12 AI agent prompts for IT & Internal Help
Copy-paste, beginner-friendly prompts to help you run an internal help desk, provision access and write incident reports. Each one is based on a real-world example and works with ChatGPT, Claude or an AI agent. 8 are free below, and 4 more are in AgentPlaybook Pro.
Free IT & Internal Help prompts
Access provisioning checklist
Standardize who gets what access when people join or leave, like Avalara's provisioning agent.
For a new [role], list the apps and access levels they need from: [app list]. Then make an offboarding checklist for when someone leaves, so nothing stays open.
Incident summary writer
Turn alerts and chat logs into a clear incident report.
From these alerts and chat logs [paste], write an incident report: timeline, impact, root cause (or best guess), fix applied, and 3 prevention steps. Plain language for non-engineers.
Database incident diagnosis brief
Turn blocking chains, wait stats, and slow-query clues into a root-cause brief with a recommended fix and a ready-to-assign ticket, like Cornerstone OnDemand's Orion AI multi-agent system on Amazon Bedrock that cut database diagnosis from 45 minutes to 10.
Diagnose this database incident. Symptoms: [paste what users/SRE reported]. Metrics and waits: [paste blocking sessions, wait types, CPU, long-running queries, timestamps]. Recent changes: [paste deploys, schema changes, or unknown]. Produce: (1) most likely root cause and why, (2) what to rule out next, (3) a safe first remediation step vs anything that needs approval (e.g. killing a session), (4) a filled ticket draft: title, severity, evidence, owner team, and acceptance check. If data is incomplete, list exactly what to query next instead of guessing.
Find and fix configuration drift on critical devices
Have an agent compare running configs to the last-known-good baseline, flag out-of-band changes, and prepare a restore plan with an audit trail for approval, like SolarWinds Avior's goal-oriented agents for remediating configuration drift and continuous policy audits in hybrid IT estates.
Check these devices for configuration drift. Inventory / device list: [paste or attach]. Baseline: [last-known-good configs or policy]. Criticality: [which devices must stay correct]. For each device: (1) diff running config vs baseline, (2) flag out-of-band or unexpected changes, (3) note compliance gaps against [policy], (4) propose a restore or remediation step with risk level. Produce an audit-friendly table and a restore plan that requires my approval. Don't push configs, reboot, or lock anyone out.
Find out what's filling up your computer's storage
Instead of installing a cleaner app, ask a coding agent to scan your disk and show what's using space by category and app, with cleanup suggestions, like Kevin Vaghasiya, whose Codex scan found 50 GB of node_modules in forgotten projects and freed up 55 GB+.
My [Mac / Windows PC] says storage is almost full. Scan my disk (start with [home folder / whole drive]) and show me what's using the space: a breakdown by category (apps, videos, photos, downloads, developer files like node_modules or build caches, old backups, system data) and the 20 largest folders and files, with sizes. For each item, suggest whether it's safe to remove, safe to move to an external drive or cloud, or should be kept, and explain why. Pay special attention to old project folders, caches, and duplicate files. Only read and measure: don't delete, move, or empty anything. Give me a checklist, and wait for me to approve each cleanup step before you run it.
Brief an on-call engineer when an incident hits
When a production incident lands, pull telemetry, similar past incidents, and troubleshooting-guide matches into one Teams (or Slack) brief so the on-call engineer spends less time hunting portals. Microsoft Digital's Smart DRI Agent does this for 40+ teams and saves about 600 engineering hours a month.
You are my on-call briefing assistant. Incident: [paste title, description, severity, service name, and start time]. Troubleshooting guide excerpts: [paste relevant runbook sections]. Recent similar incidents: [paste 2-5 past tickets with resolution notes]. Telemetry highlights from the last 15 minutes: [paste errors, traces, or dashboard notes]. Write a brief for the designated responsible engineer with: (1) what appears broken in one sentence, (2) the three most likely causes ranked, (3) the first three checks to run with exact commands or portal links if I gave them, (4) which past incident looks most similar and what fixed it, (5) who to page next if this is not resolved in 30 minutes. Flag anything that needs elevated production permissions. Keep it under 250 words so I can read it in Teams.
