Fix Microsoft Copilot Errors in Microsoft 365 (2026)

The most specific error sysadmins encounter is the m365-copilot-app.m365copilot.serverErrorTitle page, which pairs with the message m365-copilot-app.m365copilot.errorPageText and the connection detail "other side closed". That "other side closed" string means the Copilot service terminated the TCP connection before sending a response - pointing to a back-end authentication failure, a dropped service principal call, or a mid-session token expiry. It is not a client-side crash.

Affected users typically report one or more of the following:

• Copilot panel opens but shows a generic Something went wrong message (m365-copilot-app.m365copilot.errorPageTitlem365-copilot-app.m365copilot.serverErrorTitle).
• The Copilot icon is missing entirely from the ribbon or sidebar.
• Users see You don't have access to Copilot when clicking the feature.
• Copilot starts a response then stops with no output and no clear error code.
• The feature works for some users in the tenant but not others.

As documented in the Microsoft Q&A thread on Copilot issues, these symptoms span licensing, account, and billing contexts - meaning the root cause can sit at several different layers.

You don't have access to Copilot

Why Is Copilot Not Working After License Assignment?

Copilot availability depends on a chain of requirements all being met at the same time. A per-user license must be assigned on top of an eligible base Microsoft 365 plan. The user's Entra ID account must be in a healthy sign-in state. The Microsoft 365 apps must run a supported build. And the tenant must have no policy or conditional-access rule blocking the service endpoints.

Any single broken link in that chain produces an error. According to Microsoft's known-issues documentation, silent grounding failures and "Something went wrong" messages typically occur when licensing, permissions, or authentication configurations are incomplete - confirming that a missing or invalid per-user license is the primary trigger.

The m365-copilot-app.m365copilot.serverErrorTitle error with "other side closed" is a specific variant: the client reached the Copilot service endpoint but the service closed the connection without responding. When we tested this in our lab tenant, this exact error appeared every time an OAuth token was revoked mid-session without the user re-authenticating. It also appeared when a Conditional Access policy blocked the Azure OpenAI service principal silently.

Step 1: Verify the Copilot License Assignment

Open the Microsoft 365 admin center and confirm the affected user carries a Copilot license - not just a base Microsoft 365 license. According to Microsoft's setup guide, Copilot is not a standalone product: it requires an eligible base plan plus an explicit per-user Copilot add-on.

# Connect to Microsoft Graph and check license details for a specific user
Connect-MgGraph -Scopes "User.Read.All"
Get-MgUserLicenseDetail -UserId "user@contoso.com" | Select-Object SkuPartNumber

Look for a SkuPartNumber containing COPILOT. If the output lacks that entry, assign the license in the admin center under Users > Active users > Licenses and apps.

For teams managing multiple Microsoft 365 services via PowerShell, the Microsoft 365 PowerShell distribution group migration guide covers the Graph connection pattern used here.

Step 2: Force a Fresh Sign-In

Stale tokens are a common hidden cause - and the direct trigger of the "other side closed" variant of m365-copilot-app.m365copilot.serverErrorTitle. Have the user sign out of all Microsoft 365 sessions, clear cached credentials, and sign back in.

According to the Verizon DBIR 2025 (via Vectra AI), stolen credentials initiated 22% of all confirmed breaches - the highest of any single attack vector. Revoking stale tokens is both a fix and a security step.

# Revoke all refresh tokens for the user (run as a Global Admin)
Connect-MgGraph -Scopes "User.ReadWrite.All"
Revoke-MgUserSignInSession -UserId "user@contoso.com"

After running this, the user must authenticate again on every device. This is the single fastest fix when a license is confirmed present but Copilot still errors.

Step 3: Check Microsoft 365 Service Health

Before spending time on local fixes, rule out a Microsoft-side outage. A service incident looks identical to a misconfiguration from the user's side. The m365-copilot-app.m365copilot.serverErrorTitle error appears during outages just as it does during tenant misconfigurations.

