What AI on IBM i actually means in late 2026: IBM Bob and the Premium Package for i, the watsonx product family, the IBM i MCP Server, Power Autonomous Operations, AI-assisted RPG and COBOL modernization, and on-chip acceleration in Power hardware.

"AI on IBM i" is not one thing. As of September 2026 it is five distinct threads that get talked about interchangeably: IBM Bob and its Premium Package for i writing and modernizing code, IBM's watsonx product family, the IBM i MCP Server connecting AI agents to Db2 for i data, IBM Power Autonomous Operations running the infrastructure underneath, and on-chip AI acceleration in current Power hardware. Knowing which thread you actually mean is the first step to evaluating any of it seriously.

The five threads

IBM Bob and AI-assisted modernization

IBM's AI coding agent for IBM i, GA since March 2026, with Bob 2.0 and the Bob Premium Package for i shipping June 24, 2026. Explains, generates, converts, refactors, documents, and tests RPG, COBOL, CL, DDS, and SQL, and works on Db2 for i in Database Mode. Reported pricing from $20 to $240 a month.

watsonx on IBM i

IBM's broader AI platform: watsonx.ai, watsonx Orchestrate, and watsonx Assistant. General-purpose, with no IBM i connector in any of them. Since August 2026 Orchestrate can launch Bob to build agents through a Build with Bob option, currently a preview, which is the first official link between Bob and the watsonx brand. See our watsonx on IBM i overview.

IBM i MCP Server

An open-source, IBM-published Model Context Protocol server that lets AI agents and LLMs query Db2 for i data and take action on IBM i using natural language. In beta since October 2025, no GA date published.

IBM Power Autonomous Operations

An AI agent for the Power infrastructure under IBM i: monitors, diagnoses, and resolves capacity constraints with a human in the loop, through chat. Announced July 15, 2026, GA expected September 23, 2026.

On-chip AI acceleration

Matrix Math Acceleration in every Power10 and Power11 core for local inference alongside IBM i without a GPU. The entry-level Power S1112, GA July 24, 2026, is the smallest Power11 box that runs it. Covered on the AS/400 Power 11 sites.

Where each thread actually applies

Code explanation, generation, and RPG modernizationIBM Bob with the Premium Package for i, plus DevOps and modernization vendors

Db2 for i SQL work with an assistantIBM Bob Database Mode

Agents that span IBM i and other systemswatsonx Orchestrate, with Build with Bob (preview) generating the agent and IBM i access through the IBM i MCP Server (beta)

AI agent access to Db2 for i dataIBM i MCP Server (beta)

Custom generative AI over IBM i datawatsonx.ai, with OpenSearch as the vector layer because Db2 for i has no vector store

Conversational AIwatsonx Assistant (general-purpose, custom integration to IBM i)

Running the Power infrastructureIBM Power Autonomous Operations (GA expected September 23, 2026)

Local AI inference hardwarePower10 and Power11 Matrix Math Acceleration, down to the Power S1112

What is still unresolved

Three questions shape most IBM i AI designs right now and IBM has not closed any of them. Db2 for i has no vector store, so retrieval-augmented generation needs OpenSearch or another external index; our RAG on IBM i explainer covers the design. The IBM i MCP Server is still beta, so the authority and audit model for letting an agent touch production data is yours to define. And Power Autonomous Operations has no published integration points, licensing, or per-partition detail until GA. The IBM i AI watch list tracks all three with dates.

If you are just starting to evaluate AI for your IBM i shop, start with our AI project readiness checklist and AI adoption roadmap rather than picking a product first ... the right starting point depends on your data quality, staff bandwidth, and which of the five threads above actually matches your goal. Page last revised September 2, 2026.