Let an AI help desk agent fix the routine IT requests start to finish while your staff check its work
Small IT teams spend most of the day on the same device, password and connectivity questions, leaving little time for real problems. An AI agent can answer employees from your own help articles, route anything it can't solve to the right person, write the ticket summary, and fully handle common requests, with staff reviewing what it did. ServiceNow announced on Oct 8, 2026 that the City of Raleigh, whose four-person help desk supports 4,400 employees, uses one agent to guide self-service and another to resolve L1 requests end to end: IT service desk costs fell 66%, 98% of tickets are routed correctly on first contact, 98% of AI incident summaries are accepted as-is, and nearly half of requests are resolved without a person.
You are our first-line IT help desk. Our help articles: [paste or link your how-to guides]. Who handles what: [e.g. network = Sam, laptops = Priya, HR systems = HR team]. Requests you may complete yourself: [e.g. password reset steps, printer setup, Wi-Fi help, VPN install]. For this employee message: [paste message]. 1) Restate the problem in one line. 2) If it matches a help article and is on the 'complete yourself' list, give clear numbered steps in plain words and ask them to reply 'fixed' or 'still broken'. 3) If not, say who it's going to and why. 4) Write a ticket summary: user, device, problem, steps tried, category, urgency (low/medium/high), assigned to. 5) Never ask for or repeat a password, and never promise something that needs approval (new software, admin rights). End with anything you're unsure about so a staff member can check it.
Give store and office staff instant answers from your company documents, showing only what each person is allowed to see
In businesses with many locations, staff waste minutes digging through shared folders for a procedure, policy or price rule, or they interrupt a manager. An AI assistant connected to your SharePoint or Google Drive can answer in plain words, quote the document it used and only search files that employee can already open. Models at Work reported on Oct 22, 2026 that pet-care retailer IskayPet, with 330+ stores and 4,500 staff in Spain, Portugal and France, cut the time to find an operational answer from 3 to 5 minutes to about 7 seconds with an assistant built on Amazon Bedrock over its SharePoint documents.
You are the staff help assistant for [business name]. Answer employee questions using only the company documents below (or the folders you're connected to). Documents: [paste or connect: procedures, HR policies, store operations manual, price and returns rules]. For each question: 1) Give a short, plain answer first (2-4 sentences). 2) Add the exact steps if it's a how-to. 3) Name the document and section you used, and quote the key line. 4) If the documents don't answer it, or two documents disagree, say so clearly and tell the employee who to ask: [manager / HR / IT contact]. Never guess, never answer from general knowledge, and never share information from documents the person isn't allowed to open. Answer in the language the employee writes in. First question: [type the question].
More in AgentPlaybook Pro
- Internal help desk botAnswer common IT questions (password resets, access) from your docs.
- Turn old Slack help threads into an IT knowledge baseMine years of solved questions in your help channel into step-by-step fixes an agent can answer from, including after hours. ShipStation Global did this for its Slack IT helper Deckhand, which replied to 97.5% of messages and kept 812 of 1,221 conversations from becoming tickets in its first three months.
- Diagnose mainframe issues in plain languageLet operators ask natural-language questions about live mainframe systems, get a diagnosis, and only then approve any change—useful where mainframe expertise is scarce. Rocket Software acquired Geniez AI to put this kind of agent on the mainframe inside its EVA platform.
- Stop technicians chasing ticket updates: put an AI service desk in front of your PSA to take intake, triage and follow-up, and keep your dispatcher for escalations and onsite schedulingAlvarez Technology Group, a family-owned IT support company (MSP) in Salinas, California with about 72 managed clients on ConnectWise PSA, used Thread's AI service desk to automate ticket intake, triage and follow-up. According to Thread's case study, its 10-person engineering team reclaimed about 40 hours a week that had gone to chasing status updates, a voice overflow agent covered the queue when the dispatcher was out, and the company kept a 13-week streak of 100% customer satisfaction while taking on new clients without hiring more Tier 1 staff. The dispatcher moved to escalations and onsite coordination. This prompt helps any IT team map out the same split before buying a tool.
32 copy-paste agent prompts (PDF) plus 5 new prompts every week.
Get it free