Rocket expands mainframe EVA AI platform with PlanGuard

Rocket Software has expanded its EVA agentic AI platform for mainframe systems and introduced a new security layer called PlanGuard.

The update extends EVA into a broader range of operational tasks on core systems, including diagnostics, security, batch processing, application management and data work. Organisations in financial services, government, insurance, retail and telecommunications are already piloting the platform.

EVA is designed to let AI agents analyse operational data and automate selected tasks within policies set by enterprise users. PlanGuard adds a control point between AI decision-making and system execution, using identity and policy checks to limit what an agent can access and do on a mainframe.

Rocket is targeting a longstanding problem for large organisations that still rely on mainframes for core workloads: a shortage of specialist skills. In research commissioned from Hanover Research, 81% of financial services IT leaders said the mainframe skills gap was very or extremely significant, while 87% said AI would help address it over the next two years.

Another Hanover finding cited by Rocket found that 94% of financial services IT leaders rank improving IT operations with AI as a high or top priority. EVA uses natural language interfaces to help staff work through operational issues that have often required specialist knowledge, multiple tools and manual investigation.

Use cases

Rocket outlined several areas where customers are testing the software. In operations and diagnostics, EVA is being used for job failure pattern analysis, dashboard and KPI generation, workflow automation, financial reporting analysis, online banking application optimisation, and queue and backlog analysis in IBM CICS environments.

In security, the focus includes vulnerability detection, compliance work, patch management and identifying risks linked to post-quantum cryptography. For batch processing, the platform is intended to help teams monitor performance against service-level agreements, identify critical failures and investigate root causes more quickly.

Application and configuration management is another target area, including analysis of high CPU use and degradation affecting IBM Db2 data sharing, buffer pools and cache performance. The latest EVA release also supports Rocket's broader aim of unified enterprise data access and output management.

EVA is model-agnostic and uses a standards-based method to connect to z/OS data. Rocket says that approach is intended to simplify access to operational and enterprise data while keeping governance controls in place.

Control layer

PlanGuard is central to Rocket's pitch to customers seeking tighter oversight of AI activity in sensitive environments. The software is intended to keep AI-driven actions within approved boundaries, while logging activity for audit and allowing human oversight where needed.

Mainframes remain deeply embedded in sectors such as banking, insurance and government, where organisations often want to modernise operations without disrupting systems that handle high-value or regulated workloads. That creates an opening for suppliers promising practical AI uses in established infrastructure rather than only in newer cloud environments.

Rocket, which is privately held and backed by Bain Capital Private Equity, says it serves more than 13,500 customers and works with 850 partners. The business focuses on software for IT modernisation across mainframe, infrastructure, applications and data environments.

"Enterprises have relied on the mainframe to run their mission-critical workloads for decades," said Milan Shetti, President and Chief Executive Officer of Rocket Software.

"AI promises to unlock even more value from these environments, but only if enterprises can deploy it securely and without disruption. We're helping customers apply agentic AI to mission-critical systems with speed, confidence and control, closing the skills gap and putting that expertise within reach of every enterprise team," Shetti said.