The Outages Aren't Connected. The Warning For Midmarket IT Is
Microsoft, the FAA and other recent high-profile outages had different causes. But growing technology concentration, constant change and untested dependencies are forcing IT leaders to rethink what resilience actually requires.
(Image: Getty)
Over the span of a few days, Microsoft 365 went down. The Federal Aviation Administration issued a nationwide ground stop. United Airlines and Chase experienced technology disruptions. Workday had an outage.
The proximity of these events tempts us to draw a connection and ask the obvious question: What the heck is going wrong with technology infrastructure?
To be clear, there is no evidence that these outages share a technical root cause, and spending time manufacturing one would be a futile exercise.
"I don't think there is a common weakness," Pranay Ahlawat, chief technology officer at Commvault, told me in an interview this week. "These are fundamentally sort of different applications and different architectures. They all have different root causes."
As IT environments become more interconnected, dependent on fewer major providers and subject to increasingly frequent change, the recent outages will have more organizations assessing: When something fails, how far can the failure travel?
The FAA Ground Stop: A Lesson In Failure Containment
The FAA's Nov. 13 nationwide ground stop offers an extreme but instructive lesson.
The explanation for the dramatic and sudden halt evolved considerably in the days following. Officials connected the disruption to work being performed as part of the FAA's modernization of communications infrastructure, including transitioning from legacy copper to fiber.
IT leaders responsible for resilience in their organizations, likely watched the events unfold with a hawk's eye. For them, the most telling realization to emerge wasn't about the possibility of some defective router or misconfiguration.
Instead, it was the potential ripple effect, what IT and security teams often call the "blast radius."
Routine technology changes happen every day. Switches get replaced. Fiber gets installed. Software gets patched.
Those changes sometimes fail.
But the recent outages make one thing evident: resilience isn't about preventing every failure. It's about preventing an ordinary failure from becoming an extraordinary event.
And that's not just an FAA problem.
[RELATED: Building Infrastructure That Bends Without Breaking: A Primer On Resilience]
IT Is Accumulating More Moving Parts
Artificial intelligence inevitably enters the discussion, particularly when several high-profile tech failures seemed to occur in succession.
Ahlawat cautions against making that leap.
"I think we have to be careful not to draw a direct line," he said about AI's role.
He does, however, see AI affecting resilience indirectly.
Ahlawat points first to the growth in vulnerabilities increasing the frequency with which systems to be patched or updated.
Then there are AI agents. As organizations deploy agents with access to business applications, data and APIs, some of those agents will have the permissions to take action and make changes across systems.
[RELATED: AI Agents Can Take Action. Who Controls Their Access?]
And AI architecture itself adds new technology components.
“There are more data, more distributed data, new types of applications, application components, new types of services like vector databases and agents themselves,” Ahlawat said. “As business logic basically moves from systems of record to systems of intelligence, it actually creates a little bit more entropy across the systems.”
Put more plainly: AI is adding more moving parts to already complex IT environments.
The result, in his words, is more entropy.
That doesn't mean AI caused any of these outages. It means outages are occurring in increasingly complex and dynamic technology environments. Organizations will have to ensure that resilience practices keep pace with how quickly those architectures are changing.
[WATCH]: MES Computing interview with Pranay Ahlawat, Chief Technology and AI Officer, Commvault:
Vendor Concentration Changes The Equation
Complexity is one issue. Vendor concentration is emerging as another.
Ahlawat pointed to the enormous number of organizations reliant on Microsoft 365 and a relatively small collection of major cloud providers. Recent outages from AWS and CrowdStrike offer similar examples of how failures at major technology providers can have wide-sweeping effects.
Stakes are higher for the midmarket. Large enterprises have the budgets, people and infrastructure to build extensive, redundant environments. A midmarket IT department generally doesn't have the resources to maintain a failover version of every critical service in case one goes offline.
Strong resilience isn't just about redundancy everywhere. It starts with every CIO or CISO, regardless of company size, knowing the organization's crown jewels, Ahlawat said.
Technology has made some resilience options more attainable for smaller organizations, Ahlawat explained. Virtualization, cloud infrastructure and cloud-native technologies can provide recovery and workload portability capabilities that once required far more infrastructure.
The biggest challenge may be knowing where to apply them.
"I don't think they are fully aware of what their crown jewels are," Ahlawat said.
That is a strategy issue rather than simply a technology problem. The exercise isn't just labeling applications “critical.” It's mapping what happens to the business when one disappears. Does a Microsoft 365 outage take down email, for example, or does it also disrupt help desk intake, incident communications, employee support and access to critical documentation?
Recovery Plans That Aren't Tested Aren't Really Recovery Plans
Knowing what's critical is only the first step. Ahlawat identified another weakness he routinely sees: "They don't test recoveries."
Recovery testing should be treated as an operating discipline, not measured by whether a backup succeeded or relegated to an occasional compliance exercise.
Commvault describes that approach as "ResOps": making recovery readiness continuous rather than treating it as a one-time project.
People Are Part Of The Resilience Strategy
During the Microsoft 365 disruption, security experts pointed out a nasty secondary consequence of losing normal communications: attackers don't necessarily go down when corporate email does.
Matt Hull, vice president of cyber intelligence and response at NCC Group, warned that disruption and confusion create opportunities for attackers impersonating Microsoft or internal IT staff through calls, text messages and other channels.
[RELATED: Microsoft's Outage Exposed A Midmarket Resilience Blind Spot]
“Organizations should consider out-of-band communications, which is any method of exchanging information that uses a separate, independent channel or network rather than the standard system,” Hull said.
During an incident, employees need an alternative way to communicate with IT precisely when they also need to be suspicious of unusual requests arriving through alternative channels.
A Practical Resilience Checklist For Midmarket IT Teams
Based on insight from the experts we spoke with, here is a practical resilience checklist that is less of a datacenter diagram and more like an operating model:
- Identify the business's true crown jewels.
- Map the providers, services and systems each depends on.
- Understand where single-provider dependencies exist.
- Decide which workloads require failover and which can accept downtime.
- Define recovery priorities before an incident occurs.
- Keep critical contacts and recovery procedures accessible when main systems aren't.
- Establish an independent channel for outage communications.
- Test restorations rather than assuming successful backups equal successful recovery.
- Exercise what happens if a critical provider disappears for 24 hours as an operational scenario, not simply a technical one.
- Revisit those assumptions whenever the architecture materially changes.
And perhaps one question belongs in every single change review: If this goes catastrophically sideways, how far can the failure travel?