One of the biggest challenges I hear from customers is staying on top of Message Center updates. Between Copilot announcements, Teams changes, and admin center tweaks, it’s hard to tell what’s relevant and what’s just noise.
I’ve been testing a solution that helps: a Message Center agent built by Mike Francis here from Microsoft and shared on GitHub. It’s a really great alternative to the built-in Copilot in the Microsoft Admin Center, that only focuses on Message Center and it’s worth a look.
What is the Message Center Agent
The Message Center Agent, built by Mike Francis, is an agent you can deploy that does one job, allows you to search the Microsoft 365 Admin Center messages with Copilot chat using natural language prompts.
Not only can admins and 'Message Center Readers' search for messages, but you can also use the power of generative AI to summarize them and draft internal communications. Here are some examples of what you can do:
- Search for messages: Find specific messages in the Microsoft 365 Admin Center message center. Search by keywords, date ranges, and message types. Search is case-insensitive. Search by message Title (default) or message body.
- Summarize messages: Get a summary of the latest messages in the message center.
- Draft internal communications: Generate draft emails or messages to share information about updates with your team.
- Get insights: Gain insights regarding impact of updates for your organization.
- Get alternatives Get suggested alternatives for deprecated features.
- Get suggested actions: Get suggested actions for updates that require your attention.
How to Deploy It in Your Tenant
In the video I walkthrough of how I deployed the agent. Full instructions are on GitHub and I go through all of them live in this video!
You can read all the instructions here, but I have a high level view below: https://github.com/mjfusa/Message-Center-Agent
- Prerequisites
You’ll need the right roles:
- App registration: Application Administrator (least privilege), or Global Admin
- Agent provisioning: Teams Admin or Global Admin
- Agent usage: Message Center Reader, Service Support Admin, Service Admin, or Global Reader
Also:
- Enable custom app uploads in Teams Admin Center
- Install PowerShell modules and Node.js
- Use the Microsoft 365 Agents Toolkit (formerly Teams Toolkit)
- Clone the Repo and Set Up
- Download the repo from GitHub
- Extract it to a local folder (don’t use OneDrive—trust me)
- Run the PowerShell scripts in the prereqs folder to:
- Set execution policy
- Install Node and the toolkit
- Register the app in Entra and generate client ID/secret
- Deploy the Agent
- Switch to the root folder
- Run the deployment script
- Enter your client ID and secret
- Test the agent in Copilot (before pushing to production)
- Publish to Production
- Run the final script to publish the agent
- Go to Teams Admin Center → Manage Apps → Search for “Message Center”
- Approve and assign to users or groups
- Optionally manage it in Microsoft Admin Center under Copilot → Agents
Troubleshooting or Resetting
If something breaks or you need rinse and reset and start from stratch (e.g., auth errors, agent not loading), here’s how to reset:
- Uninstall the agent from Copilot
- Delete the app registration in Entra
- Remove the app from the Teams Developer Portal
Github Repo: https://github.com/mjfusa/Message-Center-Agent
My Ending Thoughts
While it’s not a full solution yet, a lot of customers are looking for more automation in this type of solution. I really do believe this is a strong step forward, it helps cut through the noise and highlights the announcements that actually matter to your team. Plus, the ability to use the agent to find and interact with posts in new ways opens up some creative possibilities.
In a conversation with Mike Francis, we even explored ideas like building a Graph connector and using Planner to automate tasks and announcements for org teams. While filming and editing, I kept thinking about how we could push this further, like integrating multi-modal agents to automate even more.
A big thanks to Mike Francis for building and sharing this. If you try it out, let me know what you’re doing with it. I’d love to see how others are using it in the field.
Microsoft