Integrations built on Microsoft Graph: what the outage teaches about custom software
A lot of business software talks to Microsoft 365: a CRM that sends email, a planning tool that reads calendars, a portal that lets users sign in with Microsoft. During the November 2026 outage, many of those integrations started returning errors. Here you can read what happened with Microsoft Graph, how to tell whether your own software was affected, and how integrations can better withstand the next outage.
Microsoft reports that Graph requests for mail data, calendars and contacts may time out. Software that relies on them will then throw errors or hang, unless it has been built to absorb downtime.
Below you will find what was affected, what to check now and which design choices make the difference. Status as of Friday 13 November 2026, 10:15.
What happened with Microsoft Graph
- Microsoft Graph is the gateway through which software accesses mail, calendars, files and users in Microsoft 365.
- According to Microsoft, Graph requests for Exchange data could time out, and with them the management of teams, channels and notifications.
- Well-known AI assistants also reported that their integration with Microsoft 365 was not working.
- Software that retries properly, uses queues and keeps the user informed copes with an outage like this far better.
- After recovery, check whether messages and syncs that failed during the outage have since been processed.
Microsoft Graph is the API that software outside Microsoft uses to talk to Microsoft 365. If you send email from your own system, put appointments in a calendar or manage users and teams, you usually do so through Graph. On Microsoft's status page, Microsoft reports that Graph requests that access Exchange Online data, such as mail, calendar and contacts, may time out. That also affects the management of teams, channels, members and apps, and the sending of notifications in Teams via Graph.
It was not limited to internal business software. According to ITdaily, Anthropic and OpenAI also reported that the integrations of Claude and ChatGPT with Microsoft 365 were not working. That shows how widely Graph is used: any service that looks into Microsoft 365 on behalf of a user needs this layer.
Was your own software affected?
Not every integration notices the outage. Microsoft refers to Graph requests for Exchange data; an integration that only reads files in SharePoint may work fine. Check per integration which data it uses. How the services themselves were affected is covered in the articles on Teams and Intune, Defender and Copilot.
Typical integrations and the likelihood that they were affected, based on Microsoft's report, status as of 13 November 2026, 10:15
| What the integration does | Likelihood of problems | What to check |
|---|---|---|
| Sending or reading email from your own software | High: Microsoft explicitly mentions mail data | Error logs, queue of unsent messages |
| Creating or reading appointments in calendars | High: calendars are stored in Exchange Online | Failed appointments, double bookings after recovery |
| Managing teams, channels or members, notifications in Teams | Present: Microsoft mentions this as a consequence | Whether notifications were posted after all |
| Reading or writing files in SharePoint or OneDrive | Smaller: opening files works, according to Microsoft | Sharing with external users and email notifications |
| Letting users sign in via Microsoft | Not listed as affected for Entra ID | Sign-in errors in your own logs |
The quickest way to find out is to check your software's logs for errors towards graph.microsoft.com since the afternoon of Thursday 12 November. A sudden rise in timed-out requests, 500-series error codes or messages saying Graph is throttling the request all point to the outage. If you have no log that shows this, that is a lesson in itself.
Design choices that make the difference
An integration with Microsoft 365 silently assumes Microsoft is always available. Most of the time that holds true. But software built for the day it doesn't stops an outage at Microsoft from becoming an outage in your own process. It comes down to a handful of choices that belong in every good integration.
- Retrying, with patience
When something fails, don't give up straight away and don't keep hammering the service. Retry at increasingly longer intervals, and respect Graph's guidance on how long to wait.
- Queue the work
An email or appointment that can't be sent right now goes into a queue and is processed as soon as Microsoft responds again. Nothing is lost, and nobody has to re-enter anything.
- Keep the user honestly informed
Show that something will be sent later rather than a technical error message or a spinning wheel. That way the user knows where they stand.
- Keep the core process separate
Accepting an order or creating a schedule must carry on, even if the email confirmation has to wait a while. Make Microsoft 365 a channel, not the foundation of your process.
If you let an AI assistant or agent work with email, calendars or files via Microsoft 365, the same applies. Make sure a failed step isn't left half-completed, and that the assistant reports it couldn't do something instead of making up an answer.
After recovery: what to check
Once Graph is responding normally again, the most important work begins: finding out what went wrong during the outage. Search your logs for requests that failed for good. An email that was never sent, an appointment missing from the calendar or a sync that stopped halfway goes unnoticed until a customer calls.
Also watch for the opposite: software that kept retrying during the outage may send or book twice after recovery. Check that your integration prevents the same action from being carried out twice. And see whether the outage exposed a blind spot in your alerting: did you know something was wrong before your users told you?
For the broader question of how dependent your processes are on a single supplier, read the article on dependency and business continuity. To find out how Appfront builds Microsoft 365 integrations, see Custom Microsoft 365 integration development. For a timeline of the outage, read the outage overview.
design choices that help an integration ride out an outage: retries, a queue, an honest notification and a separate core process.
types of Exchange data that Microsoft mentions in connection with timed-out Graph requests: email, calendars and contacts.
the incident in the admin centre under which Microsoft tracks the impact on Graph.
Frequently asked questions
About integrations with Microsoft 365 and Microsoft Graph during the November 2026 outage.
Is Microsoft Graph working during the outage?
Partly. Microsoft reports that Graph requests that access Exchange Online data, such as email, calendars and contacts, may time out. This also affects the management of teams and channels and notifications in Teams.
Why isn't my Outlook integration working?
An integration that uses email or calendars talks to Exchange Online via Microsoft Graph. That service has been disrupted since 12 November 2026, so requests may time out or return errors.
How can I tell whether my software is affected?
Check your software's logs for errors towards graph.microsoft.com since the afternoon of Thursday 12 November, such as timed-out requests or 500-range error codes.
Do messages get lost if the integration fails?
That depends on how the software is built. With a queue, they are still processed later. Without a queue, they can be lost; check this after recovery.
How do I make an integration resilient to outages?
Have the software retry at progressively longer intervals, put work in a queue, inform the user honestly and keep the core process separate from Microsoft 365.
Did my AI assistant also stop working with Microsoft 365?
It may have. According to ITdaily, Anthropic and OpenAI also reported that the integrations of Claude and ChatGPT with Microsoft 365 were not working.
Further reading
Integrations that withstand an outage
Appfront builds custom software and integrations with Microsoft 365, from Outlook and Teams to SharePoint and Entra ID. Want to know how your own integrations behaved during this outage, and what could be improved? Schedule a conversation with a developer.