AI for IT support is most immediately useful not for replacing your engineers, but for pulling the knowledge out of their heads and turning it into a written process anyone can follow. If your SOPs live in one senior tech's memory - or nowhere at all - a language model can draft a working document from raw inputs in under an hour. A human still reviews, approves, and owns every SOP before it goes live.
How to generate an SOP from tribal knowledge using AI
1 - Extract the raw knowledge
Interview the engineer who owns the process - record or take notes. Alternatively, export the last 20-30 closed tickets on the topic. You want the actual steps taken, the edge cases they have hit, and the things that go wrong. Paste these inputs directly into your AI session without editing them first.
2 - Prompt for a structured draft
Give the model a clear instruction: 'Using the inputs below, write a step-by-step SOP for [process name]. Include: purpose, scope, prerequisites, numbered steps, and a troubleshooting section. Flag any gaps where information seems missing.' The explicit flag-gaps instruction is important - it stops the model from hallucinating filler for things it does not know.
3 - Iterate with follow-up prompts
The first draft will have gaps or vague steps. Ask targeted follow-ups: 'Step 4 mentions rebooting the agent - what are the exact commands for Windows vs Linux?' or 'Add a Prerequisites section listing the access roles required.' Each round tightens the document without starting over. Keep the source engineer in the loop so their corrections feed back into the session.
4 - Human review, approval, and version control
Print or export the draft and walk through it with the engineer and a junior tech who has never done the task. The junior tech is your usability test - if they get stuck, the SOP is not clear enough yet. Fix those gaps, get sign-off from the process owner, then commit it to your documentation system with a version number and review date. The human is accountable; the AI just did the drafting.
A Worked Example: Documenting a VPN Onboarding Process With AI for IT Support
A five-person IT team had one engineer who handled every new VPN client setup. When she was on leave, tickets piled up. The manager exported 15 closed VPN tickets and pasted them - plus a 20-minute interview transcript - into an AI session. The prompt asked for a numbered SOP with troubleshooting steps and flagged gaps. The model returned a 600-word draft in under 2 minutes. Two gaps were flagged: the exact registry key for a Windows 11 client fix, and which team owns the MFA token provisioning. The engineer filled both gaps in a follow-up prompt. The final SOP went through one review session, was approved, and is now in the team wiki. Three junior techs have since completed VPN setups without escalating.
Common questions about using AI for IT support documentation
Does the AI need access to our ticketing system directly?
How do we keep SOPs accurate as processes change?
Want help building a documentation process your team will actually use?
VITI Security works with IT teams to turn scattered knowledge into structured, maintainable processes - combining practical AI tooling with managed IT support that keeps your operations running. Talk to us about where to start.