FAA Takes AI Traffic Advice Nationwide: Inside the $875 Million SMART Rollout

Posted on 13th Nov 2026 11:03:04 in Artificial Intelligence, Machine Learning

Tagged as: FAA, SMART, Air Traffic Control, AI Governance, Aviation

The Federal Aviation Administration has extended its AI-assisted air traffic planning system from a three-airport trial in the Washington region to the airspace managed by all 22 of the country's air route traffic control centers. The agency announced the nationwide expansion on November 10, 2026, barely seven weeks after the tool went live over Reagan National, Dulles and Baltimore/Washington International.

The system is called SMART — Strategic Management of Airspace, Routes and Trajectories. It is a cloud-based platform that pulls roughly 200 data streams, covering weather patterns, flight paths, traffic flow and controller staffing metrics, into a single operational picture. An AI-supported engine then forecasts where demand will outrun available capacity and recommends schedules, routes and arrival sequencing hours, days or even weeks before congestion forms.

The results the agency is promoting are modest in absolute terms but pointed: 180 flights rerouted since September 21, with average savings of about 10 minutes and nearly 80 gallons of fuel per flight. Whether those Washington-area numbers survive contact with 22 centers, a winter weather season and a controller workforce that was never consulted on the design is the open question the expansion now has to answer.

From Three Airports to the Entire National Airspace System

SMART was unveiled on September 21 by Transportation Secretary Sean Duffy and FAA Administrator Bryan Bedford as a limited deployment covering the National Capital Region, the tightly packed airspace around Reagan National, Washington Dulles and Baltimore/Washington International. Before the nationwide announcement, coverage had already widened to the Atlanta center and to Jacksonville and Miami in Florida, according to reporting by The Dallas Morning News.

The November 10 statement framed the expansion as a validation exercise that passed: "After a successful initial rollout to the National Capital Region in September, the FAA will now use SMART in the airspace of all 22 centers across the country to reduce delays, reduce fuel consumption and improve the travel experience for the flying public."

Air Space Intelligence, a Boston-based software firm, is building the system under a 12-year, $875 million contract awarded in June. The FAA's own framing is that SMART is a planning layer, not a control system. It sits alongside the existing air traffic operation rather than inside it, feeding recommendations upward into decisions that humans still own.

Scale matters here. The FAA's Air Traffic Organization handles more than 44,000 flights and roughly three million passengers every day across 29.4 million square miles of airspace, about 17 percent of the world's total. A tool that saves ten minutes on a handful of flights in the mid-Atlantic is a promising pilot. The same tool advising planners in every center is a different category of dependency.

How SMART Actually Works, and Where the Humans Sit

The operating chain is more layered than the headlines suggest, and the layering is deliberate. FAA specialists at the SMART center in Washington review the model's output first. Validated recommendations then go to the Air Traffic Control System Command Center in Warrenton, Virginia, which passes them to the facilities responsible for the affected airspace. Local managers at those facilities decide whether to act on each suggestion or set it aside.

That gives the system two human checkpoints before any recommendation reaches a controller's working picture. The FAA has been explicit about the limits: in its September announcement, the agency stated that SMART "cannot be used to replace the critical role air traffic controllers play or take over control of an aircraft." No aircraft is commanded by the model. Nothing in the architecture closes a loop between prediction and clearance.

The inputs are what make the predictions interesting, and what make them hard to audit from outside. Weather is a known quantity. Flight schedules are published. Controller staffing metrics are not — and staffing is precisely where the model's judgments get politically sensitive, because a system that predicts capacity from staffing levels is implicitly making statements about how much traffic the workforce can absorb on a given shift.

The 180 Flights, the 10 Minutes, and the Missing Passenger Data

The FAA's evidence base is thin enough that its own examples carry the argument. The agency told The Dallas Morning News that SMART recommended a shorter, faster route for a cargo flight inbound to Dallas-Fort Worth International. The newspaper noted that the agency did not identify the flight, and has not explained how the system has assisted passenger traffic — the traffic that actually generates the delay statistics the public feels.

That gap is not trivial. Cargo operators fly at different times, with different flexibility and different tolerance for rerouting than scheduled passenger carriers. A tool that produces clean wins on freight movement may still be unproven on the peak-hour passenger banks that cause the delays SMART is meant to dissolve. The 180-flight figure covers a seven-week window in one region, and it is being used to justify a national rollout.

Airline leadership has been enthusiastic regardless. United CEO Scott Kirby said that using SMART to open airspace intelligently "has the potential to significantly reduce delays and cancellations during weather events," adding that if it works as expected it would do more than anything in decades. Delta's Ed Bastian described the airline as actively engaged in development. American's Robert Isom and Southwest's Bob Jordan offered similar support, as did Airlines for America chief Chris Sununu.

