Outlook and Teams Went Dark for a Day. What Is Your Plan B?
A botched network change took down Outlook, Teams and Copilot across dozens of countries, in a month that also saw Apple Pay and the FAA go down. Microsoft says it was not a cyberattack or AI. For a small startup, the cause matters less than the question: what stops working at your company when one vendor does? Here is a one-page plan.

Think about your company's last 24 hours. How many of them ran through one vendor's login?
On Thursday, November 12, people around the world stopped being able to open Outlook. Teams and Copilot followed in part. Within about an hour, Downdetector had reports from 74 countries and territories, according to Al Jazeera. By the next morning in Europe, servers Microsoft had brought back were failing again.
It was not an isolated week. TechCrunch counted a run of outages in the past two weeks: Apple Pay went down, NASA's internal communication network went down, and the FAA grounded all flights for hours.
Big companies have teams whose job is to plan for this. Most five-person startups have a shared Google doc, if that. You do not need a disaster recovery department. You need one page and one afternoon. Here is how to fill both.
What happened at Microsoft
Here is the sequence, as reported:
- Start: Problems began around 16:00 GMT on Thursday, November 12, when users could not reach Outlook, per Al Jazeera.
- Scale: By about 07:00 GMT Friday, Downdetector showed at least 43,000 Outlook problem reports from 72 countries. Microsoft did not confirm user numbers. Government systems in Singapore and Japan were among those disrupted, and Al Jazeera reported that integrations used by OpenAI and Anthropic were also hit.
- Duration: Windows Latest reported that most servers were back online after almost 20 hours. TechCrunch described Outlook, Teams and Copilot as down for nearly two days. Recovery was uneven, so both can be true for different users.
- Cause: In an admin center statement quoted by Windows Latest, Microsoft said the incident was introduced by an issue during an earlier network capacity upgrade that surfaced after a recent network change. It added that this was a technical issue not related to cyber or AI and agents.
Microsoft paused all deployments, moved traffic and restarted servers. Some servers then failed automated health checks, which pushed more load onto the ones still running.
The lesson inside the fix: failures spread
That last detail is the part worth remembering. A server fails a health check, gets pulled out, and its traffic lands on the remaining servers. They overload, fail their own checks, and the problem grows.
Google's Site Reliability Engineering book calls this a cascading failure: a failure that grows over time through positive feedback. Its advice includes limiting retries, shedding load before servers fall over, and testing what happens when part of the system disappears.
You may not run servers at that scale. But the same pattern shows up in small companies:
- Your email is down, so customers cannot get password reset links, so they flood support, which runs through the same inbox.
- Your payment provider is slow, so your checkout retries, so you get double charges, so you get chargebacks.
- Your AI provider is down, so every feature that calls it hangs, and your whole app looks broken instead of just one button.
The fix in each case is the same idea: decide ahead of time what should fail quietly, and what should never depend on one thing.
Step one: list your single points of failure
Open a blank sheet and make four columns: Service, What breaks if it is down, Who notices first, Backup.
Go through your company in this order:
- Communication. Email, chat, video calls. If Outlook and Teams vanish, how does your team talk? How do customers reach you?
- Identity. Single sign-on. If your login provider fails, can anyone get into anything?
- Money in. Payment processor, invoicing, bank. Apple Pay going down is a reminder that even the biggest wallets fail.
- Product. Hosting, database, AI model provider, email delivery for your app.
- Records. Where your contracts, customer list and financials live.
For most early teams, three to five rows will have an empty Backup column. That is normal. The goal of this exercise is not to fill every row. It is to know which empty rows would cost you a customer or a payroll.
Your AI co-founder is ready when you are.
Foundra turns everything in this article into an actual plan. Validation, customers, pricing, launch. In one place, in your voice, in an afternoon.
Get started→$39/month. Cancel anytime.
Step two: write the one-page continuity plan
For each row that could cost you real money, write three lines.
The fallback. Keep it cheap. A personal phone group chat the team agrees to use if work chat is down. A second payment link from a different processor kept in a drawer. An exported customer list from last Friday.
The trigger. Who decides to switch, and after how long? "If work chat is down for 30 minutes, the CEO posts in the backup group."
