News analysis · Microsoft 365 outage
Microsoft 365 was down for a day. Your AI went down with it.
The Nov. 12-13 Microsoft 365 outage took down Outlook, Teams and Copilot for more than a day. Microsoft says AI didn't cause it, but every AI tool that works through your inbox stopped too. What it means for an order desk, what didn't change, and four things to set up before the next one.

In brief
On Thursday, Nov. 12, a Microsoft network problem knocked out Outlook, Exchange email, Teams and Copilot for many businesses, and the disruption lasted more than 24 hours. Microsoft says it was not a hack and not caused by AI. But anything that works through your email, including Copilot and outside AI tools connected to Microsoft 365, stopped with it. Don't switch vendors over one outage. Do write down what depends on email, give customers a backup way to send orders, and make sure your automations catch up cleanly when mail comes back.
- The Nov. 12-13 outage took down Outlook, Exchange email, Teams and Copilot for more than 24 hours.
- Microsoft says it wasn't a hack and wasn't caused by AI, but AI tools that read your mailbox stopped with it.
- Delayed mail was held for up to 48 hours, so the backlog can arrive all at once.
- Don't switch vendors over one outage. Give customers a second way to send orders.
- Ask how your automations handle a day of backlog: in order, no duplicates, late items flagged.
What happened
Starting Thursday, Nov. 12, businesses around the world lost access to Outlook, Exchange email, Teams and other Microsoft 365 tools. CBS News reported that more than 130,000 users reported problems on Downdetector on Thursday, that the disruption lasted over 24 hours, and that users reported Copilot down along with email. Microsoft's own status page for its consumer services gives a start time of 10:05 a.m. UTC Thursday and an end time of 4:10 p.m. UTC Friday. News reports put the start for most U.S. users around midday Eastern time.
Microsoft's explanation, as reported by ABC News: it was "not related to cyber or AI / agents." Microsoft said the problem came from an earlier network capacity upgrade that surfaced after a recent network change. Its status page, as quoted by Ars Technica, said it had no evidence of malicious activity and promised a full post-incident report. That report wasn't out as of Saturday morning.
Why this is an AI story anyway
AI didn't cause the outage. But look at what stopped. CBS reported that OpenAI and Anthropic each had problems connecting to Microsoft 365 tools. CNN reported that AI companies reported problems too.
That's the part that matters for a mid-size office. More and more AI work runs through email: an agent reads the orders@ inbox, Copilot summarizes a customer thread, a tool drafts the reply to an order status question. When the mailbox is down, all of that stops at once. The AI is only as available as the system it reads from.
And when mail comes back, it doesn't trickle in. Microsoft said it extended the email timeout to 48 hours to preserve delayed mail. That's good news: messages waited instead of bouncing. It also means a day's worth of orders, invoices and questions can land in a few hours. A person can sort that pile. An automation built for a steady flow may not.
The AI is only as available as the inbox it reads from.
How big was it, really?
Estimates vary, and the fair reading is "big, but not equal for everyone." Anderson Economic Group, a Michigan think tank, estimated the lost productivity could approach at least $7 billion, as reported by CNN. The same firm said many of Microsoft's users were likely dealing with annoyances, not real losses, and put the number with real disruption much lower.
Microsoft 365 users vs. users with real disruption
Millions of users. The disruption figures are a think-tank estimate, not a count.
Sources: CBS News (430 million, citing Microsoft's website); CNN (Anderson Economic Group estimates, "perhaps as many as").
CBS also talked to a manufacturer: the owner of an aluminum company in Hong Kong said that if it dragged on, the outage could delay production on orders from U.S. customers. That's the order desk risk in one sentence.
What doesn't change
Be fair to Microsoft and to the cloud. This was not a breach. Your data wasn't stolen, and delayed mail was held, not lost. Tony Redmond, whose site has recorded Microsoft 365 uptime figures since 2013, writes that uptime for Microsoft's online services is usually strong. Running your own email server in a closet would not have a better record for most companies with 40 to 400 people.
