Rocket Software brings governed AI agents to mainframe operations
Rocket Software is extending its Enterprise Virtual Assistant (EVA) platform with EVA 2.0, which introduces a PlanGuard security layer that evaluates AI agent actions on mainframe systems to prevent unrestricted access to critical resources. This governed‑AI approach lets enterprises probe and eventually automate operational workloads while maintaining strict control over what agents can do. Engineers can leverage the built‑in PlanGuard checks to enforce policy compliance before any automated operation is executed.
⚡ Key Takeaways
- EVA 2.0 will add PlanGuard, a security layer that evaluates AI agent actions on mainframe resources.
- The architecture inserts a policy‑evaluation component (PlanGuard) between the agent and the mainframe workload.
- The design trades off full autonomous automation for controlled, governed access, limiting agents from unrestricted resource manipulation.
- Integration requires routing agent requests through the PlanGuard evaluation step within the EVA platform.
- A prerequisite is that agents operate under the EVA framework; unrestricted agents will be blocked by PlanGuard.
By embedding policy enforcement directly into the AI agent pipeline, engineers can safely introduce automation to legacy mainframe environments without exposing critical systems to uncontrolled AI actions.
✅ Practical Steps
- Deploy EVA 2.0 and configure PlanGuard to define the permissible actions for AI agents on your mainframe workloads.
- Route all agent‑initiated operations through the PlanGuard evaluation API to ensure compliance before execution.
Want the full story? Read the original article.
Read on SiliconANGLE AI ↗