Learning IBM i with AI
If you are new to IBM i — or new to AI-assisted operations — this page is your starting point. The IBM i AI Platform teaches you the system as you work: you ask questions in plain English, observe live state safely, and approve every change before it happens.
You do not need to memorize command syntax or navigate green-screen menus to begin. You need curiosity and a willingness to verify.
The Observe → Recommend → Act Model
Every tool in the platform follows a three-phase safety model. Think of it as a built-in approval workflow that protects your system while you learn.
Read-only inspection of system state. Zero risk.
Generate a validated change plan for human review.
Execute the approved plan. Mutating operations only with explicit consent.
What each phase means for an admin
| Phase | What it does | Why it matters for safety |
|---|---|---|
| Observe | Gathers read-only facts: active jobs, storage use, message queues, logs, and database metrics. Tools like inspect_jobs and inspect_messages never change state. | You can explore freely on production systems. There is no risk of accidental modification while you are learning. |
| Recommend | Builds a Change Plan: validates syntax, checks prerequisites, and generates a human-readable preview of the command. Tools like manage_job and manage_user_profile in "preview" mode show you what would happen. | You see the intent before the impact. This is your chance to catch bad assumptions, incorrect job names, or unintended scope. |
| Act | Performs the actual state change: ending a job, creating a user, starting a subsystem. Tools like manage_job and manage_user_profile in "execute" mode execute only after you or another admin explicitly approve. | Mutations are gated behind human approval. The AI cannot accidentally disable a user or kill a critical job on its own. |
The separation exists because learning and doing are not the same activity. Observe lets you gather context without pressure. Recommend lets you practice decision-making without consequence. Act commits the change only when you are ready.
Why AI Changes IBM i Admin Work
IBM i has a reputation for stability, but that stability has historically required deep, specialized knowledge. Green-screen fluency, command syntax memorization, and decades of tribal knowledge created a high barrier to entry.
The IBM i AI Platform changes this equation:
- Natural language observation. Ask, "Why is subsystem QBATCH slow?" and the AI will run
inspect_jobsandinspect_subsystemsautomatically — then explain the results in plain English. - Grounded documentation. The
get_ibmi_procedure_helptool retrieves official IBM documentation, command syntax, and operational procedures. The AI answers are anchored in authoritative sources, not guesswork. - Safe experimentation. Because Act tools require explicit approval, you can hand the keyboard to a junior admin or a curious developer without fear. The platform teaches proper sequencing: look first, plan second, act last.
- Demystified vocabulary. Terms like ASP, LPAR, Subsystem, and Journal are defined in the Glossary and reinforced every time the AI explains a result.
You still make the decisions. The AI simply removes the friction of syntax and navigation so you can focus on judgement.
How to Get Started: Your 3-Step Path
Step 1: Connect
Follow the Setup & Connectivity guide to configure your AI provider and connect to your IBM i partition. No prior AI or IBM i experience is required — the guide covers API keys, ODBC ports, and SSH setup step by step.
If you are deploying for a team, use the Web Deployment Guide. For personal diagnostics, choose the Desktop Requirements Guide.
Step 2: Learn the vocabulary
Open the IBM i Glossary and skim the terms for Work Management, Storage, and Security. You do not need to memorize everything. The glossary is designed for rapid lookup: read a definition, then ask the AI to observe that part of your system.
Example: After reading about Job Queue, try asking: "Check my job queues and tell me if anything is backing up."
Step 3: Observe before you act
Browse the Tools Reference to see which tools map to the Observe, Recommend, and Act phases. Start with read-only tools only. Run inspect_system to confirm connectivity, then inspect_storage and inspect_jobs to build your first mental model of system health.
Only move to Recommend and Act tools after you are comfortable reading system state and verifying glossary concepts.
Learning Progression
Use this sequence to structure your study. Each step builds on the last.
| Stage | Resource | What you will learn |
|---|---|---|
| Start Here | This page | The safety model, the learning philosophy, and where to find everything. |
| Glossary | IBM i Glossary | Core vocabulary: jobs, subsystems, storage pools, security terms, and IFS concepts. |
| Tools | Tools Reference | The complete inventory of Observe, Recommend, Act, Query, and Help tools. |
| Playbooks | Admin Playbooks | Daily operating procedures: health checks, user management, and troubleshooting workflows. |
| Self-Study | Self-Study Runbook | Structured exercises for deepening fluency through guided scenarios. |
Beginner rule: Spend your first week in Observe only. Master the art of asking, "What is the current state?" before you ever ask, "Can you change this?"
Frequently Asked Questions
Is it safe to let AI access my IBM i?
Yes — if you respect the phase boundaries. Observe tools are read-only and safe to run on any system. Recommend tools only generate previews. Act tools require explicit human approval and are clearly labeled as mutating. The platform is designed so that the AI cannot execute destructive commands autonomously.
Do I need to be an IBM i expert?
No. The platform is built for learners. The Glossary defines every concept, and the AI explains its findings in plain language. If you are a Linux or Windows administrator transitioning to IBM i, you will find the terminology mapping especially useful.
What if the AI is wrong or hallucinates?
Always verify Recommend plans against your own knowledge or the get_ibmi_procedure_help tool. The platform grounds its answers in live system state and official IBM documentation, but you are the final authority. If a plan looks suspicious, cancel it and ask the AI to re-observe the relevant state.
Which deployment option should I choose?
- Web Application: Best for teams, shared audit trails, and governed environments.
- Desktop Application: Best for single users, offline work, and deep-dive diagnostics.
- MCP Server: Best if you want to integrate IBM i into an existing AI client or IDE.
See the index for a full comparison.
How do I know which tool to use?
Start broad, then narrow. If you suspect a performance issue, begin with inspect_system (view: "health") (Observe). If you need to understand specific library objects or authorities, use inspect_objects. The Tools Reference categorizes every tool by function and risk level.
Next Steps
- Set up your connection — AI provider, ODBC, and SSH.
- Read the Glossary — Learn the vocabulary of IBM i administration.
- Explore the Tools — Map capabilities to the Observe → Recommend → Act model.
- Follow an Admin Playbook — Apply what you learned to real daily workflows.
- Work through the Self-Study Runbook — Build lasting fluency with guided practice.
IBM i has managed critical business workloads for decades. With AI-assisted observation and a rigorous safety model, you can learn to manage it with confidence.