AI for the Mainframe: Turning System Data Into Answers
For the past year, nearly every conversation about artificial intelligence and the mainframe has centered on one question: What will AI do to COBOL?
Will it modernize legacy applications? Generate new programs? Rewrite decades of enterprise logic?
Those questions make headlines. But they may also be missing the more immediate opportunity.
During a conversation at SHARE Orlando Conference 2026, Gil Peleg, Co-founder and CEO at Geniez AI, suggested that AI’s most practical impact on the mainframe may not involve writing code at all.
Instead, it may come from helping teams understand what the platform produces in abundance: system data.
Every mainframe environment generates a steady stream of signals—SMF records, console messages, subsystem metrics, job outputs, and performance logs. For experienced engineers, those signals reveal system health, workload behavior, and emerging problems. But interpreting them still takes time.
Peleg believes that is where AI can make the biggest difference.
“We’re not trying to replace experienced people with decades of mainframe knowledge,” Peleg said. “We’re trying to remove the boring work so they can focus on the things that actually require expertise.”
In that sense, AI’s role may be less about replacing developers and more about helping operators and system programmers interpret mainframe system behavior faster. Think of it like a decoder ring for the mainframe.
Turning Mainframe System Data Into Answers
Mainframe environments produce some of the richest operational telemetry in enterprise computing. SMF records, console messages, syslogs, job outputs, and subsystem metrics all provide insight into what the system is doing. The challenge is connecting those signals quickly enough to understand what they mean.
Peleg’s approach with Geniez links AI directly to those operational data sources and translates them into answers engineers can use. Instead of navigating multiple dashboards or scanning logs, users can ask questions in plain language:
Why is CICS running slowly?
Which jobs are consuming the most CPU right now?
The system gathers relevant data in real time and builds a response based on the system’s actual behavior.
“If you want to ask why your CICS region is slow, we pull the real data from your system—DB2 locks, performance metrics, system messages—and build the response from that.”
The AI isn’t guessing, and the engineer isn’t hypothesizing. The technology is interpreting live system signals and returning them to the person in simple language.
Giving Experts Back Their Most Limited Resource: Time
Investigating performance issues often requires reviewing multiple logs, metrics, and monitoring tools before identifying the root cause. AI can compress that investigative work.
Peleg says teams often see benefits in several areas: recovering time spent on manual analysis, avoiding unnecessary capacity increases, and diagnosing incidents faster.
“Almost everyone we speak to is extremely busy with a backlog of work. If AI helps them gain back even an hour a day, that’s enormous.”
In large organizations, even small time savings can scale quickly.
Bridging the Experience Gap
Mainframe teams also face challenges with knowledge transfer. Many professionals running enterprise systems today have decades of experience interpreting system behavior. Newer engineers often lack that context.
AI can help bridge the gap by explaining what system metrics mean and suggesting the next questions to ask.
“If you’re a junior engineer and you see a metric you don’t understand, the system can explain what it means and what to ask next.”
This kind of contextual guidance doesn’t replace experience, but it can help new engineers become productive faster.
A Different Direction for AI on the Mainframe
Much of the broader AI conversation focuses on generating code or automating development tasks, and that’s all good. Those capabilities may eventually reach the mainframe ecosystem as well.
But more immediately, Peleg believes the opportunity for AI on the mainframe lies in helping professionals interpret what their systems are already telling them.
Mainframe platforms continuously generate signals on performance, workload behavior, and system health. AI can help the people responsible for those systems understand them faster.
And if that happens, the biggest impact of AI on the mainframe might be asking the right questions to gain better information that improves mainframe performance.
Next
Watch Gil Peleg’s full interview on YouTube
See how Geniez AI got started — and how works
0 Comments