AI is driving the charge for tech stack and coding language modernization. 

Created in the 1950s, Common Business Oriented Language (COBOL) became the bedrock coding language for many industries through the 1980s, including financial and health care organizations. 

COBOL still processes $3 trillion a day worldwide, nearly 95% of ATM systems are built on it and it supports more than 1.5 million transactions per second with an uptime of 99%, Jeana Bolanos, president of AI implementation and consulting firm Sales E, previously told FinAi News. 

Few software developers are well-versed with COBOL and many who know the language are nearing retirement, Paul Holland, chief technology of Astadia, a subsidiary of tech company Amdocs, told FinAi News.  

He said the financial services industry needs to pivot away from the legacy language before all COBOL developers are gone. 

As FIs look to mount the tedious task of modernizing billions of lines of code, AI stands out as way to help the transition. 

“We are in a period of enormous change in the whole concept of mainframe application modernization, all driven through AI, particularly agentic AI,” Holland said. “It’s very exciting, but it’s also relatively unproven.” 

Banks declutter, insurers exit 

Amdocs has helped more than 300 companies exit the mainframe infrastructure entirely using deterministic conversion tools that translate COBOL line for line into Java or C#, Holland said. Societe Generale and Deutsche Bank are among Amdocs clients. 

Amdocs is reporting 25% to 30% cost reductions on projects that add agentic AI into workflows intended to modernize code, he said. Mapping infrastructures is where AI is the biggest help to banks as of now, he added. 

But banks and insurers are taking different approaches, he said. 

Large financial institutions are peeling off peripheral workloads while keeping core transaction processing on the mainframe, Holland said. 

Insurance companies, by contrast, are more willing to pursue full exits, driven by the high cost of running closed books of legacy policies on expensive mainframe infrastructure. 

“Banks are not in a hurry to get rid of the mainframe,” Holland said. “What they’re looking at is, ‘Can I transition some of these workloads off, lower my mainframe footprint, maybe still keep core processing?’” 

Sunwest Bank, for one, uses COBOL indirectly through core banking vendor ecosystems, Chief Technology and Strategy Officer Ben Xiang previously told FinAi News.  

Although Sunwest is in talks with its core banking provider FIS to modernize COBOL, it is still evaluating what language to use. 

The reasoning behind keeping core capabilities is twofold, Holland said. One is the deterministic reliability of COBOL and the other is the ability to modernize at scale. 

The scale problem 

The central challenge, executives say, is that no one has used AI to modernize a mainframe at the scale most large banks operate, he said. 

Can AI modernize “20 million lines of code and 5,000 programs on a mainframe? Not today,” Chad Jones, chief revenue officer of Astadia mainframe practice, told FinAi News. 

The AI-driven processes can compress what would have been a five- to 10-year redevelopment project to roughly two to three years, but at smaller scope — for now. 

Register here for the FinAi Lending Summit, set for Oct. 7-8 in Las Vegas.