A survey of 1,319 IT professionals who work with mainframes finds that while mainframes remain a strategic platform, the level of criticality also limits the degree to which respondents are willing to rely on artificial intelligence (AI), for now at least, to manage them.

Conducted by BMC Software, the survey finds 94% of respondents view the mainframe as a long-term platform or a destination for new workloads, with an equal percentage planning to continue to invest. Overall, 69% said capacity is growing on the platform, but only 18% said that growth is being driven by new applications.

When it comes to AI, there is a significant amount of interest. A total of 40% of respondents are planning to invest in AI agents to help manage AI operations, with 36% planning to rely on third-party AI agents. Top use cases for AI include performance tuning (37%), problem detection (36%), automated testing (35%), documentation (33%), root cause analysis (32%), code explanation (32%) and incident resolution (32%).

Top AI areas of investment over the next 12 months are developer assistance (43%), tools for preserving and transferring knowledge (41%), analytics for performance tuning (36%), creating agents to manage the mainframe (36%), AIOps for incident management (33%) and using agents from others to manage the mainframe (32%), the survey finds.

Survey respondents, however, also have concerns, including high implementation costs (41%), security and privacy (39%), data integration issues (37%) and regulatory/compliance challenges (22%).
Nevertheless, implementing AI ranks second (49%) on the list of top priorities for the coming year, well after compliance/security (61%) and slightly ahead of cost optimization (46%). In fact, 68% of respondents expect to see value from AIOps on the mainframe either within less than six months (26%) or a year (42%).

The degree to which AI is trusted is still relatively nascent. The majority of respondents today rely on AI to create alerts or generate recommendations but less than a quarter are comfortable enough to allow AI of any kind to complete a task on its own.

Matt Whitbourne, vice president of product management and design for the BMC Automated Mainframe Intelligence (AMI) portfolio, said that while the perceived risks associated with AI agents are high, other classes of AI tools are already helping to optimize mainframe environments in a way that also serves to reduce the level of expertise that might otherwise be required.

As AIOps continues to mature, it’s also expected that many of the roles within the IT teams that manage mainframes will also continue to evolve, he added. In fact, AI should help more of the IT teams that manage mainframes to adopt best DevOps practices as mainframes are more deeply integrated with distributed computing environments, he noted. The mainframe as a platform will always be unique, but the level of expertise required to manage it continues to decline, added Whitbourne.

At this juncture, it’s not so much a question of whether IT teams that manage mainframes will adopt AI so much as how soon and to what degree. Given the mission-critical nature of the workloads that run on mainframes, the appetite for experimenting with any unproven technology is, as always, comparatively limited.