In technology, we tend to lump talent into a single bucket. But there are three distinct layers at play: skills, knowledge, and experience. Skills are the tools. Knowledge is how you learn to use them. Experience is knowing which tool to use and when to use it, often under pressure.
That last one is the problem.
Accepting the Experience Gap
Training programs, documentation, and modern interfaces have made it easier than ever for a new generation to acquire technical skills. You can learn commands, systems, and workflows faster than before.
What you can’t learn quickly is experience. Experience is earned gradually, through mentorship, repetition, failure, and time. It’s the late-night outages. The subtle anomalies. The gut instinct that tells you which metric matters and which one is noise. It’s apprenticeship, not instruction.
And right now, that mentorship is thinning out, with engineers who have built up deep mainframe expertise moving on. What they take with them isn’t just knowledge, it’s judgment. Younger, capable, and motivated professionals are stepping in, but they don’t have the same level of work context as their predecessors. That context is what makes decision-making efficient.
The result? A widening experience gap at exactly the wrong time.
Bridging With AI
Mainframe environments are more complex, more integrated, and more critical to high-volume digital operations. At the same time, organizations are under pressure to deliver higher productivity with fewer resources. That’s a dangerous mismatch: rising expectations paired with declining experiential depth.
Historically, software hasn’t solved this. It’s been good at surfacing data—dashboards, alerts, logs—but not at interpreting it. It assumes the user brings experience to the table. It hands you thousands of metrics and expects you to know which ones matter.
One customer recently described a situation where several database objects had entered an exception state. The software correctly identified the affected objects and their status, but that was only the beginning. The database administrator (DBA) assigned to the issue spent hours trying to determine the right recovery procedure. As the manager put it, this happens more often now because the people who immediately knew how to resolve these situations and who had built up that judgment over time are no longer always available on demand.
In this case, the software showed what was wrong, but not what to do next. A more task-oriented system, guided by AI, could have reduced the time to resolution by recommending likely steps, surfacing relevant context, and helping the DBA move from diagnosis to action.
But what if that information firehose approach is wrong?
Look at how everyday tools have evolved. Navigation apps don’t just show you a map anymore. They tell you how long your commute will take at 5 p.m. on a Tuesday versus 8 a.m. on a Saturday. They factor in patterns, anomalies, and historical behavior. They quietly embed experience-based recommendations into the interface. That’s the model.
In mainframe environments, systems generate enormous volumes of data and thousands of signals across performance, availability, and workload. Experienced engineers know how to interpret those signals. They recognize patterns. They spot what’s “normal” for a given time of day, workload, or business cycle.
Fixing Root Causes With Synthesized Recommendations
Thankfully, AI can replicate that pattern recognition. By layering machine learning and anomaly detection on top of existing observability data, software can begin to distinguish signal from noise. It can learn what “normal” looks like across thousands of variables and flag deviations that actually matter. More importantly, it can guide the user, not just alerting that something is wrong, but suggesting where to look and why.
That’s a shift from tools to teammates. Some vendors are already moving in this direction — building systems that don’t just monitor, but recommend. Instead of requiring users to build complex dashboards based on years of experience, the software highlights the most relevant metrics automatically. Instead of vague alerts, it provides context, probable causes, impacted systems, and suggested next steps.
This is how you begin to close the experience gap. Not by replacing people, but by augmenting them. By capturing fragments of collective experience (i.e., patterns, decisions, outcomes) and embedding them into software that improves over time. Every anomaly detected, every issue resolved, every pattern confirmed becomes part of a growing knowledge base that future users can benefit from.
It’s not perfect. And it’s not complete. But it’s necessary. Because the alternative is expecting a new generation to rebuild decades of experience from scratch, under increasing pressure in increasingly complex systems. That’s not realistic, and it’s not scalable.
Wisdom at the Speed of AI
The opportunity in front of the mainframe industry is bigger than just modernization. It’s about rethinking how software supports human decision-making. It’s about designing systems that assume less experience and compensate with smarter guidance.
Skills can be taught. Knowledge can be documented. But experience has always been the bottleneck. Until now.
With thoughtfully applied AI, we have a chance to compress the timeline — not by shortcutting experience, but by redistributing it. By making the wisdom of the past available in the workflows of the present. That’s smart thinking. And it’s exactly what the mainframe needs.
Toine Michielse brings over 40 years of mainframe experience to the table. His career has been focused on data management in all facets, starting on IMS and Db2 for z/OS as a programmer, database administrator and Db2 for z/OS System engineer. He has worked both for customers and for the IBM Db2 development lab as a lab advocate. Before joining Broadcom, he worked at Swiss Re where he was the lead mainframe architect and led the mainframe capacity team, as well as a team of specialists that provided education and consultancy to the various Swiss Re development units. At Broadcom, he leads a talented team of technical consultants that is responsible for the EMEA region and covers data management, DevOps, and security.