Two months after Anthropic‘s Code Modernization Playbook sent IBM stock to its worst single-day loss in more than 25 years, IBM is telling financial institutions that the predicted COBOL exodus is not materializing. 

“Firms aren’t ditching COBOL,” Skyla Loomis, general manager of IBM Z Software, told FinAi News. Emerging technologies and AI actually “lower the skill barriers to working with COBOL.”  

In fact, IBM Z’s mainframe revenue grew by 51% year over year in the first quarter, contributing to total revenue of $15.9 billion, up 9% YoY, according to the company’s earnings report on April 22. 

Software revenue rose by 11% YoY to $7.05 billion, infrastructure revenue rose by 15% to $3.33 billion and the company’s generative AI book has reached about $25 billion, up by 10% YoY, according to the earnings presentation. 

“AI strengthens the mainframe case, it does not weaken it,” IBM Senior Vice President of Software and Chief Commercial Officer Rob Thomas wrote on LinkedIn following the earnings release. 

That message is central to IBM’s pitch to FI clients evaluating modernization strategies in the wake of Anthropic’s Feb. 23 announcement, which positioned Claude Code as a tool to automate COBOL exploration and migrate workloads off the mainframe. IBM shares fell 13.2% on the day of the announcement. 

Co-operative approach

Rather than positioning its watsonx Code Assistant for Z directly against Anthropic’s Claude, IBM incorporated Claude into its enterprise development framework, IBM The platform does not rely on a single AI model, but uses IBM’s Granite models, Anthropic’s Claude, Mistral AI and Meta‘s Llama, according to the company. 

“IBM Bob was built to give enterprises flexibility and choice, rather than locking them into a single model,” Loomis told FinAi News. “Different LLMs have different strengths, and no single model is optimal for every task across the software development lifecycle.”  

Selective modernization

However, mainframe modernization requires more than language translation. 

“IBM Z applications are among the most complex and mission-critical systems in an enterprise, often spanning millions of lines of code and requiring deep expertise in mainframe languages, middleware and runtime behavior,” she said. “Other AI solutions often lack systemwide awareness across the mainframe application landscape, which means the outputs from them won’t have the fidelity needed to accelerate modernization.” 

The path most FI clients are employing is selective augmentation rather than wholesale migration, Loomis said.  

“Some clients might take a selective modernization approach, retaining COBOL for core transaction processing where it delivers maximum value, while refactoring or exposing specific capabilities through APIs, often using Java, to increase flexibility and integration,” she said. “For those clients that are choosing to augment with Java, the platform itself offers the non-functional attributes necessary for success across every hosted language.” 

IBM’s watsonx Code Assistant for Z, which helps create new or updates existing COBOL code, is used by about 200 enterprises, according to the company. 

Royal Bank of Canada, for example, used watsonx Code Assistant for Z to “proactively identify dependencies, data flows, structure and organization of existing applications,” Loomis said, building “an in-depth blueprint for the modernization and management of changes to core system applications.” 

Egypt’s National Organization for Social Insurance and ANZ Bank are also tapping the platform. 

National Organization for Social Insurance “observed up to a 94% reduction in time to analyze and locate superfluous COBOL code and programmatic routines, reducing the identification time from approximately 8 hours to nearer to 30 minutes,” Loomis said.  

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