Skip to content

IT: Triage support requests and answer from the knowledge base

Deflect tier-1 IT tickets with instant KB answers — and for anything that needs a human, collect the right information and raise the ticket automatically.

Stage: Copilot Studio · For: Maker, IT Admin · Level: Intermediate · Time: 4–5 hours

📐 Full blueprint & test plan → IT Helpdesk Triage Agent — the copy-paste system prompt, topic specs, and test-case table behind this build.

Which harness? This one uses the standard harness

Every Copilot Studio agent runs on a harness — the engine underneath it. This walkthrough builds on the standard harness: you author the topics and rules, the flow is predictable, and for Microsoft 365 Copilot-licensed users it's covered inside Microsoft 365 channels (no extra credits) — the lowest-cost choice for structured, rules-based work.

Want the same use case to reason through the task on its own — planning multi-step work, recovering when a step fails, and working across your files? Build the GitHub Copilot harness version instead. It's the autonomous engine to grow into; the honest tradeoff is that it bills Copilot Credits for all usage and a Microsoft 365 Copilot license never covers it. Compare the engines · estimate the net cost.

When to use this

IT helpdesks spend a disproportionate amount of time on repeatable tier-1 questions: password resets, VPN access, software requests, "how do I…" queries. These are answerable from a knowledge base. The cost isn't the answer — it's the ticket volume and the routing noise.

A triage agent changes the economics: it answers from the KB first, and only creates a ticket when the issue genuinely needs a human. When it does raise a ticket, it collects the structured information the helpdesk actually needs (category, urgency, description) before the engineer ever sees it — no back-and-forth to gather basics.

Why Stage 6: This agent does more than answer questions. It makes a triage decision (KB answer vs. ticket) and takes an action (creating a ticket via Power Automate). That logic and the action layer require Copilot Studio — Agent Builder can't do it.

What you'll need

  • Copilot Studio access and a Power Platform environment
  • Your IT knowledge base in SharePoint or as uploaded docs (KB articles, known issues, how-to guides)
  • A ticketing system with a Power Automate connector (ServiceNow, Jira, Zendesk, or a SharePoint list as a lightweight alternative)
  • The required fields your helpdesk needs to process a ticket (category, urgency, short description, contact)

Try it now — the prompt

Run this prompt in Copilot Chat to scaffold the persona, topics, and triage decision flow — it works because it makes the answer-vs-ticket boundary explicit, which is the decision the whole agent turns on.

Two questions to resolve with the helpdesk team first:

  1. What are the top 20 requests that don't actually need a human? These should be answered by knowledge, not a ticket.
  2. What information must be on every ticket before the team picks it up? Those are the fields your triage topic collects.
Design a Copilot Studio IT helpdesk triage agent.
Knowledge base: [our IT SharePoint KB / these documents].
Top self-serviceable requests: [list your 20].
When a ticket is needed, collect: category ([hardware / software / access / other]),
urgency ([blocking / degraded / low]), and a short description.
Then submit via Power Automate to [ServiceNow / Jira / our SharePoint list].

Give me: persona instructions, 3 topics to configure, and the triage
decision flow — when does the agent answer vs. when does it create a ticket?

Step by step

  1. Create the agent. Name it (e.g. "IT Help"), write instructions that set the triage behaviour:

    You are the IT Help agent for [Company]. First, try to answer the user's request from the IT knowledge base. If the knowledge base covers it, answer with a clear resolution step and the source article. If the issue needs a human to resolve — or if the user says the KB answer didn't work — guide them through raising a support ticket.

  2. Add the knowledge source. Point the agent at your IT KB (SharePoint site or uploaded articles). Test the top-20 self-serviceable queries directly in the Test pane before building any topics.

  3. Build the triage topic. This is the core of the agent. The flow:

    • User describes an issue
    • Agent checks if knowledge base covers it → if yes, answers
    • If the user says it didn't work, or the issue is clearly action-required (e.g. "I need a new laptop") → moves to ticket collection
    • Collects: category, urgency, description
    • Confirms the details with the user before submitting

  4. Connect the Power Automate action. Build a flow that takes the collected fields and creates the ticket in your helpdesk system (or posts to a SharePoint list if no dedicated tool). Pass back the ticket reference number.

  5. Build the confirmation topic. After ticket creation, the agent confirms: "I've raised a ticket for you — reference [number]. The IT team will be in touch within [SLA]. Is there anything else?"

  6. Test end-to-end. Walk the happy path (KB answer works), the escalation path (KB answer doesn't help → ticket raised), and the direct path (user asks for something that clearly needs a ticket immediately). Confirm ticket numbers appear in the target system.

Screenshots

We deliberately don't ship screenshots that go stale — the Microsoft Copilot UI changes often. Follow the numbered steps above, which we keep current. Maintainers can regenerate fresh captures with the Playwright tool in tooling/screenshots/.

Make it better

  • SLA awareness. Update the confirmation message with the actual SLA by urgency tier ("blocking issues are typically picked up within 2 hours"). This reduces "where's my ticket?" follow-ups.
  • Known issues broadcast. Add a topic that checks a "known issues" SharePoint page — if the user's issue matches a known outage, the agent can say "we're already aware and working on it" before they raise a duplicate ticket.
  • Closed-loop follow-up. Connect a Power Automate flow that messages the user when their ticket is resolved. The agent doesn't have to be the endpoint — it can be the start of a loop.

Watch out for

  • KB article quality. The agent surfaces what's in the articles — if KB articles are incomplete, outdated, or written for IT engineers rather than end users, the answers will reflect that. Run a plain-language review of your top-20 articles before launch.
  • Collecting too much in the triage topic. Start with the minimum fields the helpdesk actually uses. Every extra question increases drop-off. Add fields after launch when you see what the team asks for repeatedly.
  • Power Automate error handling. Test what happens when the PA flow fails (connector down, authentication error). The triage topic should catch failures and offer a fallback: "I wasn't able to raise the ticket automatically — please email [helpdesk@] with these details."

Where this leads (the ramp)

Answering from the KB and raising clean tickets is a strong tier-1 deflection play. When ticket volume is high and you need to prove — continuously — that the agent resolves rather than misroutes, Azure AI Foundry's evaluation and monitoring is the upgrade.

Next: Foundry: evaluate and monitor continuously

Ready to build? Use the solution template.

The IT Helpdesk Triage Agent solution template gives you the system prompt, triage topic flow, Power Automate action spec, test cases, and deployment checklist for this exact pattern.

Was this page helpful?

Found an issue or want to suggest an improvement? Open a GitHub issue  ·  Star on GitHub