The Union That Was Not in the Room

The National Air Traffic Controllers Association, which represents more than 15,000 members, told Flying magazine that it was not involved in the design, testing or implementation of SMART and therefore cannot speak to its effect on the air traffic control workforce. That is a striking omission for a system whose entire purpose is to shape the workload of the people the union represents.

The FAA and NATCA have already clashed this year over staffing. The agency lowered its targeted controller headcount without the union's input, and facilities remain understaffed after years of attrition and difficult hiring. The FAA says it has recruited 4,000 controllers into the pipeline in 18 months, reached roughly 11,000 controllers in towers and cut more than five months from time-to-hire. Those are real gains against a deep hole.

Research cited by critics of the rollout points at a specific hazard: people tend to lean harder on automated recommendations precisely when they are overworked. In a facility running short-staffed, a suggestion that arrives pre-validated by a national center is easy to accept and hard to second-guess in the moment. The FAA's answer — that controllers may decline any recommendation — assumes the cognitive slack to decline exists on the busy shift, which is the shift where the tool matters most.

Twenty-Five Lawmakers, a Ninety-Day Window, and a Letter Still Unanswered

On October 9, a group of 25 Democratic lawmakers led by Representative April McClain Delaney of Maryland wrote to Bedford and Duffy asking the FAA to pause the Washington rollout. The letter, co-led by Alexandria Ocasio-Cortez, Don Beyer, Kweisi Mfume, Suhas Subramanyam and Sarah Elfreth, described an "egregious lack of transparency" about how SMART's AI has been tested, how it is being integrated and what its automated recommendations could do.

The lawmakers asked why the National Capital Region was chosen at all, citing the fatal January 2025 midair collision between a U.S. Army helicopter and a passenger jet over the Potomac. They said they had not seen safety evaluation or testing reports and were not aware of training efforts for aviation safety personnel. They set an October 19 deadline for answers. The FAA's November 10 expansion announcement did not address the questions.

Several specific accountability items remain open: which officials approved the rollout, whether any safety measures were compressed to hit a September launch, how the system performs in off-nominal conditions such as severe weather or equipment outages, and who bears responsibility — controller, FAA or vendor — if a SMART recommendation contributes to a loss of separation. How accepted, rejected and overridden recommendations get logged and audited is also unanswered.

The timing compounds the concern. Reporting by Politico and The New York Times described the DC deployment as a 90-day evaluation, with the system initially handling only same-day delays rather than the multi-week lookahead it is being built for. Duffy said in August that full deployment was targeted for spring 2027, with different capabilities slow-rolled rather than switched on together. Expanding to 22 centers inside that evaluation window inverts the sequence the agency itself described.

What This Means for Aviation Regulation Everywhere, Including India

SMART is the clearest test yet of a governance model that regulators worldwide are circling: keep the model advisory, keep a human in the loop, and call that sufficient. The FAA's design does exactly that, and the criticism it is drawing is not that the model is dangerous in isolation but that advisories carry weight whether or not anyone signed for them. Where a controller is stretched, an authoritative suggestion is a soft form of control.

That lesson travels. India's airspace is growing faster than its controller corps, and Indian carriers are adding long-haul capacity to the United States, which means Indian passengers sit inside the delay and fuel economics SMART is meant to improve. Any aviation authority weighing a similar planning layer will face the same three questions the FAA is now being asked: what evidence justifies the expansion, who was consulted before it happened, and who is accountable when the recommendation is wrong. Getting the procurement and the safety case right at the pilot stage is cheaper than retrofitting accountability after a national rollout.

For enterprises in any safety-critical sector, the FAA's rollout offers a template and a warning. The template is a two-tier human review chain plus an explicit never-touches-the-controls boundary, which is more disciplined than most corporate agent deployments. The warning is that a successful pilot on a narrow slice of operations is weak evidence for a system-wide expansion, especially when the measured benefits are minutes and gallons rather than safety outcomes, and when the workforce that must absorb the tool was never part of designing it.

The next real signal will not be a press release. It will be whether the FAA publishes the evaluation data behind the 180 reroutes, whether the 10-minute average holds across centers with different traffic mixes, and whether the accountability question gets a written answer before a controller has to make that judgment live.

Sources

• Ars Technica — FAA expands AI advice to manage air traffic congestion across entire US airspace
• AeroTime — FAA expands SMART air traffic tool to all 22 US en route centers
• Flying Magazine — FAA Must Pause Use of AI-Based SMART ATC System, Lawmakers Say
• AVweb — Lawmakers Seek Pause on FAA AI Tool
• US Department of Transportation — SMART launch announcement, September 21, 2026
• Superpower Daily — FAA takes AI traffic advice nationwide, with controllers retaining the final say