On Friday 13 November 2026, Microsoft's email had been partly down worldwide for two days. Along with email, Copilot and the AI connectors on Microsoft 365 went down too. If your business runs email, documents and AI on a single supplier, an outage takes out everything at once. That calls for a contingency plan and a deliberate choice about how dependent you want to be.
What happened
The fault was in Exchange Online, the service behind Outlook, and in what is built on it. According to The Rundown AI on 13 November:
- Outlook, SharePoint, Copilot and Defender XDR were affected. Teams chat and meetings mostly still worked.
- In the United States, roughly 130,000 reports about Outlook came in at Thursday's peak.
- OpenAI and Anthropic reported that their Microsoft 365 connectors were broken.
- A Taiwanese financial firm went back to confirming trades by fax.
Microsoft reported encouraging recovery progress, but after more than twenty hours still could not name a root cause. How long the outage lasted in total and what went wrong was not yet known when the source was written. It is not the first this year: in January, Microsoft 365 was down in North America for nearly ten hours.
Your AI shares the fate of your platform
Copilot and the connectors from Claude or ChatGPT to your mailbox and documents only work as long as Microsoft 365 works. When email goes down, the AI layer goes down with it. A second AI tool won't help if it hangs on the same data through the same connector.
Our view: depending on one supplier is not wrong, but it should be a choice. Know what stops and how you keep working.
What you do now
- Write a contingency plan for a day without Microsoft 365. An agreed alternative channel, phone numbers of customers and suppliers outside Outlook, and who decides to switch over.
- Map your AI connectors. Note for each AI application which service it relies on and what happens if that service drops out. Critical work gets a manual fallback.
- Make the trade-off deliberately. Discuss with management which processes may run on one supplier and where you want spread or your own copy of the data.
How it works under the hood
Exchange Online is more than email: Copilot, SharePoint and Defender XDR rely on it. Recovery was uneven. Servers Microsoft brought back overnight failed again under traffic. On Friday, Microsoft began rolling back a network change; that is a recovery step, not a confirmed cause. To stop delayed mail from bouncing, Microsoft extended the timeout window to 48 hours.
If you build with agents or connectors yourself, assume an external service will drop out:
- Let a connector retry after a timeout and then stop cleanly.
- Log which source failed, so a silent failure gets noticed.
This news touches chapters 24 and 23 of the course AI(l) or nothing by Newway. If you follow the course, this subject returns there at a calmer pace.
What you can do with this
Want to know what this means for your business? Plan an introduction: half an hour, no sales pitch, but an honest answer to the question where to begin. Rather read on first? The course AI(l) or nothing is free and starts with the basics.
Written in response to a publication by The Rundown AI: https://www.therundown.ai/articles/microsoft-s-outlook-outage-hits-day-two
S.O.S