Modernize mainframe applications with generative AI

You can look behind a sleek mobile application or commodity interface and find the mainframe that runs the show deep beneath the integration and services layers of major enterprise application architectures.

Critical applications and systems of record rely on these core systems as part of their hybrid infrastructure. Any disruption to ongoing operations could be disastrous for the ongoing operational integrity of your business. So much so that many companies are afraid of real change.

However, as technical debt accumulates, change is inevitable. To achieve business agility and meet competitive challenges and customer demands, companies must fully modernize these applications. Instead of delaying change, leaders should explore new ways to accelerate digital transformation in hybrid strategies.

Don’t blame COBOL for modernization delays

Perhaps the biggest obstacle to mainframe modernization is talent shortage. Many of the mainframe and application experts who have created and added to enterprise COBOL codebases over the years have moved on or are likely to retire soon.

What’s even scarier is that it will be difficult to recruit the next generation of talent because computer science graduates who have learned Java and modern languages ​​cannot imagine developing mainframe applications naturally. To them, the work may not seem as glamorous as mobile app design or as agile as cloud-native development. In many ways, this is a somewhat unfair trend.

COBOL was created long before object orientation existed. Much less service-oriented or cloud computing. With a concise command set, the language shouldn’t be difficult to learn or understand for new developers. And there’s no reason why mainframe applications can’t benefit from agile development and small, incremental releases within DevOps-style automated pipelines.

grasp What teams have been doing with COBOL over the past few years This is what makes change difficult to manage. Developers endlessly added to and created logical loops in a procedural system that needed to be checked and updated as a whole, rather than as components or loosely coupled services.

When code and programs are tied together on a mainframe in this way, the interdependencies and potential points of failure are so complex and numerous that even experienced developers cannot solve them. This makes COBOL app development more difficult than it needs to be, leading many organizations to look to mainframes prematurely.

Overcoming the limitations of generative AI

We’ve seen a lot of hype around generative AI (or GenAI) recently due to the widespread availability of large-scale language models (LLMs) like ChatGPT and consumer-grade visual AI image generators.

Although many exciting possibilities are emerging in this field, there is a nagging “fantasy factor” when applying an LLM to critical business workflows. When AI is trained with content found on the Internet, it can provide persuasive and believable dialogue, but often not completely accurate responses. For example, ChatGPT recently cited a hypothetical precedent from a federal court that could sanction lazy lawyers who use it.

There are similar issues with trusting chatbot AI to code business applications. A generalized LLM can provide reasonable, general suggestions on how to improve an app, how to easily fill out a standard registration form, or how to code an Asteroids-style game, but the functional integrity of a business application depends on the machine on which the AI ​​model was trained. It largely depends on your training data. with.

Fortunately, production-focused AI research had been underway for several years before ChatGPT emerged. IBM® has been building deep learning and inference models under the watsonx™ brand, and as a mainframe originator and innovator, we have built trained and calibrated observational GenAI models for the COBOL to Java transition.

The latest IBM watsonx™ Code Assistant for Z solution accelerates mainframe application modernization using both rules-based processes and generative AI. Now, development teams can support developers in application discovery, automatic refactoring, and transition from COBOL to Java through highly practical and enterprise-focused use of GenAI and automation.

Modernize your mainframe applications in 3 steps

For mainframe applications to be as agile and flexible as other object-oriented or distributed applications, organizations must make them a top-level feature of their continuous delivery pipeline. IBM watsonx Code Assistant for Z helps developers bring COBOL code into the application modernization lifecycle through three steps:

1. discovery. Before modernizing, developers need to know what needs attention. First, the solution takes inventory of all programs on the mainframe, mapping each program’s architectural flow diagram with all data inputs and outputs. The visual flow model makes it easier for developers and designers to spot dependencies and apparent dead ends within the code base.
2. Refactoring. This step is about breaking down the monolith into a more consumable form. IBM watsonx Code Assistant for Z explores long-running program code bases to understand the intended business logic of the system. This solution refactors COBOL code into modular business service components by separating instructions and data into individual processes.
3. conversion. This is where the magic of an LLM tailored to your enterprise COBOL to Java transformation can make a difference. The GenAI model transforms COBOL program components into Java classes, allowing true object orientation and separation of concerns, allowing multiple teams to work in parallel and agile. Developers can then focus on improving their Java code right in the IDE, with AI providing predictive suggestions, similar to the co-pilot functionality found in other development tools.

Intellix Take

We are generally skeptical of most vendor claims about AI. Because often the claims about AI are simply automation under another name.

Compared to learning all the nuances of the English language and guessing the factual basis of words and paragraphs, mastering the syntax and structure of languages ​​like COBOL and Java seems like a perfect fit for GenAI’s domain.

Generative AI models designed for the enterprise, like IBM Watsonx Code Assistant for Z, can reduce modernization efforts and costs for the world’s most resource-constrained organizations. Applications on known platforms containing thousands of lines of code are ideal training bases for generative AI models such as IBM watsonx Code Assistant for Z.

Even in resource-constrained environments, GenAI can help teams remove modernization obstacles and empower modern mainframe developers to significantly improve the agility and resilience of their most critical business applications.

For more information, see other posts in the Intellyx Analyst Thought Leadership Series.

Accelerate mainframe application modernization with generative AI

©2024 Intellyx BV Intellyx has editorial responsibility for this document. No AI bots were used to create this content. As of this writing, IBM is an Intellyx customer.