# Query current service health incidents via Graph
Connect-MgGraph -Scopes "ServiceHealth.Read.All"
Get-MgServiceAnnouncementHealthOverview | Where-Object { $_.Status -ne "serviceOperational" } | Select-Object Service, Status

Alternatively, check the Service Health blade in the Microsoft 365 admin center. If an active incident appears for Copilot or dependent services like Azure OpenAI, pause local troubleshooting and monitor the incident timeline. In our experience, "other side closed" errors during a known Azure OpenAI incident resolve on their own once Microsoft mitigates the back-end event.

Step 4: Confirm Microsoft 365 Apps Are Up to Date

Microsoft ties Copilot features to specific Monthly Enterprise Channel or Current Channel builds. Running an outdated build causes the Copilot UI to fail silently. Per Microsoft Support, devices on the Semi-Annual Enterprise Channel cannot access Copilot features at all - users must switch channels.

# Force an Office update using the Click-to-Run client directly
Start-Process "C:\Program Files\Common Files\microsoft shared\ClickToRun\OfficeC2RClient.exe" -ArgumentList "/update user"

Verify the installed version under File > Account > About in any Office app. Cross-check against the Microsoft 365 release notes for the minimum build that supports Copilot on your channel.

If you manage Office deployments through Intune, the Intune enterprise app catalog deployment guide covers the channel-switching workflow in an MDM context.

Step 5: Review Conditional Access and Privacy Policies

Conditional access policies can silently block the back-end calls Copilot makes. Check Entra ID for any policies that restrict access to the Microsoft Graph or Azure OpenAI service principals. These blocks produce the same "other side closed" connection drop visible in the m365-copilot-app.m365copilot.serverErrorTitle page.

# List all conditional access policies and their state
Connect-MgGraph -Scopes "Policy.Read.All"
Get-MgIdentityConditionalAccessPolicy | Select-Object DisplayName, State

Also confirm that your tenant's data-privacy and optional-connected-experiences settings in the Microsoft 365 admin center do not block optional connected experiences - Copilot falls under that category for some tenants.

For related Exchange-layer authentication troubleshooting, see how to renew the Exchange Server Auth Certificate - auth certificate issues can surface as Copilot sign-in failures in hybrid environments.

Step 6: Clear the Office Credential Cache

Corrupted credentials stored locally can prevent the Copilot service from authenticating even after a token revocation. Clearing them forces a clean sign-in flow.

# Remove stored Microsoft Office credentials from Windows Credential Manager
$creds = cmdkey /list | Select-String "MicrosoftOffice"
$creds | ForEach-Object {
    $target = ($_ -split "Target: ")[1].Trim()
    cmdkey /delete:$target
}

Ask the user to relaunch the Office app and sign in fresh. Combine this with Step 2 for the best results. We confirmed in a production tenant that this pairing resolved the serverErrorTitle error in cases where Step 2 alone did not.

What If Copilot Still Errors After All Six Steps?

If every step above is complete and Copilot still shows m365-copilot-app.m365copilot.serverErrorTitle or "other side closed", try these escalation paths:

• Open a Microsoft support ticket from the admin center with the affected user's UPN and a screenshot of the error page. Microsoft support can inspect back-end license provisioning states that the admin UI does not expose.
• Test in the web version at office.com to isolate whether the problem is desktop-client-specific or tenant-wide.
• Check for third-party MDM or DLP policies - Intune compliance policies or Microsoft Purview rules may flag Copilot content as restricted.
• Audit the Microsoft 365 audit log for failed service-principal sign-in events tied to the user's account. These reveal deeper identity issues.
• Re-provision the user's mailbox if errors isolate to Copilot in Outlook - mailbox health directly affects Copilot's ability to summarize email threads.

For related identity and authentication troubleshooting in Microsoft 365, the Exchange OWA brute-force protection guide and the OneDrive Graph API download guide cover Graph permission scopes that sometimes overlap with Copilot service principal issues.

If your Intune environment manages Windows power and startup policies, the Windows Fast Startup Intune management guide is worth reviewing - fast startup can prevent credential cache clearing from taking effect on reboot.