
When the Planes Can Fly but the System Can't
When the Planes Can Fly but the System Can't
On Friday, November 13, the FAA halted departures across the United States after communications problems affected air traffic control centers. The nationwide ground stop lasted more than four hours.
Passengers were waiting. Aircraft were available. Pilots, controllers, and airport employees were at work.
Much of the capability required to move people across the country was still present. But using that capability safely depended on relationships that had become unreliable.
An aircraft's ability to fly is only one part of what makes a flight possible. Its departure has to fit into a changing system of other departures, arrivals, routes, handoffs, and available capacity. Decisions made in one place affect what can safely happen somewhere else.
When the information supporting those decisions becomes unreliable, the consequences extend well beyond the equipment experiencing the problem.
That's what makes this worth examining as a coordination system.
What stopped working?
The immediate question after an incident like this is understandable. What broke?
Early explanations and speculation have included both aging infrastructure and recently introduced AI. The technical cause needs to be established through investigation. We don't need to settle that question to examine the dependency the disruption exposed.
Air travel requires participants to maintain a sufficiently consistent understanding of a situation that never stops changing.
Where are the aircraft? Where are they going? Which routes are available? What instructions have been issued? Who is responsible for the next decision?
Different participants need different information, but their decisions have to remain compatible.
A pilot can be fully capable of operating an aircraft while lacking the information or authorization required to proceed. An airport can have an available runway while the wider network cannot safely accept another departure.
The capability exists within a structure that determines when and how it can be used.
We often treat that structure as background infrastructure. We notice the aircraft, the people, and the visible activity. The relationships making their work possible become much easier to see when they stop supporting it.
The ground stop was part of the system working
There's an important distinction between the communications disruption and the decision to halt departures.
The disruption impaired the operation. The ground stop limited the additional work entering it.
That response imposed enormous costs on passengers, airlines, airports, and everyone depending on their schedules. It also exercised a boundary around what the system was permitted to do under uncertainty.
Continuing to launch aircraft would have added commitments to a network already experiencing difficulty coordinating them.
A departure creates obligations elsewhere. Other participants must make room for it, manage its progress, and receive it. Those obligations continue after the initial decision has been made.
This is a useful way to think about work more broadly.
Starting something commits more than the person or team starting it. It creates demands on the participants who must review it, approve it, integrate it, support it, or respond when it fails.
Organizations make these commitments constantly. Their ability to recognize when they should stop making them varies considerably.
Most organizations keep accepting work
Consider a software organization whose developers can produce changes faster than the organization can review and release them.
Product decisions remain unresolved. Review queues grow. Teams discover dependencies after work has started. People disagree about which changes are ready and which commitments take priority.
The developers continue producing.
From the perspective of individual activity, work is progressing. From the perspective of delivering useful software, the organization is accumulating unfinished obligations.
A similar pattern appears when sales commitments outrun delivery capacity, when hiring moves faster than onboarding, or when leadership introduces more initiatives than teams can reconcile.
Each participant may have a reasonable explanation for continuing. Sales has a target. Engineering has a backlog. Leadership sees an opportunity. Each decision creates work that other participants must absorb.
The question of whether the whole system can support another commitment often has no clear owner.
Even when someone recognizes the problem, they may lack the authority to slow intake. They can describe the growing burden without being able to change the conditions producing it.
The resulting delays are then attributed to the people carrying that burden.
Load changes what a system can tolerate
At lower volumes, people can compensate for incomplete information and weak handoffs.
Someone remembers a conversation. A manager knows whom to call. An experienced employee catches a conflict before it becomes expensive. A team absorbs an unexpected request by moving something else.
These adjustments are part of how organizations function. They also have limits.
As more work enters the system, the number of decisions, dependencies, and exceptions grows. The people holding the operation together have less time to reconcile each one. Information becomes outdated before it reaches everyone who needs it.
Eventually, continuing at the same pace creates more uncertainty than participants can resolve.
AI makes this relationship particularly important.
An agent can generate another proposal, initiate another workflow, or complete another task much faster than the surrounding organization can determine what should happen next. Its output may be useful. Receiving that output still requires capacity, information, and decisions elsewhere.
Before introducing that acceleration, we need to understand where its consequences will land.
Who can authorize the resulting work? Who can see the demands accumulating across teams? What happens when two participants act on incompatible information? What can be paused, and who has permission to pause it?
Those questions belong to the design of the system receiving the capability.
Repairing the component doesn't restore the operation
There's another consequence that becomes visible in an interruption like this.
Once communications are restored, the operation has to resume from the situation the disruption created.
In air travel, that means working through the relationships among aircraft, crews, gates, passengers, and schedules. A resource available before the interruption may no longer be available when service resumes. One delayed movement changes the possibilities for later movements.
The same thing happens inside organizations.
An approval finally arrives, but the team has moved to another priority. A dependency becomes available after the release window has passed. A service returns while everyone is still reconciling work performed during the outage.
Recovery requires participants to establish what is true now and make their next decisions from that state.
This is why the duration of a technical failure tells us only part of its operational cost. The interruption changes the work around it, and those changes must also be resolved.
What makes capability usable?
The FAA incident will require a technical explanation and a careful examination of how the disruption became so widespread.
It also gives us a visible example of something that happens much less visibly in ordinary organizations.
People, teams, software, and AI agents depend on a structure that makes their capabilities usable together. That structure determines what they can know, which decisions they can make, what obligations they create, and how they respond when conditions change.
We spend considerable effort increasing what individual participants can do.
We give much less attention to whether the systems receiving that capability can support the resulting work.
A nationwide ground stop makes the dependency impossible to overlook. In most organizations, the work keeps entering, the unresolved obligations keep growing, and people are asked to make up the difference.
Understanding when a system can accept more work is part of understanding what it can accomplish.