Large language models capable of writing and refactoring computer code have proliferated over the past year. These tools don't really replace programmers -- writing code is as much art as science. But they can make programmers dramatically more efficient by handling tedious tasks.
International Business Machines (IBM -1.05%) has been rolling out its watsonx artificial intelligence (AI) platform over the past few months. The company is focused on helping its enterprise customers train and deploy generative AI models while keeping data privacy and regulatory requirements in mind. On Thursday, IBM officially launched watsonx Code Assistant. This new generative AI-powered assistant is not a general-purpose tool. Instead, it's tailored for very specific use cases.
One of those use cases is converting decades-old code that runs on IBM's mainframe systems to a more modern programming language. This may seem like a boring application of such a transformative technology, but it represents an enormous opportunity for IBM to keep its mainframe systems relevant and generate additional revenue from its enterprise customer base.
From COBOL to Java
COBOL, or Common Business Oriented Language, is a programming language that has been around since 1959. COBOL isn't commonly used to write new applications, given the preponderance of modern programming languages available. However, it's still prolific in certain industries. IBM estimates that 70% of global banking transactions are processed on mainframe systems running COBOL. Another estimate puts the total number of lines of COBOL code in daily use at around 800 billion.
The problem is that no one learns COBOL anymore. It's not widely taught in U.S. universities, and even if it was, learning COBOL is likely viewed as a bad career move. This puts companies running COBOL-powered workloads in a bind. Rewriting decades-old COBOL code requires specialized talent that's hard to find, and it creates a risk of disrupting mission-critical systems. As time goes on, the talent pool of programmers capable of modernizing COBOL applications dwindles.
IBM has a solution with watsonx Code Assistant for Z. This specialized version of its code assistant uses generative AI to map out a COBOL application and its dependencies, split up a COBOL application into modular parts, and convert those parts into modern Java code. The product doesn't yet support validation testing, but that feature is planned for a future release.
This new AI tool from IBM accomplishes multiple things. First, it gives mainframe customers a path to modernize their applications without abandoning the mainframe as a platform. Second, it allows IBM to expand its relationship with mainframe customers. And third, it will provide a boost to IBM's consulting business. IBM's consulting arm will help customers identify applications ripe for modernization, and it will work alongside other parts of the company to build custom solutions for clients with specialized needs.
One big benefit for clients is data privacy. Some of these COBOL-powered applications are at the heart of sensitive and critical systems in the banking and financial services industry. By partnering with IBM and using its watsonx platform, clients can ensure their data remains secure.
An enterprise-first AI strategy
Much like its hybrid cloud strategy, IBM's AI strategy centers around helping clients boost productivity, become more efficient, and lower costs. The cost in time, money, and potential for disruption to manually convert legacy COBOL applications to a modern language is likely too high for many mainframe customers to seriously consider taking that path. With this new AI tool, mainframe customers have another option for modernizing mission-critical applications.
IBM's mainframe business is resilient, but the lack of COBOL developers has become a thorny problem for clients. This new AI tool will help clients bring their mainframe applications into the 21st century, strengthening IBM's mainframe business in the process.
