What happens when AI moves from advising and assisting to acting? That question came up in my latest conversation with Anthony DiStauro, Senior Director of Architecture for AI at BMC Software. In parts one and two of this series, we explored AI as both a trusted advisor and a collaborative partner. In this final discussion, the focus shifts to AI agents that can not only recommend actions but also begin initiating them within the mainframe environment.
The Shift to Proactive AI Agents
Today, most AI systems still wait passively for a prompt, responding to questions or providing recommendations when asked. Agentic AI moves beyond this, with the ability to continually monitor systems, follow reasoning paths over patterns, and take action (within boundaries). AI agents can operate around the clock, learn from each experience, and gradually improve performance.
Importantly, agents can be designed to be proactive without stepping in for humans, augmenting teams and handling repetitive or time-consuming groundwork while people focus on higher-level priorities.
Real-World Impact: Mainframe Examples
In our discussion, DiStauro described several areas where proactive AI agents are already showing measurable impact for mainframe teams.
For example, in managing legacy COBOL code, AI agents are beginning to move beyond simple explanation. They autonomously review vast codebases, pinpoint inefficiencies, and even propose or draft improvements, then submit the work to human developers for rapid review and approval. The tedious groundwork is automated, freeing developers to focus on modernization and quality initiatives.
A concrete proof point shared by DiStauro illustrates just how transformative this can be. By orchestrating three specialized AI agents, one for code generation, another for test creation, and a third for deployment, he built an entire application in just four days, without writing a line of code himself. In this scenario, the human role shifts from coder to architect, leveraging agent orchestration to multiply productivity and accelerate outcomes.
Preserving institutional knowledge is another domain where these digital collaborators shine. As experienced mainframe professionals retire, AI agents become invaluable by observing and internalizing the methods used to resolve complex problems. They effectively become digital knowledge keepers, supporting new team members, helping train future talent, and ensuring that operational wisdom remains within the organization as the workforce evolves.
Human in the Loop
Despite increasing autonomy, humans remain critical to the process. One useful way to think about these agents is as digital assistants, capable of preparing and proposing actions but still requiring human approval before executing high-stakes tasks. Having robust human oversight, transparency, and audibility reassures teams that they can always see what decisions were made and why. In highly regulated mainframe environments, this level of governance isn’t optional; it’s essential.
Multi-Agent Collaboration and Open Standards
Another theme from our conversation was the shift from a single, monolithic AI toward teams of specialized agents working together. When a major IT incident occurs, agents focused on networking, databases, and applications can coordinate quickly and efficiently.
Technologies like the Model Context Protocol (MCP) simplify this orchestration. In simple terms, MCP gives agents a standardized way to connect, share context, and solve problems together. For mainframe teams, that means faster, more cohesive responses without complex integration hurdles.
Steps to Adoption
Organizations exploring proactive AI agents don’t necessarily need sweeping changes to begin experimenting. Many organizations begin by deploying agents in advisory roles and gradually introducing them into higher-impact tasks as confidence grows. Trust can be built by choosing well-defined, non-critical processes first, and expanding as reliability is demonstrated, all under human supervision.
Empowerment Through Proactive AI
Across this series, we’ve explored AI as advisor, partner, and now actor. Each step represents a deeper level of trust and integration. Proactive AI agents mark a turning point, where insight no longer waits for execution, and action is guided by intelligence infused with organizational knowledge. With humans firmly in control, enterprises can move faster, innovate more boldly, and ensure their mainframe environments remain resilient and future-ready. The result is not automation for its own sake, but a smarter, more responsive enterprise where intelligence drives meaningful action.
To learn more, watch the full conversation on Techstrong TV.
