We are rolling out a very large Office365 tenant and I've been thinking about ways to manage the user communities on our tenant.
Machine learning is also something I am interested in, and my recent discovery of AzureML studio also fed into this project idea
What it does
Our app grabs actual user audit log data from O365 and uses that as the basis for clustering that user information. Then we survey users, and use a BI dashboard to learn more about the type of users in a cluster, so that we can assign a useful name to that cluster. Once a cluster has been named, we create Yammer groups for the clusters (by hand), and then we run code to add everyone on the tenant to the Yammer group for their cluster
How we built it
We used only Office365 tools. With the exception of some PowerShell to upload data to SharePoint, this whole app was built in the browser and in the PowerBI client.
Challenges we ran into
Adding users to the group using Flow was complicated. Getting clustering to work took a lot of experimentation.