Can AI Agents Truly Replace Mainframe Veterans? Analyzing Rocket’s Move

Key Highlights

• Rocket Software has acquired Geniez AI to integrate generative and agentic AI directly into mission-critical mainframe operations.
• The deal aims to accelerate Rocket EVA by embedding AI agents that can monitor live systems and diagnose incidents using natural language.
• Geniez AI brings specialized capabilities designed to allow real-time data access while keeping sensitive data securely on-premises.
• We see this acquisition as a direct response to the deepening skills shortage in traditional mainframe environments.

The News

Rocket Software announced its acquisition of generative AI startup Geniez AI to augment mainframe professionals with expert AI agents. These agents are designed to operate directly within the mainframe environment to monitor systems, diagnose incidents, and execute workflows using natural language. The acquisition aims to accelerate the development of Rocket EVA, the company's agentic AI platform for core enterprise systems. (Press release: https://www.rocketsoftware.com/en-us/news/rocket-software-announces-acquisition-of-geniez-ai-to-accelerate-agentic-ai-for-mainframe-modernization)

Analyst Take

The narrative of an operational mainframe skills crisis fundamentally misreads how technical competencies have modernized. Rather than facing an unfillable talent void, organizations are actively redefining what running enterprise systems requires. Findings from the 2026 BMC Mainframe Survey (https://www.bmc.com/info/mainframe-survey.html) confirm that data centers are not stalling out: 94 percent of respondents view the platform as a long-term foundation or one attracting new workloads, and 69 percent report sustained growth in general-purpose capacity.

The idea that data centers cannot survive unless universities teach COBOL or bespoke System Management Facility (SMF) log triage reflects an outdated narrative. The modern mainframe leverages universal software engineering disciplines. By standardizing on mainstream languages like Java and Python, standard IDEs, and cross-platform Git and DevOps toolchains, IT teams can hire directly from the broader software engineering talent pool rather than searching for scarce legacy specialists. HyperFRAME's 1H 2026 research reinforces this point: while 24 percent of enterprise infrastructure leaders report a significant mainframe skills gap, only 4 percent identify mainframe systems as their most significant gap to overcome, placing it well behind network and security expertise (33 percent), automation and DevOps practices (26 percent), and cloud platforms (25 percent).

Where tribal knowledge previously created operational exposure, modern organizations are actively neutralizing the gap through tooling. The 2026 BMC survey, which has over 20+ years of longitudinal data, highlights a deliberate shift toward using AI, automation, organizational intelligence, and cross-platform tooling specifically to resolve skills shortages. Rather than requiring decades of memorized runbooks to troubleshoot environments, 40 percent of mainframe leaders are investing in building agents to manage the platform, while 36 percent are adopting third-party management agents. Teams are simultaneously using generative AI for automated documentation generation, performance tuning, and anomaly detection. The broader Infrastructure and Operations market, based on HyperFRAME survey data, is moving in the same direction: 56 percent of organizations expect commercial off-the-shelf software to be their most widely deployed infrastructure software category within 12 to 24 months, compared with 23 percent expecting internally developed proprietary software to lead.

Because survey respondents strongly favor advisory AI that recommends actions over autonomous execution, the required skill set has shifted from manual syntax memorization to governed operational decision-making. With Millennials and Gen Z already accounting for roughly two-thirds of the mainframe workforce, software-driven knowledge retrieval and modern developer toolchains have turned the retirement cliff into a scalable, contemporary engineering practice.

This leaves enterprises exposed when diagnosing obscure system errors or updating complex job control language. A recent McKinsey analysis highlighted that the technical talent gap remains a primary barrier to enterprise modernization. Rocket Software is responding directly to this reality by acquiring Geniez AI. This Israeli startup, founded in 2025, builds generative AI tools specifically for the mainframe environment. The acquisition brings a team of engineers who have built agents that interact with legacy systems using natural language.

What was Announced

Rocket Software acquired Geniez AI to embed agentic capabilities into its Rocket EVA platform. The Geniez AI framework is architected to allow large language models to interact securely with real-time mainframe data. It supports multiple mainframe data sources such as DB2, IMS, MQ, and VSAM. The platform aims to deliver a natural language chat interface that operators can use to query system status and troubleshoot issues.

