By Kiran Bhujle, Global Head of Cyber, SVAM International Inc.; part-time lecturer in Enterprise Risk Management and Technology Management, Columbia University School of Professional Studies
It is 1:15 p.m. on Thursday, Nov. 12, 2026, and the security operations center of a midsize hospital system has gone quiet in the wrong way. The alert feed in Defender, which normally scrolls like a stock ticker, has stopped. The analyst on shift assumes the console has frozen and reloads it. Nothing. She messages her lead on Teams. The message sits with a spinning circle.
I'm describing a composite, not a specific client, but some version of that scene was playing out in thousands of organizations at once. Email was down. Teams, the fallback, was down with it. The security console was dark. And when people turned to Microsoft's status page to find out what was happening, it wouldn't load.
The CISO in that building has a 5 p.m. call with her board chair. She has three and a half hours to answer a question she has never been asked quite this way: What, exactly, do we do when the vendor is the outage?
The rehearsal
After a faulty CrowdStrike update in July 2024 took down about 8.5 million Windows machines (Microsoft's count) and disrupted critical services around the world, Jen Easterly, then director of CISA, called the incident “a useful exercise, like a dress rehearsal” (The Record).
Thursday was the next rehearsal.
At 12:54 p.m. Eastern time, Microsoft's 365 Status account posted that it was investigating problems reaching Exchange Online mailboxes “in the Americas.” Within minutes, that became “users in any region.” Shortly after 1 p.m., Downdetector showed more than 130,000 Outlook reports, with tens of thousands more for Teams and Microsoft 365 (Al Jazeera, citing Reuters). Microsoft's service-health page listed Exchange Online, Teams, SharePoint Online, Copilot and Defender XDR as affected; users also reported OneDrive and the admin center failing, and for a stretch the status page itself would not load (TechRadar). A separate Teams problem in North America came and went that morning (Newsweek).
Nobody attacked Microsoft, as far as anyone has said. Microsoft’s early updates described an internal cascade: servers failed automated health checks and were pulled from service, shifting their load to the remaining infrastructure.
Recovery stretched into the following day. Microsoft later said it had addressed the underlying technical issue while services gradually returned.
The larger risk-management point does not depend on the precise technical cause. When a platform carries communication, collaboration, identity and security functions for the same organization, an operational failure can become an enterprise-wide event.
Same bug, different outcome
Go back to July 2024 for a moment, because the most instructive story from that rehearsal is not about who went down. It’s about who didn't.
When the CrowdStrike update hit, New York City’s 911 systems kept running. Not because the city got lucky, but because critical systems were protected differently. The city isolates and tests updates in a sandbox before they reach anything that matters. “In this case, there's no luck,” city CTO Matthew Fraser said at the briefing afterward “It's good planning and practice.”
That is the argument in two sentences. Or, as Daniel Bizo of the Uptime Institute put it after the CrowdStrike incident: “You can outsource processes, you can outsource jobs, but you can't outsource responsibility.”
All of that was knowable in the summer of 2024. Which brings us back to the CISO and her 5 p.m. call.
The 5 p.m. call
The board chair will not ask about health checks or Exchange Online. He will ask the kinds of questions that map directly to the COSO ERM framework we teach in Columbia’s Enterprise Risk Management program.
Who decided we'd run communication, identity and security through one vendor? That's governance. Was the concentration a conscious trade-off between efficiency and resilience? That's strategy and objective-setting. Was the dependency on the risk register with an impact measured in hours and dollars? That’s performance. After the CrowdStrike incident, did anyone reassess it? That’s review and revision. And if the vendor’s own status page fails, how do we learn what is happening and communicate with employees? That's information and communication.
The honest answer to most of those, in most organizations, is uncomfortable. Third-party risk programs have been built around the vendor as an attack path. Questionnaires ask about encryption and SOC 2 reports. They may spend less time asking what happens when the vendor simply stops. Thursday was an operational failure at a hyperscaler, spread through concentration, and almost no one's assessment had a line for it.
Regulators increasingly recognize the same concentration problem. The EU now directly oversees designated critical ICT providers under the Digital Operational Resilience Act (EBA), while U.K. regulators began overseeing Microsoft and three other major technology providers as critical third parties in 2026 (FCA). U.S. banking regulators likewise emphasize that using a third party does not diminish a firm's responsibility for managing the resulting risk (Federal Reserve, FDIC and OCC interagency guidance).
What the rehearsal taught
Strip away the vendor name, and the lesson is straightforward: Critical dependencies need to be managed as risks in their own right. Concentration remains a risk even when every individual component performs well. When one platform carries email, collaboration, identity and security telemetry, a single failure can remove both the ability to work and the ability to see what is happening.
The response is not necessarily to abandon an integrated platform. It is to identify where shared fate creates an unacceptable blast radius. Put loss of the primary collaboration or identity platform on the risk register. Map the processes that stop with it. Keep incident plans and critical contacts somewhere the vendor does not control. Maintain an out-of-band communication channel. Where necessary, separate critical controls from the platform they monitor. And test the scenario.
Then revisit the assessment after the next incident. A risk-management process that never changes its estimate when new evidence arrives is not really a review process.
Opening night
The hardest part of enterprise risk management isn’t identifying risks. It’s keeping attention on the ones that haven’t hurt you yet, or haven't hurt you recently. Thursday’s outage will produce a flurry of post-incident reviews, but the value of a rehearsal lies in what changes afterward. The organizations that come out ahead will be the ones that took Easterly literally: a rehearsal is only useful if you take notes and change the script before opening night.
The vendor will change. The questions will not: What do we depend on? Who decided that? What happens when it fails? How will we communicate if the vendor's own channels are unavailable? And have we actually tested the answer?
Those are not questions to ask after the next outage. They are the work of resilience before opening night.
Views and opinions expressed here are those of the authors, and do not necessarily reflect the official position of Columbia University School of Professional Studies or Columbia University.
About the Program
The Master of Science in Enterprise Risk Management (ERM) program at Columbia University prepares graduates to inform better risk-reward decisions by providing a complete, robust, and integrated picture of both upside and downside volatility across an entire enterprise. For both the full-time and part-time options, students may take all their courses on Columbia’s New York City campus or choose the synchronous online class experience.
The application deadline for the M.S. in Enterprise Risk Management program is May 1. Learn more about the program here.