The FAA is preparing to launch SMART, a cloud platform that uses AI to forecast traffic flows and flag potential conflicts before they form. The Wall Street Journal reports it was built by Air Space Intelligence and will cost the government $875 million over 12 years, deployed first across the Washington metropolitan region and extended to other regions after that. The acronym stands for Strategic Management of Airspace, Routes, and Trajectories. It arrives after years of a nationwide controller shortage the agency has not closed — the problem the government keeps reaching for technology to relieve.
What the system actually does is narrower than the name suggests. SMART is described as an extension of the FAA's existing traffic management systems rather than a replacement for them. It ingests airline schedules, weather, airport capacity, the state of the airspace and operational constraints, and from those five inputs predicts how traffic will move and where it will collide with itself. Prediction, not separation. The people who tell an aircraft to turn are still the people who tell an aircraft to turn.
That distinction is the whole story, and the framing around SMART quietly elides it. The FAA's shortage is a shortage of certified controllers — humans licensed to hold legal responsibility for keeping aircraft apart. Software that anticipates congestion three hours out reduces how much of a controller's attention gets burned on foreseeable mess. It does not produce a controller. Read charitably, SMART is a workload instrument sold into a staffing vacancy; read less charitably, it is what an agency buys when the thing it actually needs takes years to train.
The money is smaller than the headline number implies. Spread across 12 years, $875 million is roughly $73 million a year — real money for a single platform, but modest beside the broader overhaul of the country's aging air traffic control systems that the government is running in parallel, and modest for a system meant to reach every region eventually. A 12-year horizon is also a long commitment to one vendor's architecture in a field where the underlying models are replaced every 18 months.
Set it next to what the FAA said earlier this year, when it announced what it called a "bold new" hiring plan and claimed it would resolve the multi-year staffing gap. Two answers to the same problem, announced months apart, neither one contingent on the other. If the hiring plan works, SMART becomes an efficiency layer on a fully staffed system. If it doesn't, SMART becomes the thing standing in for the staff. Nothing in the public description says which assumption the $875 million was priced against.
The more interesting question is one the announcement does not touch: what authority SMART's output carries. A system that predicts conflicts before they arise is either advisory, in which case a controller may ignore it and the FAA has bought a very expensive dashboard, or it is weighted into decisions, in which case someone has to define what happens when the model and the controller disagree, and how a cloud platform that reshapes traffic flows gets certified to the standards aviation applies to everything else in the chain. The project is summarized publicly in a one-page description. For $875 million of public money over 12 years, one page is not a specification; it is a posture.
Washington was chosen as the first deployment, and the schedule for everywhere else is unstated. That sequencing matters more than it looks: whatever SMART learns about flow prediction in its first region becomes the template the rest of the country inherits, and the contract runs long enough that the FAA will still be paying for it more than a decade from now. The agency has bought a decade of one company's judgement about how American airspace should be sequenced, before it has told anyone what that judgement is allowed to override.