Techstrong TV (DevOps.com)

Bringing Coordinated AI to the Mainframe

The discussion centers on BMC’s Automated Mainframe Intelligence (AMI) portfolio, which is evolving from siloed generative AI tools toward a coordinated, agent‑driven intelligence layer that spans development, operations, and governance across mainframe environments. By embedding AI assistants and autonomous agents directly into the mainframe stack, BMC aims to turn isolated code‑explanation features into collaborative workflows that multiple developers, testers, and offshore teams can share, audit, and repeat. Key insights include the need for a central orchestration plane that enforces policies, role‑based access, and audit trails while allowing agents to act across subsystems such as databases, security modules, and capacity planners. The panel highlighted how AI‑augmented capacity planning can balance performance improvements against cost trade‑offs, and how emerging interaction standards like MCP and MCPS will replace ad‑hoc APIs to ensure reliable, vendor‑agnostic communication. Notable examples cited were the AMI assistant’s code‑explanation capability, the agentic pipeline that can commit code from a generative model directly into a mainframe, and the distinction between simple data‑sharing APIs and MCP servers that execute actions on behalf of users. Priya emphasized risk avoidance in mission‑critical mainframes, while Matt stressed a federated orchestration model that mirrors existing team responsibilities. The broader implication is that enterprises can modernize legacy mainframes in place—leveraging AI to boost efficiency, reduce manual errors, and maintain compliance—without falling into vendor lock‑in. Standardized protocols and a governance‑first approach are essential for scaling these capabilities across heterogeneous environments.