The customer message. Write the status update now, while calm. Two sentences: what is affected, when you will update next. Post it on a status page or social account that does not run on the vendor that is down.
Keep the plan somewhere you can reach when your main tools fail. A printed copy and a file outside your main workspace both count. Some teams keep it next to their runway model and investor updates in a planning tool like Foundra, as long as that tool sits outside the stack that might fail.
Step three: make your own product fail politely
If you build software, the outage month is a prompt to look at your own product through a customer's eyes.
- Time out, do not hang. Every call to an outside service should have a time limit. Ten seconds of spinning followed by a clear message beats a frozen screen.
- Degrade one feature, not the app. If your AI summary feature cannot reach its model, hide the button and show the rest of the page.
- Cap retries. Retrying forever is how a small blip becomes your own cascading failure. Use a few retries with growing gaps between them.
- Have a kill switch. A setting that turns off a feature without a code deploy. Microsoft's first move was pausing deployments. Yours might be switching off the one feature that depends on the vendor that is down.
- Test it once. Block one outside service in a test environment and click through your app. You will find at least one surprise.
Is this a threat or an opening?
Experts quoted by TechCrunch disagree on what is driving the run of outages, and investigations are ongoing.
Dave DeWalt, a cybersecurity investor and former McAfee CEO, thinks the outages are probably AI-related in an indirect way: AI is surfacing vulnerabilities faster, companies are patching quickly, and that rush is disrupting systems in the short term. He expects demand for startups in AI-driven vulnerability management, automatic fixes, safe patch testing and resilience for critical systems.
Anusha Iyer, founder of security startup Corsha, pointed to the FAA's modernization, where new systems overwhelmed older hardware. Her advice: segment legacy systems more tightly, control access to them, and phase new systems in.
Both views lead to the same founder takeaway. If your product pushes changes into a customer's systems, staged rollouts, easy rollback and clear status updates are now selling points. Put them on your sales page, and be ready to show them in a security review.
After any outage: three small habits
Ask for the credit. Windows Latest notes Microsoft 365 business customers can request service credits under Microsoft's service level agreement. Most vendors offer something similar, but few send it automatically. Check each contract.
Write a short post-incident note. What broke, how long, what the customer saw, what you will change. One paragraph is enough. Do it within a week while memories are fresh.
Re-check the plan every quarter. Your stack changes fast in the first two years. A backup you set up in March may point at a tool you cancelled in June.
Frequently asked questions
What caused the November 2026 Microsoft 365 outage? Microsoft said the incident came from an issue during an earlier network capacity upgrade that surfaced after a recent network change. It said the outage was not related to a cyberattack or to AI and agents.
How long was Outlook down? Reports differ. Windows Latest said most servers were back after almost 20 hours. TechCrunch described Outlook, Teams and Copilot as down for nearly two days. Recovery varied by user and region.
What is a cascading failure? It is a failure that grows through feedback, such as overloaded servers being removed and pushing their traffic onto others until they fail too.
Does a small startup need a business continuity plan? Yes, but a short one. List the services that would cost you customers or payroll if they failed, then write a fallback, a trigger and a customer message for each.
Can I get money back after a vendor outage? Often, in the form of service credits. Microsoft 365 business customers can request them under its service level agreement. Check your own contracts, because credits usually must be requested.
Were the recent outages cyberattacks? The Director of National Intelligence said they do not seem related to state-sponsored terrorism, and many companies said the causes were technical. Investigations are ongoing.
Sources
- These tech outages are giving Y2K, AI agent style, TechCrunch, November 19, 2026
- Microsoft hit by global outages: What happened and why?, Al Jazeera, November 13, 2026
- Microsoft doesn't want you to blame AI for the Microsoft 365 outage, insisting it was a botched network upgrade, Windows Latest, November 14, 2026
- Addressing Cascading Failures, Site Reliability Engineering, Google
You just read the theory. Ready to build the thing?
Foundra is your AI co-founder. It turns an idea into a validated business plan, a go-to-market, and your first 10 customers. In an afternoon, not a semester.
$39/month. Cancel anytime. Works in 20 languages.