking about financial institutions, you have to know that when you input ‘A,’ that you’re going to get output ‘1’ every single time without variance,” she said. “Companies can migrate away from COBOL to other [enterprise resource planning] systems, but they might not get the same 99% uptime, which can be costly.”

To replace COBOL, companies and banks must migrate to a language that is equally as deterministic, she said. 

Where does AI fit in? 

While AI systems cannot replace COBOL systems, they can help financial institutions map their systems visually. 

Major AI businesses like Anthropic are aiding companies in the discovery process, Bolanos said, noting that AI can make discovery drastically faster. 

“It’s not that you couldn’t have done these things before, it’s that it would have taken years to do this,” Bolanos said. “The tedious tasks,” including writing the test, compiling information and mapping data can be done quickly now with AI. 

A CTO perspective 

Like most FIs, Sunwest Bank interacts with COBOL indirectly through core banking vendor ecosystems, Chief Technology Officer Ben Xiang told FinAi News. 

“The discovery and documentation phase of any legacy modernization is where the real time gets burned,” Xiang said. “AI tools that can map dependencies, surface undocumented logic and identify risk areas faster than manual review are genuinely useful.” 

The best use of AI for financial institutions is “analysis and planning” and the tech “wouldn’t let them generate production code for regulated systems without significant human validation and phased testing,” Xiang said. 

If the financial services industry were to move away from COBOL, the $3.7 billion bank said that they might use Java and C# or .NET for transaction-oriented workloads and Python for batch and analytics, he said. 

“The choice depends on runtime requirements and team capabilities more than language preference,” Xiang said. 

Register here for the inaugural FinAi Banking Summit, taking place March 2-3 in Denver. View the full event agenda here.