BMC has released survey findings indicating that mainframe teams are increasingly using artificial intelligence for operational insight and recommendations, while remaining markedly more cautious about letting the technology execute changes independently.
The 2026 BMC Mainframe Survey drew responses from more than 1,300 mainframe practitioners and decision-makers globally, according to the company. BMC said 94% of respondents expressed long-term confidence and continued investment in the mainframe, while 45% identified implementing AI technologies as a top priority.
The research is directly relevant to IT automation because it describes where organizations are willing to apply AI in systems that still run critical workloads. Respondents cited uses including performance tuning, problem detection, database reorganizations, IMS queue management and documentation generation. Those are operational tasks where a recommendation can be useful even if a person remains responsible for the final change.
BMC’s figures show the gap between advice and autonomous execution. Forty percent of respondents said they were willing to have AI recommend code-management actions, but 23% were willing to allow it to complete them. For database reorganizations, 43% were comfortable with an AI recommendation, while 21% were comfortable with autonomous completion.
That difference is significant for teams building AIOps and automation programs. A recommendation can shorten investigation time or surface a likely next step without altering a production system. An automated change needs additional confidence in the data, the model, the controls around it and the possible impact on an application or dependent system.
The survey also found that 36% of respondents plan to invest in creating agents to manage the mainframe, with another 32% planning to invest in third-party agents. BMC did not say how quickly those planned investments will turn into production deployments, and the figures should be read as survey responses rather than a measure of actual installed technology.
Certificate management illustrates the continuing role of conventional automation. BMC said 43% of respondents use in-house automated certificate-management solutions, 31% use product-based automated solutions and 25% still rely on manual effort. The company tied that issue to shorter TLS certificate lifecycles, which can increase the volume of renewal work and the risk of disruption without a process.
For mainframe operations leaders, the release points to a staged model for AI adoption: use it first to identify issues, prioritize work and recommend actions, then expand automation only where governance, testing and rollback procedures are strong. That is less dramatic than full autonomy, but it is the operational posture reflected in BMC’s survey findings.
The survey also identifies implementation costs, security and privacy, data integration, and regulatory or compliance requirements as recurring concerns. Those factors are relevant to automation decisions because a recommendation is only useful when it is based on operational data that teams can govern and trust. BMC did not provide a breakdown by industry, organization size or mainframe platform, so the findings should not be generalized beyond the surveyed population.
For teams planning the next step, a useful distinction is whether an AI feature observes, recommends or executes. Each stage calls for different testing and accountability. The report’s results suggest that many organizations are currently concentrating on the first two stages while they build the processes needed to expand autonomous action responsibly.