FAA Plans $875 Million AI System to Predict Congestion Before It Delays Flights

The Federal Aviation Administration is investing $875 million over 12 years in an AI-based system designed to identify air-traffic bottlenecks before they produce widespread delays. The contract went to Boston-based Air Space Intelligence, which will build SMART, short for Strategic Management of Airspace, Routes, and Trajectories, alongside a broader data and flow-management replacement for the FAA’s command center systems.

The program addresses a persistent shortage of air traffic controllers and pressure on an aging management infrastructure. The initial deployment focuses on Washington, D.C., where several major airports share crowded airspace. The FAA’s broader objective is national coverage across the U.S. airspace system, according to descriptions of the contract and the planned rollout.

SMART is designed to analyze airline schedules, weather conditions, airport capacity and other operational data, then recommend ways to manage traffic flows. Air Space Intelligence describes the underlying platform, Flyways AI, as a forecasting and simulation system that models future airspace conditions and identifies demand-capacity imbalances before they occur.

During the first test, the software will provide route and traffic-management advice through existing FAA systems. Controllers will retain authority over operational decisions, while pilots and airlines will continue receiving instructions through established channels. That limits the initial system’s direct role, but it also gives the FAA a way to evaluate its predictions inside live operations without handing safety decisions to an automated tool.

The FAA is preparing to begin the first operational test around the Washington area as soon as September 21. The region offers a demanding trial environment because airports, military activity, weather disruptions and high passenger volumes compete for limited airspace and runway capacity.

The immediate test is aimed at preventing congestion rather than eliminating delays after they occur. Recommendations could help traffic managers adjust routes, timing or demand before a bottleneck spreads across the network. The passenger impact will therefore be indirect: travelers won’t interact with SMART, but a successful system could improve schedule reliability and reduce cascading disruptions.

The advisory design leaves human controllers responsible for final decisions, but it also creates a question about accountability when an AI recommendation contributes to a bad outcome. Air traffic decisions involve changing weather, incomplete information and rapidly shifting demand, making it difficult to separate a model’s error from a human judgment call or a flawed operational rule.

That concern matters because the system’s underlying models and validation process have not been fully detailed in the public descriptions available for the pilot. The FAA will need to assess prediction accuracy, explainability, failure handling and the effect of recommendations on controller workload. A system that spots congestion early but produces too many false alarms could add noise to already pressured operations.

The contract extends beyond SMART’s Washington trial. It also covers a Flow Management Data and Services system intended to replace the current platform used by the FAA’s Air Traffic Control System Command Center in Virginia. That makes the project an infrastructure overhaul as well as an AI deployment.

A national rollout would cover the FAA’s full airspace network, with the advisory platform connecting command-center operations, regional facilities and airlines. Air Space Intelligence’s product materials describe support across command centers, air route traffic control centers, terminal control facilities and carriers. The scale raises integration, cybersecurity, procurement and training demands that a successful regional test alone won’t settle.

The Washington pilot will provide the first practical evidence of whether SMART can forecast congestion accurately enough to improve decisions. Key measures include delay reduction, prediction precision, controller workload, system uptime and performance during severe weather or unusual traffic patterns.

The FAA’s investment reflects a wider effort to use software to stretch limited airspace and staffing capacity, but the system’s value will depend on operational results rather than its contract size. If SMART produces useful warnings without distracting controllers, it can support a broader modernization effort. If its recommendations are difficult to trust or integrate, the agency will face a costly lesson before expanding nationwide.