Sponsored by Broadcom
It’s not a surprise that the epicenter of AI adoption is software development. Developers are always looking for ways to work smarter — to eliminate repetitive jobs, to automate manual tasks or to remove annoying blockers. AI code assistants and agentic AI are powerful new tools in a developer’s toolkit that helps them work smarter than ever envisioned in the pre-AI world.
These tools are so effective, 90% of developers surveyed by DORA Research acknowledged using them and more than 80% believe they have increased their productivity. Many large enterprises, including ones in highly regulated industries like banking and financial services, are experiencing pervasive use across their development organizations as experimentation has given way to incorporating them as a core part of engineering strategy.
The buzz was evident at the recent GitHub Universe conference where Broadcom joined our hosts, GitHub, in showcasing the breathtaking pace of AI innovation. GitHub Copilot, their market-leading code assistant that runs in VS Code, was center stage, and our mainframe interoperability demonstration with Code4z had a big wow factor. GitHub also introduced Agent HQ, a platform that serves as a centralized command center for managing and orchestrating multiple AI coding agents within the GitHub environment.
The key takeaway from the event? It’s a fast-moving market, so aligning with the right partners and tools is critical for long-term success.
Beyond Code Generation
Early use cases focused on AI assistants as “code generators,” so some vendors in the mainframe space applied them to rewriting mainframe applications in languages like Java. Broadcom, however, recognized that much of the value in AI is with tasks outside of code writing, like application analysis, debugging, and testing.
Mainframe development is more about editing mission-critical, multimillion-line applications than it is about generating new code.
Broadcom’s guidance is that mainframe teams should apply the technology to their biggest problem areas — typically staffing and business agility. How do we attract the next generation of talent and get them up to speed quickly? How do we equip them to deliver changes at a velocity and quality that matches or exceeds that of any other platform?
Top Use Cases Addressing Mainframe Challenges
Forrester’s recent study of mainframe developers, sponsored by Broadcom, found that on average, it takes developers over three days just to determine where to make code changes, with over 20% requiring a week or more. Anecdotal feedback paints an even bleaker picture, with reports of well over a month to gain a sufficient level of confidence to begin implementing the changes. With many applications ranging in the millions of lines of code maintained over decades with less-than-robust documentation, code complexity presents one of the biggest friction points in mainframe development.
The good news is that this is exactly the type of problem generative AI is suited to solve. For example, it can explain complex business logic in plain language (whichever language the developer finds most comfortable) and, with access to application analysis, it can identify dependencies, downstream impacts of code/schema changes, and reusable business logic well-suited for APIs.
Using code assistants as a virtual “pair programmer” reduces code understanding cycle time and improves overall speed and business agility. And while new-to-mainframe talent has the most to gain as they have the biggest knowledge gap — even the most experienced developers find cycle time improvements.
As many z/OS applications are hampered by poor code coverage, AI can uplevel quality by facilitating developer-oriented regression and unit testing. Code assistants help with a variety of tasks such a suggesting test cases based on the potential impacts of a change, and writing unit tests in the source code language. Improving code coverage, in concert with better-informed developers equipped with code understanding, makes deploying changes of any kind — source code, JCL, schemas, APIs — less daunting and more rewarding.
Like tests, documentation can be limited and outdated as well. Code assistants offer in-line annotations in plain language. While the documentation can include business logic, data flows, dependencies, etc., the main difference is that it’s summarized and it’s in-line, rather than in a separate panel designed for natural language scrolling. By adding it to the source code, it becomes searchable, interactive documentation.
In summary, adopting these use cases addresses both challenges: attracting the next generation of talent and delivering code at a higher velocity. Reducing friction in this way enables developers to spend more time in the productive flow state, which keeps them happy, too.
Fast Tracking Adoption
To take advantage of code assistants and agentic AI, a contemporary developer experience like VS Code is an effective prerequisite. Code4z, which is built on VS Code, offers seamless access to assistants like GitHub Copilot, enabling AI to be applied to any application development use case: analysis, editing, debugging, building, testing, deploying, etc. A bonus for Endevor users is that Code4z offers a Model Context Protocol (MCP) extension that connects Endevor code, documentation, and artifacts to the code assistants safely and securely. This facilitates both application-specific, as well as portfolio-level scenarios (e.g., track a variable across applications).
In addition to helping new-to-mainframe talent learn about the platform (e.g., languages, databases, subsystems), AI code assistants can also help veteran mainframe developers learn how to use modern tools like Code4z. For example, “How can I analyze application performance using Code4z?” or “How can I create a CICS API using Code4z?”
Mainframe teams should be embracing code assistants where they help drive business agility, ensuring the long-term vitality of their applications. There’s never been a more exciting time to be a mainframe developer.
To experience AI code assistants in action, request a Code4z workshop for your team.
Dave McNierney is a developer advocate with over 25 years of experience in enterprise software development. A regular blogger and presenter, he raises awareness of new and emerging practices, like DevOps, that reduce delivery friction and improve developer productivity and career satisfaction