Step 7: Deploy to Teams

7. Deploy and test in a Teams channel

Previous step: 6. Excel data filtering (advanced)

Start over: README

Final checks before deployment

Before deploying the agent to Teams, review the checklist below.

• Agent Instructions saved
• SharePoint Knowledge source connection verified
• Email connector authentication completed
• ThinQ MCP tool connection verified
• Work IQ MCP tool connection verified
• AI Prompt tool (HTML_Response_Generator) saved
• Basic behavior verified in the test panel

1. Publish the agent

1-1. Start publishing

Click the Publish button in the upper-right corner of the agent overview page.

1-2. Confirm publishing

Review the contents in the publish popup and click Publish.

Publishing saves the agent’s current settings and makes it ready for deployment.
You can continue editing the agent after publishing. Republish after editing to apply changes.

Publishing may take 1-2 minutes to complete.

2. Deploy to a Teams channel

2-1. Configure channel deployment

After publishing is complete, go to Settings → Channels in the left menu.
Alternatively, find the Channels section on the agent overview page.

2-2. Select the Microsoft Teams channel

From the channel list, select Microsoft Teams.

2-3. Configure the Teams channel

2-4. Add the agent to Teams

Click the Open in Teams button, or add the agent to Teams using one of the methods below.

Method A — Add from the app store (for organization deployment):

1. Open the Microsoft Teams app
2. Click Apps in the lower-left → Built for your org tab
3. Search for the agent name and click Add

Method B — Add with a direct link:

1. Copy the direct Teams add link provided by Copilot Studio
2. Paste it in Teams chat and install

2-5. Register the agent in a Teams channel

To register the agent in a specific Teams channel:

1. Go to the target Teams channel
2. Click + (add a tab) at the top of the channel
3. Select App tab or search for the agent
4. Complete installation

3. Test behavior in Teams

After the agent is added to Teams, validate the end-to-end flow with the scenarios below.

Scenario 1 — Look up appliance information (ThinQ MCP)

@[Agent name] Recommend an LG robot vacuum model

→ ThinQ MCP lookup → HTML card response

Scenario 2 — Today’s work briefing (Work IQ MCP)

Summarize my schedule and tasks for today

→ Work IQ MCP lookup → HTML report card response

Scenario 3 — Search internal documents (SharePoint)

Find internal documents related to [search keyword]

→ SharePoint search → response with citations

Scenario 4 — Look up and send email

Check my recent emails and send my team lead an email saying today's work is complete

→ Email lookup → draft creation → send after approval

Scenario 5 — Personalized content (Excel + AI Prompt, advanced)

I'm a new employee. Teach me how to use Teams

→ Excel filtering → Personalized_Content_Generator → HTML response

4. (Optional) Copilot Studio monitoring

If you have time left in the lab, review conversation logs in Copilot Studio.

4-1. Open the monitoring tab

Copilot Studio → select the agent → click Monitoring in the left menu

4-2. Items to check

💡 Monitoring tips

• If the agent calls the wrong tool or does not follow the Instructions, analyze the cause in the conversation logs and update the Instructions.
• If the tool call order is not what you intended, make the workflow description in the Instructions more specific.

Lab complete 🎉

Congratulations! You have completed all steps.

Here is a summary of what you implemented in this lab.

Key takeaway: You can implement complex business automation using only Instructions + tools + MCP, without Flows, variables, or Topics.
The quality of an agent’s autonomous orchestration is proportional to the specificity of its Instructions.

← Back to README