Anyone has experience with Confluence Copilot connector?

[G_A_]

I know it might be a long shot but hoping for some luck :).

We have recently started trials of MS 365 Copilot and connected it to the native Confluence connector and following this guide - Confluence Cloud Microsoft 365 Copilot connector | Microsoft Learn - We are using the Oauth 2.0 method to approve the connection.

The problem that we are currently facing, is with the access permissions - the content in Confluence is indexed using the Site admin permissions (that account is the one with the M365 global admin and the Atlassian Org admin permissions and so the only one that can technically approve the connection) but then the content is also indexed based on the permissions of that account and not on a system level like the guide is promising. 

I am sure that the problem is on MS side as the owners of the connector, but I am trying to gather some more info to prove to their support where the issue might be. 

From the Oauth app side - any chance that someone might be catching a missing step in the guide? An action that would force the individual users in Copilot to authorize the connection to use their own account to do the search in the indexed content.

Currently, the users are able to see indexed content in Copilot that is blocked for them in Confluence, because the site admin account has access to it.

Any idea on how to improve the connector without using the service account authentication type method?

Thanks in advance

[John Funk · Accepted solution]

Hey @G_A_ 

Since no one has responded yet, you should probably create a support ticket with Atlassian to get their feedback/guidance. 

https://support.atlassian.com/contact/#/

[G_A_]

Thanks for replying. 

Its a bit of a scenario as its hard to find people with experience on both systems. 

Ill try Atlassian support and hope for some luck.

[Eric J Trumbull]

Hi @G_A_ - we use this setup and have experienced the same problem this week.

We've had this in place for a few months but this was the first instance of Page names and summaries being shared with users that don't have access to the content in Confluence.

Excited to hear if you get any info back from Atlassian or M365 support. We'll update you here if we figure anything out as well.

- Eric

[G_A_]

Hi @Eric J Trumbull ,

I have a support ticket with MS that has been escalated to a different engineer, so the issue is still open. 

Since the connector is a Microsoft product, I am guessing they will need to update application before it can be properly used - page hierarchy, folders and child pages seem to break its understanding of the Space structure. 

I'll update here if something changes.

Guy