Mainframe professionals are increasingly moving from AI experimentation to operational deployment, using generative AI and machine learning tools to gain system insights and recommend actions, according to a new global survey of more than 1,300 mainframe practitioners and decision makers. The survey found that implementing AI technologies remains a top priority for 45% of respondents, but the overall tone is one of cautious realism. Organizations are no longer asking only what AI can do; they want to know where it can be trusted, how it can be governed, and where it will deliver measurable business value.
Mainframes have long depended on rules-based automation and structured monitoring to keep systems running. Yet the latest generation of AI, particularly generative AI, is transforming the operator experience. Instead of manually scrolling through logs, metrics and historical incident reports to diagnose a failure, operators can now receive a clear, natural-language explanation of the likely root cause and a recommended set of next steps. These AI-driven tools digest past issue resolutions, system documentation, and organizational knowledge to produce contextually relevant guidance. The result is that a junior operator can act with the confidence of a veteran, while senior experts can focus on the most complex edge cases.
From enthusiasm to pragmatic adoption
The report says the mainframe community is moving from 'AI enthusiasm to pragmatic adoption.' That shift is visible in the way AI is being positioned across the enterprise. After much discussion, planning and investment, most mainframe executives are not ready to turn over full control to AI. Instead, they are adopting a human-in-the-loop model in which AI acts as an advisor, not an executor. An operator or administrator reviews the AI's suggestions, validates them, and then implements the changes. This approach builds trust incrementally and ensures that human judgment remains at the center of critical decisions.
One senior technology executive quoted in the study says the shift represents a significant change in mindset: 'We are seeing a significant shift as organizations have gone from asking how they can use AI to asking where they can trust it, how it can be governed, and where it delivers measurable value.' He adds that the path to greater AI autonomy on the mainframe will be earned through trust, with humans remaining in the loop to oversee and implement AI recommendations. In short, AI has not yet gained the full trust of the mainframe world, which is why human oversight remains an essential part of operations.
Top concerns about AI-driven mainframe solutions
Despite the excitement, there are real obstacles. The survey identifies four main concerns among mainframe teams implementing AI-driven solutions. The top concern is high implementation costs, cited by 41% of respondents. Security and privacy follow closely at 39%, with data integration issues at 37% and regulatory or compliance concerns at 22%.
The cost challenge is not surprising. AI adoption often requires new software licenses, specialized infrastructure, and skills development. On the mainframe, where workloads are highly sensitive and performance expectations are extreme, the cost of getting AI wrong can be steep. Security and privacy concerns also loom larger on the mainframe than on other platforms because these systems process some of the most valuable data in the enterprise, including financial transactions, customer records, and core business applications. Data integration issues reflect the difficulty of connecting AI models to the mainframe data assets that may reside in multiple subsystems, databases, and file formats.
The focus on security is also tied to the growing use of digital certificates. The survey notes that as AI-based tools and AI agents become more common, the need for more digital certificates will grow sharply. These certificates must be issued, renewed and managed frequently, and many organizations still rely on manual processes or home-grown automation to handle the workload. The report warns that the volume and complexity will soon expose the limitations of those approaches.
Digital certificate management remains a challenge
According to the survey, most organizations use either in-house automated solutions or manual management for digital certificates. That presents a risk as AI integration grows. AI-based applications need secure connections to mainframe services, and each connection may require a certificate. If certificates are not managed properly, the result can be outages, security gaps, or compliance failures. The report suggests that organizations should start investing in robust certificate lifecycle management, including automation and centralized oversight, before the demand overwhelms current practices.
Key findings across the enterprise
The survey includes several other findings that show how AI is being used across mainframe operations. In particular, it focused on data security, agentic AI, AIOps time to value, GenAI-assisted tools, and knowledge transfer.
Data security and recovery
Organizations place a high priority on securely connecting AI to their data. Data integration issues rank as the third-highest concern when implementing AI solutions, and modernizing data management is seen as the third most important AIOps capability. Data recovery also grew in importance, with 35% of respondents listing it as a top priority, up 4 percentage points from the previous year. The emphasis on recovery suggests that as enterprises rely more on AI-driven operations, they are paying more attention to resilience and backup processes.
Agentic AI gaining momentum
AI agents are emerging as a major investment area. The report identifies a group of 'Leaders' who are prioritizing AI technologies, and many are already planning to invest in autonomous or semi-autonomous agents. Over the next two years, 40% of Leaders plan to invest in creating their own agents to manage the mainframe, while 36% plan to invest in third-party agents for that purpose. These agents are expected to handle tasks such as monitoring, incident triage, and routine maintenance, with human oversight as needed.
AIOps delivers time to value
Artificial intelligence for IT operations, or AIOps, is producing measurable results on the mainframe. Of those using AIOps, 68% of respondents overall say they have seen time to value within one year. Among regular users, that number rises to 74%. The report says that finding causes and determining how to fix issues remain the top challenges in mainframe operations, and AI, especially generative AI, directly addresses these challenges. The survey found that 58% of Leaders who prioritize AI technologies name the implementation of GenAI solutions as the most important AIOps capability.
GenAI-assisted tools as trusted advisors
Rules-based logic combined with AI and machine learning has improved problem detection and even enabled proactive remedies before issues affect service. However, operations teams are still left to determine root causes and choose appropriate fixes. GenAI-assisted tools fill that gap by providing contextual advice on what actions to take next. The tools ingest past issue resolutions, documentation, and institutional knowledge to suggest next steps in natural language. This makes the system a trusted advisor that draws on years of accumulated experience to guide operators at every skill level.
AI for knowledge transfer
With many experienced mainframe professionals retiring, knowledge transfer is a pressing concern. AI is not replacing mentoring, but it is making it more effective. The survey found that 40% of respondents are using AI for documentation and knowledge transfer. Among those who prioritize staffing and skills and are hiring new staff to address skills gaps, 49% are using AI assistants to help train employees. By providing immediate access to documented procedures, past solutions, and best practices, AI enables less experienced mainframers to gain confidence and become productive more quickly.
Source: Network World News