So the lesson isn't "leave Microsoft." Prof. Steven Watt, who teaches cybersecurity at the University of St Andrews, made both points after the outage: walking away from the cloud isn't realistic or desirable, but a continuity plan that runs on the same platform that just failed isn't a plan. He added that the stakes rise as AI agents start acting inside these same platforms.
Can you get money back?
Possibly, but don't count on it. Microsoft's service credits apply when a single service, like Exchange Online, drops below 99.9% uptime in a month. Redmond notes that claims can't be made on quarterly performance or across multiple services, and that many tenants don't bother because of the work involved in gathering evidence. If you want to try, ask your Microsoft partner or reseller now, while the dates are fresh. The credit, if any, will be small next to a day of lost orders.
Four things to set up before the next one
- Write down what depends on Microsoft 365. Email, Teams, shared files, Copilot, and every tool that reads a mailbox: order entry, AP invoice capture, status replies. If you have AI agents, list the inbox each one reads.
- Give customers a second way to send an order. A phone number they already know, a portal, or a backup address on a different provider. Tell your top customers now, not during the outage.
- Make your automations catch up cleanly. When mail returns, an order agent should work through the backlog in order, skip duplicates (customers often resend), and flag anything that arrived late so a person checks the ship date. Ask whoever built yours how it handles a 24-hour backlog. If they don't know, test it.
- Keep a person's path open. If your order desk can still take an order by phone and enter it into the ERP by hand, you lose speed, not orders. That's one reason we keep a person approving every order in any automation, as we explain in human-in-the-loop automation.
If you already rely on Microsoft's AI, our page on Copilot vs. a custom agent for order entry covers how the pieces fit, and Microsoft Copilot billing going metered by default covers the Dec. 1 billing change.
Common questions
What caused the Microsoft 365 outage on Nov. 12, 2026?
Microsoft said it was a technical issue from an earlier network capacity upgrade that surfaced after a recent network change. It said the outage was not related to cyber or AI and that it saw no evidence of malicious activity. A full post-incident report was promised.
How long was Microsoft 365 down?
CBS News reported the disruption lasted over 24 hours. Microsoft's consumer status page lists 10:05 a.m. UTC Thursday, Nov. 12 to 4:10 p.m. UTC Friday, Nov. 13. Timing varied by company and region.
Did Copilot and other AI tools stop working?
Yes. Users reported Copilot down along with Outlook and Teams, and CBS reported that OpenAI and Anthropic each had problems connecting to Microsoft 365 tools. Any AI that reads your mailbox stops when the mailbox does.
Was email lost during the outage?
Microsoft said it extended the email timeout to 48 hours to preserve delayed mail. Messages waited and arrived late rather than bouncing, which can mean a large backlog when service returns.
Can we get a refund or service credit?
Microsoft's service credits apply when a single service falls below 99.9% uptime in a month. Claims can't combine services or use quarterly numbers, and many tenants don't pursue them. Ask your Microsoft partner or reseller.
Should we move off Microsoft 365?
For most mid-size companies, no. This wasn't a breach and Microsoft's uptime is usually very good. Instead, list what depends on it, give customers a second way to send orders, and make sure automations can catch up on a backlog.
Sources
- Microsoft 365 Status, consumer service history, checked Nov. 14, 2026
- CBS News, Microsoft outage coverage, Nov. 13, 2026
- ABC News, Outlook down: Microsoft outage, Nov. 13, 2026
- CNN, Microsoft global outage live coverage, Nov. 12, 2026
- Ars Technica, Microsoft blames failed server health checks, Nov. 2026
- Science Media Centre, Expert reaction to Microsoft outage
- Office 365 for IT Pros (Tony Redmond), Microsoft 365 uptime, Apr. 22, 2026
Microsoft 365 Copilot vs. a custom agent for order entry
Human-in-the-loop automation: why a person should approve every step
Microsoft Copilot billing is going metered by default. What to set before Dec. 1