Unlocking mainframe expertise with AI

Mainframe systems aren't going anywhere, but the expertise to maintain them is becoming harder to find. Mainframes power almost 70% of the world’s production IT workloads, spanning 28 industries across more than 70 countries. Yet the specialized knowledge needed to optimize their performance is increasingly scarce.

As veteran mainframe professionals retire, a growing knowledge gap is forming. While modern computer science curricula tend to emphasize languages like Java and Python—which are now supported on mainframe platforms like IBM Z—the real challenge lies deeper.

The skills shortage centers on understanding mainframe-specific terminology, core concepts and complex subsystems that differ from the ones found in cloud and distributed environments. Yet enterprise training budgets often prioritize these architectures over mainframe concepts and designs.

The result? A widening divide between ongoing mainframe reliance and the shrinking pool of professionals who have command over the unique operational environment of mainframe systems like IBM Z®. For business leaders, this shortfall threatens both operational continuity and the capacity for innovation.

Leveraging artificial intelligence (AI) to bridge the skills gap has emerged as a transformative solution. Generative AI (gen AI) combined with automation can streamline how system programmers, operators and developers acquire knowledge, and it can enhance productivity, efficiency and work quality, regardless of their experience level.

Adopting this approach could be a game changer for many global industry leaders. The approach might be important especially for professionals in industries like finance, healthcare, government and retail, where mainframes support high-volume, high-stakes operations that demand reliability, security and scalability.

AI, particularly generative AI, offers a unique opportunity to address these issues by making mainframe knowledge more accessible. By harnessing the power of large language models (LLMs), retrieval-augmented generation (RAG) and automation frameworks, organizations can reduce skill acquisition for new professionals. Also, this approach empowers experienced users to focus on strategic tasks that drive business value.

Here’s how AI can contribute to mainframe education and productivity.

New-to-mainframe users often face steep learning curves and struggle with complex tasks. For instance, imagine a developer who, after hours of searching through documentation, is uncertain about whether they’ve found the correct solution. Their uncertainty slows workflows and drives them to seek confirmation from a subject matter expert (SME), as even minor mistakes can have serious consequences in the mainframe environment.

What this developer needs is a reliable and accessible mentor—and this is where AI steps in. Conversational AI provides users with a natural language interface to access curated information and step-by-step guidance. Instead of interrupting SMEs with questions, users can interact with an AI-powered assistant to:

• Understand foundational mainframe concepts

• Access information about product updates and new features

• Follow step-by-step procedures for completing tasks

• Receive clear explanations of error messages and system behavior

Empowered by a retrieval-augmented generation (RAG) framework, such systems can provide grounded and accurate responses by leveraging trusted knowledge sources. This method not only improves the reliability of answers but also helps mitigate risks like misinformation.

Some best practices for implementing RAG systems include ensuring they draw from high-quality, domain-specific sources and keeping the AI updated regularly. A key advantage of RAG is that it helps avoid hallucinations. These safeguards help ensure that AI remains a reliable and trustworthy resource for mainframe learning and support.

While AI assistants help users learn, automation empowers them to act. By providing guided workflows, automation allows early-tenure users to perform tasks effectively, even without deep prior knowledge. 

The benefits of automation extend beyond beginners; experienced mainframe professionals also stand to gain. For instance, consider a seasoned systems programmer faced with configuring a rarely used security setting. While they might have the foundational knowledge, it’s possible that years have passed since they last performed this task, requiring a refresher on the steps, dependencies and nuances of the system.

Without proper guidance and practices, this process can become time-intensive, risking delays to critical updates and increasing the likelihood of errors.

Automation and AI-driven tools bridge this gap by offering step-by-step assistance, automating repetitive workflows and providing real-time recommendations based on the context of conversational AI. These tools empower both early-career developers tackling new problems and seasoned veterans revisiting rarely used skills. By reducing errors, improving efficiency and minimizing frustration, automation enables mainframe professionals to focus on strategic, high-value initiatives.

Equally important is the need for automation solutions to be able to integrate with the organization’s existing investments. Enterprises often rely on well-established automation frameworks and technologies like job control language (JCL), restructured extended executor (REXX) and Ansible, having invested heavily in these systems over the years. To maximize value and ensure adoption, automation must support these foundational tools, providing compatibility and seamless integration rather than demanding costly replacements.

AI solutions should be designed to cater to the unique needs of different user groups, ensuring that each group benefits from a tailored experience that aligns with their skill levels and goals. For instance:

• Early-career professionals can benefit from AI-guided, hands-on experiences that focus on reinforcing learning through manual task execution, enabling them to build confidence as they engage with foundational concepts.

• Mid-level users might rely on AI-driven automation to handle routine tasks efficiently, saving time and minimizing the risk of errors while freeing them to focus on more complex activities.

• Experienced professionals could leverage AI to explore advanced features, streamline complex operations or even encode their expertise into reusable automation assets that can benefit the broader team.

Creating role-specific assistants ensures that the AI experience aligns with users’ responsibilities and skill levels. These purpose-built assistants address questions and fulfill specific automation requirements, enabling users to seamlessly complete tasks and navigate intricate processes with an engaging conversational AI experience.

Different AI assistants can be designed to meet the needs of various user types.

When developing AI solutions, simplifying development and maintenance processes is critical. As AI systems continue to evolve, they should aim to provide support while minimizing new layers of complexity in deployment. Achieving this goal requires AI systems that are intuitive, interoperable with existing tools and flexible enough to adapt to diverse use cases. As a result, automation benefits are accessible and sustainable for every user group.

Here are some approaches.

Low-code platforms allow domain experts to create, modify and publish automation assets without extensive programming knowledge. These assets can then be incorporated into AI assistants, making them available to a broader user base.

For example, existing automation scripts in Ansible, JCL or REXX can be transformed into reusable skills.

A centralized repository ensures that automation assets are easily discoverable and reusable across teams. This method not only accelerates task execution but also fosters knowledge sharing and standardization.

When deploying AI solutions for mainframe environments, selecting an architecture is crucial. Cloud-based options can reduce runtime costs and simplify scalability, but might face integration challenges with mainframe systems behind firewalls. On-premises deployments align well with mainframe security requirements, providing robust automation capabilities and direct integration. However, this approach often requires significant upfront investment in GPU hardware and IT infrastructure.

Many organizations find a hybrid cloud approach to be optimal, using the cloud for general knowledge functions and on-premises deployment for secure automation tasks. The key is aligning your AI support infrastructure with your organization's security requirements, existing mainframe investments and automation goals.

AI solutions for mainframe environments offer a promising path forward as organizations face the challenge of maintaining critical systems with a diminishing pool of specialized talent. IBM is bringing this vision to life through the IBM z17®, its next-generation mainframe designed for enterprise-scale AI.  

With built-in AI capabilities powered by the Telum® II processor and the new Spyre™ AI accelerator, the z17 enables real-time AI inferencing directly on the system. AI assistant and agent tools like watsonx Code Assistant® for Z and watsonx Assistant® for Z deliver intelligent support for developers and IT teams. Also, integration with IBM Z Operations Unite streamlines operations with AI-powered incident detection and resolution.

Designed to meet the needs of both seasoned professionals and users new to the platform, these tools help bridge the skill gap, increase productivity and spur innovation. By embedding AI directly into the mainframe, IBM is helping democratize mainframe expertise, making it a renewable resource that ensures these systems continue delivering value for well into the future.