Your Business Continuity Plan Lives in the Thing That’s Down

A dashed cloud holding a plan, a chat bubble, and an envelope, with a solid copy of the plan kept outside it

On November 12, Microsoft 365 went down. Outlook, Exchange, SharePoint, Teams, and Copilot, for most of a day, in dozens of countries. Nobody attacked anything. A network change went wrong at a company you don’t run.

That is the scenario most SaaS outage business continuity planning still misses. Not because the plan was bad. Because the plan, the call tree, and the crisis team’s meeting room all lived inside the outage.

What Happened on November 12

Microsoft says the outage began at 1:26 PM Eastern on November 12. Its advisory listed Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot, and Defender XDR. Microsoft reported “broad recovery and stability” around midday on November 13, with some regions still being restored, according to BleepingComputer.

Al Jazeera reported that Downdetector logged complaints from 74 countries and territories within about an hour, including government systems in Singapore and Japan and Australian federal agencies.

Microsoft called it “a technical issue not related to cyber or AI / agents,” traced to an earlier network capacity upgrade that surfaced after a later change. It has promised a full post-incident report.

You can’t patch your way out of somebody else’s outage. You can only plan for it.

This isn’t a Microsoft problem. Swap in Google Workspace, Slack, Zoom, Salesforce, Atlassian, or your AI provider. We saw the same pattern with the AWS outage last fall, which I talked through on Managing Uncertainty, Episode 314.

The Dependencies Your Business Impact Analysis Never Asked About

A business impact analysis (BIA) asks each team what it does and what it needs to do it. People name the big systems: the claims platform, the ERP, the EHR.

Nobody names email.

This year, one of our clients had an outage at its email security provider that stopped external email. When the business continuity lead went back to the BIA, external email had never come up. Not in a single interview.

The systems everybody uses are the ones nobody lists. Three go missing most often:

  • The collaboration layer. Email, chat, video, the shared drive. Everything depends on it, so nobody thinks of it as a dependency.
  • SaaS that arrived through projects. New applications increasingly show up as SaaS, often after the decisions about them are made. Nobody owns recovering the integrations between them.
  • AI tools. Engineers at one client told us that if their AI coding tools went down, they would wait for them to come back. Copilot went down on November 12 with everything else.

Your executives’ list of critical processes is usually half the real list. The same is true of systems. That’s why validating your BIA results matters more than collecting them. Add one question to every BIA interview: what do you use every single day that you’d never think to mention?

When the Plan Is Stored Inside the Outage

The bigger problem isn’t that work stops. It’s that the response stops.

Old View New View
The plan is in SharePoint, so everyone can find it The plan is in SharePoint, so no one can open it when SharePoint is down
The crisis team meets in Teams The crisis team needs a channel that doesn’t run through the vendor it’s planning around
Runbooks live in the wiki One vendor outage takes the work, the ticket queue, and the instructions for fixing it
We’ll email employees an update Email is the thing that’s down

We see each of these in real programs. One client keeps its recovery runbooks in the same vendor’s wiki that runs its ticketing and its code. Another runs incident command in Teams, with a fallback chat for the response team and no answer yet for reaching thousands of employees without Microsoft. At another, the business continuity lead learned about a major outage from a coworker, because they had never been on the incident notification list.

Years ago we ran a ransomware tabletop with about forty people from a mature client. The control team sent a fake attacker attachment that read, “If you open this document, please raise your hand.” It got forwarded to the whole crisis team, and three senior executives opened it. We declared their machines, and the Teams chat running the response, compromised. You could see the room deflate. I described it on Managing Uncertainty, Episode 128.

If your crisis team meets inside the outage, you don’t have a crisis team. You have a group chat that’s down.

SaaS Outage Business Continuity: Four Questions for the Next 90 Days

Answer these before the next outage answers them for you.

  1. Where does the work run? Add the collaboration layer, project-driven SaaS, and AI tools to the BIA. For each one, name the owner, who decides to switch to the workaround, and how long each critical process can run without it.
  2. Where does the plan live? Keep the plan, the call tree, and your top runbooks somewhere that isn’t the vendor you’re planning around: offline, printed, or on a different provider. Then check that someone can actually open it.
  3. Where does the team meet? Name an out-of-band communication channel for the crisis team and a way to reach every employee that doesn’t run through your email. Test how many people it actually reaches.
  4. Who answers the questions? For your three most-used platforms, name the owner, the backup, and the one place employees go for updates. Put your business continuity lead on the incident notification list.

Then test it. Run a 72-hour SaaS outage tabletop exercise where the first inject takes away email and chat, and watch what the team reaches for. That’s your real plan. When the gaps surface, turn the after-action findings into a 90-day plan with an owner and a date on each fix. And if your recovery planning still assumes the attacker is the only threat, our post on what ransomware recovery plans miss covers the same blind spot from the other side.

This is the broader point I made on Episode 327: a technology failure is a business crisis, not an IT ticket.

The Work Moved. The Plan Didn’t.

For years we planned for losing the building, then the data center, and we got good at both. The work doesn’t happen there anymore. It happens in email, chat, the shared drive, and tools your people opened this morning without thinking. Somebody else runs every one of them.

Find out where your plan lives before the next outage tells you.

Keep Going

A few ways to go deeper if this was useful.

Bryan Strawser Avatar

About the author

Let’s build your Resilience Roadmap

Our team has spent two decades helping the world’s leading organizations prepare for and navigate disruption. We’ll bring that experience to yours.