The Federal Aviation Administration switches on an artificial intelligence system over Reagan National, Dulles and Baltimore/Washington sometime today, and the single most useful thing anyone can tell you about it is what it does not do.
It is not a collision-avoidance system. It does not watch radar returns and warn a controller that two aircraft are converging. It will not touch any of the failures that killed 67 people in this exact airspace twenty months ago. Every account of this rollout has led with the money and the acronym. Almost none has led with the boundary, and the boundary is the thing a reader who flies out of DCA actually needs.
What SMART Is
The system is called SMART, for Strategic Management of Airspace, Routes and Trajectories. It is cloud software that ingests airline schedules, weather forecasts, runway capacity and operational constraints, then predicts where congestion will build and proposes alternatives: a different route, a different altitude, a different departure time. Transportation Secretary Sean Duffy has described it flagging problems as much as 45 days out. Controllers and airline dispatchers can take the suggestion or ignore it.
The contract behind it went to Air Space Intelligence, a Washington company that beat Palantir and Thales in a field of roughly a dozen bidders, as DroneXL reported when the award landed in June. It runs twelve years and $875 million. We covered the terms of that deal and the risk of building a national system on a product proven on a single airline’s dispatch desk when it was signed. Today is the part where it meets real traffic.
The Two Pitches Cannot Both Be the Headline
Read the public framing and the private reassurance side by side and something does not line up.
Publicly, this is transformation. Futurism characterized it as nearly a billion dollars to make AI the brain of air traffic control, and Duffy has gone back to Congress asking for another $6 billion to $10 billion to push AI further into the system, on top of $12.5 billion already committed to infrastructure. Quartz framed the launch around cutting the delays and cancellations that have made domestic flying miserable.
Privately, it is inert. Airline officials told Politico their misgivings eased only after the FAA assured them that SMART changes no procedure for controllers or for carriers, and that its output arrives as alternative route information inside systems that already exist. That is a promise that nothing will be different.
Both statements are being made about the same product in the same week. A tool cannot be the brain of American airspace and also a suggestion box that alters no workflow. The honest version is that SMART is somewhere in between and the FAA has not said where, because the useful answer and the reassuring answer point in opposite directions. Whether it earns its price depends entirely on how often controllers and dispatchers act on what it says, and there is no published standard for when they should, no account of how an advisory routing model gets certified before it runs over a major metropolitan area, and no metric the public can check in six months.
Why the Boundary Matters Here of All Places
The FAA chose the Washington terminal area to go first. That is defensible on the merits. It is congested, it has three major airports operating in overlapping airspace, and it is where the agency’s own command center sits.
It is also where American Eagle Flight 5342 and an Army Black Hawk collided on approach on January 29, 2025, killing all 67 people aboard both aircraft. After a year of work, the National Transportation Safety Board blamed deep systemic failures: the helicopter was flying too high and sat in the jet’s landing path, and in the tower that night one controller was working the helicopter position and the fixed-wing position at the same time, two jobs that are supposed to be two people.
Nothing in SMART addresses that. Strategic flow management operates on schedules and routes over hours and days. It does not fix a tower running one controller short, because it is not a controller. The FAA has been clear that the system is not intended to solve the staffing shortage, and the agency is roughly 1,500 controllers below its own target of 12,563. That gap is the reason flights get cut at major airports, and it is a hiring and training problem that software does not touch.
None of this makes SMART a bad idea. Better congestion forecasting is worth having, and if it trims delays it will have done something real. Our position is narrower and it is about honesty: an agency that is 1,500 people short should not let a billion-dollar software launch stand in for the staffing plan, and it should not let the public infer safety improvements from a tool that provides none. The FAA knows the difference. Today’s coverage mostly did not draw it, and the flying public will draw the wrong conclusion by default. If SMART works, the number to watch is delay minutes in the Washington area this winter. If the agency starts citing it when someone asks about controller headcount, that is the tell.