The Geniez technology relies strictly on existing mainframe security controls like RACF. It uses patented methods designed to ensure all data remains on-premises. The software does not train models on customer data. It instead connects frontier AI models in the cloud to on-premises systems in real time. The framework includes a Mainframe Security Posture Management module designed to manage exposure lifecycles from discovery to remediation. It also features a monitor module architected to process SMF records for real-time system observability.

The framework also aims to deliver agentic workflows that react automatically to system events. The product is architected to deliver an agentic development experience directly within the developer's Integrated Development Environment. This means engineers can write or refactor legacy code without leaving their standard tools. The included Python SDK is designed to simplify integration with enterprise applications like Microsoft Teams. Rocket Software plans to integrate these engineering capabilities directly into Rocket EVA to accelerate its artificial intelligence roadmap.

The implications of this technology go beyond simple convenience. Managing a z/OS environment typically requires memorizing obscure commands and being deeply familiar with green-screen interfaces. An operator diagnosing a batch processing failure might spend hours sifting through logs. Geniez AI changes this interaction model entirely. An administrator can simply ask the system to explain what a specific job does and suggest a fix for any errors. This approach lowers the barrier to entry for junior staff. They can manage mission-critical infrastructure without needing thirty years of prior experience.

We view this acquisition as a pragmatic capability grab. Rocket Software is buying immediate engineering credibility in the agentic AI space. Geniez AI previously raised six million dollars in a seed round. They built a product that gained traction quickly among Fortune 50 companies. Their founders previously built a mainframe data management company (Model9) that BMC Software eventually acquired. Gil and his team at Geniez understand enterprise computing's stringent security requirements. By acquiring them, Rocket Software bypasses the trial and error typically associated with building generative tools for highly regulated environments.

Security remains the primary concern for any organization connecting artificial intelligence to core banking or insurance systems. Mainframe administrators are notoriously risk-averse. They guard their systems fiercely. Geniez AI addresses these concerns by avoiding traditional data lake migrations. The software does not extract data to feed an external model. It enables direct access governed by the mainframe's native security protocols. If a user does not have RACF permission to view a dataset, the AI agent cannot view it either. This architectural decision is crucial for adoption.

Enterprises want the efficiency gains of artificial intelligence without compromising their security posture. Rocket Software recognizes this delicate balance. By integrating Geniez AI, they provide a modernization path that respects the mainframe's rigid operational boundaries. The focus is on augmenting the human operator. The AI agent handles the tedious log analysis and routine maintenance tasks. The human retains ultimate control over system changes.

This level of integration alters the operational dynamic. We see a shift from reactive troubleshooting to proactive system management. When an agent can monitor live systems and execute predefined workflows when it detects an anomaly, mean time to resolution drops significantly. The system becomes somewhat self-healing. Human operators shift from fighting fires to managing AI agent parameters. This transition is vital for IT departments facing flat or shrinking budgets.

This acquisition is a smart move. Rocket Software secures specialized talent while expanding its addressable market. Organizations struggling to maintain their core systems now have a viable technological bridge. They can onboard newer talent and equip them with tools that translate complex system behavior into plain English. The skills gap will not close on its own. It requires a fundamental shift in how we interact with legacy infrastructure.

Looking Ahead

The integration of generative tools into legacy environments represents a fundamental reimagining of enterprise operations. The key trend we'll watch is how quickly conservative industries, such as financial services and government, adopt these agentic frameworks. These sectors run the global economy on mainframes, yet they are historically hesitant to introduce external connectivity to core systems. We believe Geniez AI's patented security architecture, which relies on native mainframe controls, will be the primary catalyst for overcoming this institutional resistance.

The announcement positions Rocket Software aggressively against competitors like IBM, Broadcom, and BMC. These rivals are attempting to modernize the developer experience and operational tooling all from different starting points. However, the industry is moving toward a truly autonomous agent that executes workflows via natural language. Going forward, we will closely monitor how the company performs in integrating the Geniez codebase into the broader Rocket EVA platform. The challenge lies not in the technology itself, but in seamlessly unifying disparate user interfaces and operational workflows.

HyperFRAME will track how the company performs on user adoption and security validation in future quarters. If Rocket Software can prove that these agents reduce mean time to resolution without introducing new vulnerability vectors, this acquisition will prove highly prescient. The long-term viability of the mainframe depends entirely on bridging the interface gap between decades-old architecture and modern operational expectations.