AI for IT
IT is one of the most under-discussed AI use cases. Most playbooks focus on customer-facing work, but internal IT spends huge amounts of time on triage, documentation, and translating tech for non-tech stakeholders. AI helps with all of it — but the security and access-control parts still need a human in the loop.
What AI handles well
Triage helpdesk tickets
The problem. 80 tickets in the queue Monday morning. Most are duplicates, password resets, or "have you tried turning it off and on again." A few are real incidents. Sorting takes hours.
What AI does. Use AI to categorize incoming tickets — quick fix, escalation needed, knowledge-base candidate, real incident. Generate suggested first responses for common patterns. Humans approve before sending.
Runbooks from working systems
The problem. You inherited infrastructure with no docs. The previous IT person knew it cold. Now you're reverse-engineering on the fly.
What AI does. Have AI generate runbooks from your existing systems — paste config files, deployment scripts, recent incidents. AI structures them into "what this is, when to use it, common pitfalls, escalation path."
Security policy drafts
The problem. Compliance asks for a written policy on AI tool use, BYOD, data retention, whatever. Generic templates from compliance vendors don't fit your stack.
What AI does. Use AI to draft policies tied to your actual setup (your tools, your access controls, your incident response). Iterate with security/legal before publishing.
Incident postmortems that don't sound like blame
The problem. Outage happened. Postmortem is due. Default templates produce defensive narratives instead of actual learning.
What AI does. Use AI to structure a blameless postmortem: what happened (timeline), root cause (technical, not human), what helped, what didn't, action items with owners. Output is honest, structured, actionable.
Onboarding/offboarding checklists
The problem. New hire starts Monday. They need 12 accounts provisioned. Different roles need different access. You miss things and have to fix later.
What AI does. Generate role-specific onboarding checklists. Same for offboarding — what to revoke, what to archive, what to transfer. Human reviews and approves before execution.
Vendor evaluation for the next tool ask
The problem. A team requests a new tool. You need to evaluate it against existing tools, security implications, integration cost, and budget. The default is "we already have something similar."
What AI does. Use AI to structure the evaluation: feature parity, security posture, integration cost, switching cost, alternatives. Output: a defensible recommendation, not just "no."
Translating tech for non-tech stakeholders
The problem. You need to explain the cloud migration to the CFO. Or describe a security risk to the board. Tech-speak doesn't land. Oversimplifying loses signal.
What AI does. Use the explain-it prompt — translate technical concepts to the business audience without dumbing it down. Frame around what matters to them: cost, risk, time, dependencies.
Knowledge base articles from recurring tickets
The problem. You're solving the same 5 questions every week. A KB article would help. But writing them = the work you don't have time for.
What AI does. Have AI draft KB articles from your previous ticket replies. You polish, add screenshots, ship. Then deflect tickets with the article.
Your AI stack
Start with the foundation. Add specialized tools as the work calls for them.
Foundation LLM
Specialized add-ons
Prompts ready to use
Get started in 35 minutes
Pick an LLM and load your stack documentation
15 minClaude Project with: your tools, your network setup, your access policies, your common ticket patterns. Now any prompt about your IT environment is pre-loaded.
Generate a runbook for the system you most fear losing context on
10 minPick the system with the worst docs that's most important. Use the docs-from-code prompt. Iterate until you'd trust it for a 3am incident.
Set up automated triage for your top-5 ticket types
10 minIn your ticketing system or via a custom prompt, create response templates for the most common ticket types. AI drafts, you approve and send.
Common mistakes
Treating AI security suggestions as authoritative. AI can suggest plausible-sounding security policies that miss your specific compliance requirements. Validate against your actual standards.
AI-generated runbooks that don't match your actual infrastructure. AI doesn't know your specific config. Spot-check that paths, hostnames, and procedures are real.
Skipping the human escalation step for sensitive incidents. AI categorizes incidents based on text. A "minor issue" ticket can hide a security breach. Always have a human review for severity.
Pasting credentials, API keys, or sensitive infrastructure config into public AI. Use enterprise tiers, anonymize, or skip.
Using AI to bypass judgment on access requests. "User wants admin access" — AI can generate a justification template. Whether to grant it is your call, not AI's.