Your employees already live in Microsoft Teams, so that’s where an IT agent should meet them. Set up correctly, InvGate Service Management’s Virtual Service Agent answers questions and resolves requests right inside a Teams chat, before anyone opens a ticket. Here’s how to turn it on, give it the knowledge it needs, plan the handoff to your team, and track what it actually solves.
What is a virtual service agent?
A virtual service agent is an AI tool that interacts with users in natural language to handle IT requests, answer questions, and resolve common issues. Unlike a traditional chatbot, which primarily provides information or follows predefined conversational flows, a virtual service agent can use ITSM workflows, knowledge sources, and connected systems to perform authorized actions, resolve requests, and escalate cases to human agents when needed.
What your virtual agent will do inside Microsoft Teams
Once it’s live, employees can do all of this without leaving a Teams conversation:
- Create new requests and retrieve existing ones.
- View request details and their My Requests list.
- Add and read comments, and get notified when something changes.
- Accept or reject approval requests.
- Ask questions in natural language and get answers pulled straight from your knowledge base.
How to set up a virtual agent in Microsoft Teams, step by step
Follow these five steps to set up InvGate Service Management Virutal Service Agent g in Microsoft Teams.
Step 1: Turn on the agent and connect it to Teams
First, enable the AI service under Settings → AI Hub → AI Service, selecting InvGate AI Service. Then go to Settings → Integrations → Virtual Service Agent and add Microsoft Teams as a channel. Connecting InvGate and Microsoft Teams requires app and your Microsoft Entra Tenant ID; you’ll register the app and add your Tenant ID on the Microsoft side to finish the connection.
You will also provide a name and avatar for the agent so it reads as part of your IT team rather than a generic bot.
Step 2: Feed it the knowledge it answers with
This is the step that decides how much the agent can actually resolve: it only solves what it can find. Three things feed it.
Start with your knowledge base. Keep articles current and give them clear, literal titles that match how people describe their problems — “Reset your VPN password,” not “Remote access policy v3.” The agent reads these articles and builds a short summary of each one, so the cleaner your articles, the better its answers.
Then there’s Knowledge Discovery, which looks at resolved tickets and pulls out Snippets: small, reusable answers to problems your team has already solved. Nothing goes live on its own — an admin reviews, edits, accepts, or rejects each Snippet one by one, so you stay in control of what the agent is allowed to say.
The more of your real, recurring problems you cover here, the fewer conversations end in a handoff.
Step 3: Help it understand what people ask for
The agent also needs to route each request to the right place. Give your categories and fields clear names and descriptions, because the agent uses that wording to read intent. To sharpen this, AI Hub’s Keywords Generation suggests keywords for each catalog category, so when someone describes a problem in their own words, the agent maps it to the correct category and gathers the right details up front.
Step 4: Decide what happens when the agent can’t solve it
Some requests need a person, and the handoff should be invisible to the employee. When the agent can’t resolve something, it creates the request with the full conversation attached and drops it into the assignment and approval rules your service desk already runs.
Whoever picks it up sees the whole thread, so the employee never has to explain the problem twice. Decide in advance which categories should escalate and to whom, so these handoffs land with the right team on the first try.
Step 5: Test it and measure what it resolves
Before you open the agent to the whole company, test it against your most common tickets: pull the top request types from the last few months and ask them the way an employee would. Fix the gaps you find, then roll it out.
Once it’s live, the Virtual Service Agent report in AI Hub Reports (Reports → AI Hub) is where you see whether it’s working with AI service chat KPIs — tickets deflected, conversation volume, adoption by channel, and the topics people ask about that your knowledge doesn’t yet cover. Treat that last list as your backlog: it tells you exactly which articles or Snippets to add next.
Frequently asked questions
Does InvGate use Teams conversations to train AI models?
Conversations in Teams are kept out of the training of third-party language models, and customer data is never shared or made public — everything is stored in isolated, encrypted environments. The one case where customer data trains anything is InvGate’s own non-generative AI features: when one of those needs to learn from your data, that training happens per customer and only for your use. Your data is never pooled with anyone else’s.
Does the virtual agent work outside Microsoft Teams?
Yes. The same agent runs in WhatsApp, Slack, and in InvGate Service Management’s self-service portal.
