200 Data Streams and an $875 Million Bet: The FAA's SMART AI Goes Live Over Washington

The most consequential AI deployment in the United States this month is not a chatbot, an image generator, or a coding agent. It is a piece of air traffic control software running quietly above the Washington, D.C. region. On September 21, the Federal Aviation Administration switched on SMART — the Strategic Management of Airspace, Routes and Trajectories platform — at Reagan National, Dulles International, and BWI Marshall, the opening move of what the agency intends to become a nationwide rethink of how American airspace is managed. National television coverage followed within the week, and so, almost immediately, did the first political backlash.

What SMART actually does

SMART is a predictive system, not an autonomous one. According to the FAA, the platform centralizes roughly 200 separate data streams — airline schedules, real-time runway utilization, gate holds, active flight paths, high-altitude weather patterns, and live airspace constraints — and uses machine learning to forecast where the National Airspace System is about to clog. Where today’s tools largely react to congestion after it forms, SMART is designed to spot bottlenecks, routing conflicts, and schedule clashes before aircraft ever push back from the gate.

The output is a set of recommendations. The system can suggest that controllers direct flights to tweak their paths or speeds so that an entire arrival bank lands without severe delays, and it can propose schedule adjustments hours ahead of the problem. Crucially, the FAA has been careful to state the boundary: SMART cannot replace air traffic controllers and cannot take control of an aircraft. Every recommendation is reviewed by FAA personnel before anything happens.

Technically, SMART operates as an enhancement layered on top of the Flight Data Management System (FMDS), using FMDS data to coordinate schedules and trajectories pre-departure. The agency frames the two systems together as finally giving controllers a single pane of glass for critical data, with proactive delay prediction as the payoff.

The $875 million vendor behind it

The platform was built by Air Space Intelligence (ASI), a Boston-based company that beat out both Palantir and Thales for the $875 million FAA contract awarded in June 2026. ASI is not a newcomer to the problem: its Flyways AI platform has already been used in commercial airline operations to optimize flight paths, and the FAA’s selection of a specialist predictive-analytics firm over defense-tech giants was read at the time as a statement about what the agency wanted — trajectory forecasting, not enterprise data plumbing.

Transportation Secretary Sean Duffy has championed the program publicly, and FAA Administrator Bryan Bedford personally showcased the system at the Department of Transportation, describing how real-time data would better inform how controllers oversee the airspace. The rollout at the three D.C.-area airports is explicitly a pilot for a national expansion; the FAA has said it plans to extend the technology to additional facilities across the country if the Washington deployment performs.

“Guinea pigs”: the backlash begins

The deployment did not wait long for a critic with a title. Representative Don Beyer, whose Northern Virginia district sits under some of the busiest airspace in the country, has formally called on the FAA to suspend the rollout until the system can be vetted more thoroughly by human operators on the ground.

“The FAA should not use my constituents as guinea pigs for an unproven AI air traffic control system,” Beyer said, citing the roughly 24 million people who live in the affected region. His letter to the agency raises a specific and pointed concern: air traffic controllers have reportedly been told that SMART can adjust schedules and, in some circumstances, potentially even routes. The gap between “advisory tool” and “system that reshapes the actual traffic plan” is exactly where Beyer is aiming his fire.

The controllers’ union has added a quieter but arguably more damaging observation. NATCA, which represents nearly 20,000 controllers and engineers, has stated that it was not involved in the design, testing, or implementation of SMART, and therefore cannot speak to its effect on air traffic operations. That is not an accusation of danger — but for a safety-critical system, excluding the frontline operators from the build process is the kind of process failure that independent reviews tend to remember.

The context is impossible to ignore. American air traffic control has been under intense scrutiny since a series of high-profile incidents and near-misses, chronic controller understaffing, and the aging of infrastructure that in some facilities still runs on paper strips and 1980s-era hardware. The FAA’s modernization push — SMART being its most visible symbol — is a bet that predictive software can squeeze more capacity out of constrained airspace without new runways or new controller hiring at scale.

The stakes, and what to watch

Strip away the politics and the underlying wager is straightforward: the FAA believes that forecasting congestion 30 to 60 minutes earlier, and acting on that forecast before departure, can meaningfully cut delays and cancellations in the most constrained airspace in the country. If the D.C. pilot shows measurable delay reduction without operational incidents, expect the national expansion to accelerate quickly — and expect every other air navigation service provider in the world to study the results.

If, on the other hand, the system produces recommendations that controllers learn to distrust — or worse, if a schedule adjustment ever gets linked to an operational event — the suspension calls will grow from one Northern Virginia congressman to something much broader. The FAA’s insistence that humans review every recommendation is the guardrail the entire program rests on.

Either way, the era of AI quietly managing meta-level flow decisions in U.S. airspace has begun. It started with 200 data streams, three airports, and $875 million. Whether it ends with a modernized National Airspace System or a suspended program under congressional investigation will be decided in the crowded skies over the Potomac.