Reaching New Frontiers: Modernizing Mainframe Environments with Generative AI
IBM’s Rich Larin on how generative AI solutions address the need for efficiency in mainframe application modernization and fill in growing skills gaps
By John Morell
Reaching New Frontiers: Modernizing Mainframe Environments with Generative AI
IBM’s Rich Larin on how generative AI solutions address the need for efficiency in mainframe application modernization and fill in growing skills gaps
By John Morell
Mainframes have long been the staple of critical business and government operations, handling everything from large-scale transaction processing to managing statistical data. Flashier computing advancements get more attention, but mainframes continue to be the heart of many industries. Although they’re older technology, advancements have helped mainframes stay relevant. Case in point: generative AI.
With most of the mainframe world moving to hybrid cloud strategies, generative AI promises to be the link that modernizes mainframe applications. However, there is some hesitation. The fact that mainframes are used in mission-critical environments such as banking, finance, healthcare and government means there’s a need to be cautious about using AI tools in that space. But the overall advantages of generative AI to modernize applications show that, when carefully applied, this can be a transformative tool to improve mainframe management.
AI, with its translation skill, is able to show us that a more modern programming language makes sense in mainframe environments.
3 Big Challenges With Mainframe Apps
Most mainframe operators know the value of keeping their hardware current, with many using the latest Z system. But while these operators ensure their hardware is cutting edge, they may not be as diligent about modernizing their applications. “If they're agile, some will update their mainframe applications each quarter. But most will only do so yearly,” says Rich Larin, product lead for IBM watsonx Code Assistant for Z. “It leaves many challenges for using some of these decades-old applications.” Here are three modernization challenges in legacy mainframe apps that can benefit from generative AI:
Slow and cumbersome development
Agility is the key to modern cloud-native DevOps. Gen AI helps by pushing multiple commits per day on a mainframe, helping to decompose large complex applications into smaller, manageable chunks. As a result, a giant mainframe app problem becomes easier to handle.
Lack of mainframe skills
In today’s environment it often feels that the mainframe workforce is a rapidly shrinking talent pool. And when they are available, these professionals command a premium for their time. The advantage of generative AI is that it brings language translation to the talent equation.
There are a multitude of Java developers available around the world, but not as many are fluent in COBOL, which can make it difficult to find skilled mainframe developers. But now, generative AI can be used to translate code from one language to another—for example, from COBOL to modern languages like Java. “Use of Java on a mainframe has always been possible, but difficult,” says Larin. “AI, with its translation skill, is able to show that a more modern programming language makes sense in mainframe environments.”
Generative AI tools are creating an expanded, efficient and productive mainframe workforce, and the developer talent pool has increased exponentially in size. It makes millions of enterprise Java developers capable of working on mainframe applications. The paradigm for talent has shifted.
Developing a plan
Sometimes app modernization efforts are stymied by not having a clear plan of steps to follow. Using gen AI can help by automatically rewriting and translating applications from an older language to one that’s more accessible. If users want to keep some of these apps in COBOL, they have that ability. Gen AI is a tool that helps you quickly determine the needs of your applications and show you what works best for your particular situation. “It's allowed us to pay more attention to mainframe application modernization than we've had to in the past,” says Larin.
The Advantages of Generative AI
Besides its language translation abilities, generative AI can be used in various ways to improve the mainframe experience. This includes creating sample data for testing processes in order to protect an organization's genuine data; finding anomalies in mainframe logs to identify unusual patterns and security threats and analyzing historical data to predict potential failures of hardware and software.
Making an informed decision on a product as important as generative AI for mainframe applications requires full proof of value or proof of concept testing to get a sense of how the solution would work on a particular system.
One of the advantages of IBM’s watsonx Code Assistant for Z, for instance, is that it is brought to prospective customers and demonstrated using their own code and data. They see it in action in their environment, which allows them to make a knowledgeable decision.
4 Insights Into the watsonx Code Assistant for Z Process
Larin explained how watsonx maximizes the capability of generative AI to create a smoother app modernization process. Here are the four steps that take operators from slow, error-prone operations to fully modernized mainframe applications:
Understanding
As data is collected, what Rege refers to as “noise” can be added to it. The process begins by removing identifying information. Then some data is randomized. So, for example, with census data, ages could be randomly increased or decreased so that it becomes unidentifiable.
Refactoring
Once a candidate is identified, refactoring extracts a specific COBOL application into a COBOL business service.
Understanding
Transforming
From this point, the client can keep the candidate in COBOL, or they can choose to translate it into Java, which involves the generative AI component of watsonx. This means a multiple million-line application is translated into a couple thousand lines of COBOL business service.
This creates a form factor that AI can manage, and from there, the corresponding Java architecture can be mapped out. The data is sent to the prompt in COBOL, and it returns in Java. The developer can take the easier flow of Java to modernize COBOL apps while making them more modular.
Understanding
Validation
This is perhaps the most critical step since many clients are dealing with banking, healthcare, and government applications that have no room for error. The developer takes that return and reviews it for mistakes and tests the results to validate them. Automation provides the ability to run test cases to see if the new Java script produces the same results as one would in COBOL.
If It’s Not Broken, Do You Fix It?
Many mainframe operators realize the need to modernize apps but they’re not completely sure about what to do. A modernization project has risks: It can go over budget, requiring numerous consultants, and it could create countless complications that affect other parts of the system.
But the question that might need to be asked is: What if you do nothing? Apps become harder to update, end user expectations rise and the legacy language talent pools continue to shrink. The answer is clearly using a generative AI solution that tackles modernization with less complexity, risk and cost.
The generative AI revolution in mainframe management is just beginning. The space is evolving quickly and as it continues, capabilities and languages will be added, and more exciting developments are just over the horizon. “Today's products will continue to evolve,” says Larin. “It's possible that every six months we may see big improvements. Clients are giving us ideas on what they'd like to see and overall it's an exciting time.”
