What happened

IBM’s chief performance marketing officer, Ritika Gunnar, told SiliconANGLE that the company’s AI operating model now puts agentic workloads at its core. She explained that as AI systems begin to act autonomously, they become "always on" and require constant support across data, applications and infrastructure. The model is built around four components, data, workloads, automation and hybrid cloud, and stresses the need to integrate with existing enterprise systems such as SAP and Salesforce. Gunnar said that without full observability of AI-infused workloads, organisations cannot pick the right model or agent for a given task.

Why it matters

Enterprises that run large-scale batch jobs on z/OS face a new set of pressures. Agentic workloads generate a continuous stream of decisions, logs and prompts that must be observed and governed in real time. If the data feeding those agents is not governed, the agents may make poor decisions, leading to missed service-level targets or compliance gaps. Gunnar highlighted the importance of shifting governance left, ensuring the right people have timely access to the right data while preserving provenance and ownership. She also noted that human judgement remains essential; tacit knowledge held by experienced staff must be captured in prompts or other contextual inputs to guide agents where they are not yet mature.

What this means for enterprise IT teams

IT teams will need to extend monitoring and automation beyond traditional batch schedulers to cover AI-driven agents. Observability tools must capture agent decisions, evaluate performance against business outcomes and feed the results back into model training. Integration layers will have to bridge mainframe workloads with cloud-native AI services, demanding secure APIs and data pipelines. Teams must also implement real-time data governance that tracks who accessed which data and when, to satisfy both regulatory and operational requirements. Human-in-the-loop processes will need to be codified so that operators can intervene when an agent’s decision falls outside acceptable parameters. Finally, cost monitoring becomes critical, as always-on agents can consume resources continuously and affect the overall cost-effectiveness of the AI deployment.

In the words of Ritika Gunnar, chief performance marketing officer, IBM: "As you’re more successful in using AI, especially as AI is taking action, you now have workloads that are continuously being utilized"

The bottom line

IBM’s focus on always-on agentic workloads reshapes the AI operating model and adds complexity for mainframe-centric organisations. Enterprises must invest in end-to-end observability, governance and integration to keep AI agents aligned with business goals. Tools that provide visibility into both traditional batch jobs and emerging AI agents will be critical to managing risk and cost on the mainframe platform.

Sources

  • siliconangle.com