AI for Mainframe Management: Our Roadmap and Principles

The mainframe is not going away. According to the Arcati Mainframe Navigator 2025 User Survey, more than two-thirds of mainframe-running organizations report that the majority of their revenue runs on IBM Z — and 57% are actively growing capacity. Yet 62% cite a skills gap, and 70% of professionals have 20+ years of experience, with many nearing retirement. AI/ML in production on the mainframe sits at just 13%, even as 48% rank it a top priority.

That gap is where SMT Data operates. With over 35 years of mainframe performance, capacity, and cost expertise in our ITBIaaS platform, we are building AI capabilities that are practical, governed, and grounded in trusted data.

Our AI Roadmap: Four Steps

We are taking an iterative approach: AI must earn its place at each stage and each step will build on the previous ones.

Step 1 — Open Access via MCP

We are building an MCP (Model Context Protocol) server so customers can connect using their own AI tools: ChatGPT, Claude, or any other AI assistant. Immediate access to our mainframe capacity knowledge base. Open standards, no lock-in.

Step 2 — Grace: Mainframe Chat Assistant

Grace is a domain-specialized AI assistant. Ask it plain-language questions about your mainframe dat a such as “are we at risk of hitting capacity limits before our next hardware refresh?” or “where are we wasting capacity we are already paying for?” As mainframe expertise concentrates in a generation nearing retirement, Grace makes that knowledge accessible to the whole team.

Step 3 — Intelligent Alerts & Recommendations

When ITBI detects an anomaly or emerging issue, it does not just raise an alert, it delivers a recommendation on what is causing it and what to do. Guidance grounded in decades of real-world mainframe expertise, not generic alerting rules.

Step 4 — Predictive Forecasting

Insight becomes foresight. ITBI will be able to forecast where capacity constraints are heading, so teams can make adjustments before they become problems.

How We Approach AI as an Organization

Serving banking, insurance, government, and other large enterprises running mission-critical infrastructure means every AI decision has a direct impact on our customers’ compliance.

Five principles guide how we build:

• Security and data ownership. Strict access controls ensure the right data reaches the right people, no further. Customer data is never used to train external AI models.
• Data quality is the foundation. AI is only as reliable as the data beneath it. Mainframe SMF data is complex and requires deep domain knowledge to normalize, that is exactly what ITBI guarantees before any AI touches it.
• Cloud transparency. ITBI runs in the cloud, and so does our AI. We are deliberate about where data is processed and stored, data residency is a compliance obligation, not an afterthought.
• AI cost discipline. Running AI at scale has real costs. We design for efficiency, ensuring AI overhead is proportionate to the value delivered. In a world where every MIPS is scrutinized, that matters.
• Human judgment stays in the loop. When ITBI flags a capacity risk or recommends an action, a qualified person must evaluate and own that decision. We build AI to sharpen human judgment not replace it. On a platform this business-critical, that is not a limitation. It is the only responsible way to operate.

Ambitious by Design. Responsible by Principle.

Our roadmap and our principles are two sides of the same coin. The four steps describe where we are taking ITBI. The five principles describe how we get there — without cutting corners on data integrity, security, transparency, cost discipline, or human accountability. We are ambitious about what AI can do for mainframe management. We are equally clear-eyed about what it takes to do it responsibly.