The mainframe is often looked at as technology of the past, but over 70% of Fortune 500 companies still rely on it given its ability to modernize and adapt to new environments that meet the industry’s growing technological advancements. But as mainframes move deeper into these hybrid environments, developers cannot just rely on mainframe expertise alone. Integrating AI — especially agentic AI — in mainframe environments before your people and processes are prepared can create or worsen existing problems. A recent report found that even as companies invest in technology and skills to prepare mainframe developers for the future, 35% of those already working with AI on the mainframe say existing skills gaps are hindering progress, underscoring the need for a more multiskilled approach.
Developers need to be on top of the latest rules around AI deployment, build on their existing skills and technical fluency, and develop the confidence to experiment safely as new technologies enter the enterprise landscape.
Like learning to drive, deploying AI on the mainframe requires more than just knowing the mechanics. It takes confidence, real-world practice and trust in the system.
Learning the rules of the road
Enterprise security teams are increasingly concerned about AI-driven threats, which Gartner identifies as the fastest-growing category of attacks. However, governance and compliance efforts have not kept up with the rapid pace of AI development. Limited understanding, low prioritization and gaps in technical oversight are creating significant risks, especially since AI systems depend on sensitive data and continually evolving code. As with any transformative technology, the rules and regulations around AI are evolving, and developers must stay current, especially in regulated environments.
By gaining a better understanding and awareness of the latest information, developers will be able to monitor the mainframe using the latest organizational guardrails designed to protect systems and vulnerable data. Effective guardrails and governance frameworks ensure employees use AI responsibly, helping organizations avoid regulatory breaches, legal exposure, security risks and the financial or reputational damage that can result.
As experienced professionals retire and misconceptions around “legacy” systems persist, organizations face a growing skills gap. Investing in mainframe education isn’t just a technical necessity; it’s a strategic move to safeguard key business applications and infrastructure, ensure continuity and maintain competitiveness in a rapidly evolving landscape.
Security as the seatbelt for AI
As with any major technology shift, security must be a cornerstone of agentic AI adoption. These systems need robust safeguards to ensure safe and reliable operations. One critical measure is role-based access control, which limits agent permissions based on their specific functions, reducing the risk of unauthorized actions. This is reinforced by secure credential storage, encryption and multi-factor authentication to protect both system integrity and user authorization.
Equally important are controls governing how agents interact with data and execute tasks. Techniques such as input validation, output monitoring and behavioral constraints help prevent misuse of language model components, while human oversight for high-impact actions adds an extra layer of protection. Prompt engineering is equally critical, as poorly structured prompts can lead to unpredictable or unsafe actions by AI agents. By applying disciplined prompt design alongside input validation, organizations can ensure that AI-driven processes remain accurate, secure and aligned with business objectives. This ensures that sensitive or disruptive operations are carefully reviewed before execution.
Finally, continuous monitoring, observability and auditing provide transparency and accountability across the lifecycle of agentic systems. These practices strengthen trust and integrity, enabling organizations to deploy AI confidently while maintaining compliance and control.
Driving readiness through core skill integration
To succeed in the AI era, developers need a blend of core and future-ready skills grounded by foundational training. When used responsibly and thoughtfully, AI tools can be used for closing that gap and fast-tracking modernization efforts — provided they are deployed intentionally.
Mainframe fluency anchors AI deployments in deep system knowledge. AI literacy and agentic AI fluency enable developers to build, train and manage AI agents that align with business logic and enterprise goals.
The challenge isn’t a lack of educational resources — there’s an abundance of training available. The real barrier is perception: mainframe skills are often seen as outdated or ‘untrendy.’ Yet these skills remain highly valuable. In fact, specialized expertise in areas like COBOL often commands higher salaries than more common languages such as Java because qualified professionals are in short supply.
A study by the Futurum Group revealed that while educational institutions are producing more mainframe-trained graduates, 61% of the respondents still report a significant skills gap between what’s taught and what’s needed in practice. Showing the need for continued skill integration, and not just foundational training.
Confidence behind the wheel
In the same way drivers can know the rules of the road, without confidence out on the road, they are just as much at risk. The same goes for developers dealing with deploying a new technology like AI.
Mentorship, simulation tools and community learning can all help close the confidence gap, by giving mainframe developers safe-play zones where they can explore AI deployments without fear of impacting production.
Organizations can be critical enablers by creating the infrastructure and culture that support experimentation and skill growth. Building confidence is just as vital as building a strong skillset; both are necessary to move beyond cautious pilots to confident production-scale deployments.
Driving isn’t just about moving forward, it’s about being prepared for the unexpected, knowing when to accelerate and trusting your tools.
The rise of AI isn’t just about automating tasks, it’s about empowering people to engage with systems they once found intimidating. For many, mainframe modernization has been slowed not by lack of vision but by a lack of confidence, uncertainty around where to start, what might break or how to fill gaps left by retiring experts. But when AI agents become part of daily workflows, that fear starts to fade. Developers begin to trust their footing. And teams start making decisions with clarity instead of caution.
With the right foundation, the mainframe becomes not just something developers maintain — but something they drive.
This article is published as part of the Foundry Expert Contributor Network.
Want to join?
Read More from This Article: Skilling mainframe developers for an AI-first future
Source: News