FAA Tests AI Tool to Predict Congestion and Cut U.S. Flight Delays

The Federal Aviation Administration is moving into a new phase of air traffic modernization with planned tests of an artificial intelligence-assisted system intended to spot congestion earlier and help prevent delays from rippling across the U.S. flight network.

What the FAA is testing and where it fits in air traffic control

Published coverage indicates the FAA’s test effort centers on software designed to forecast where and when the airspace system will run into trouble, using a combination of airline schedules, weather information, airport capacity and operational constraints. The concept is to predict bottlenecks early enough for traffic managers to adjust departure times, reroutes, or flow restrictions before disruptions spread.

The system at the heart of the effort is widely described as SMART, short for Strategic Management of Airspace, Routes, and Trajectories. The FAA has positioned SMART as a decision-support capability, not an automation layer that directly controls aircraft. Early testing is expected to focus on recommendations provided to human traffic managers rather than autonomous execution.

The initiative also connects to longer-running FAA NextGen modernization programs that already use time-based management and automated decision support. Those include the Traffic Flow Management System (TFMS), Time Based Flow Management (TBFM), and Terminal Flight Data Manager (TFDM), which are used to help sequence traffic and manage throughput from gate to gate. The new AI-driven layer is intended to strengthen strategic planning by improving forecasts and identifying conflicts sooner.

Contract details and why 2026 is a key moment for rollout

The FAA has publicly described SMART as part of a broader modernization plan that includes replacing legacy flow management data services used at the Air Traffic Control System Command Center in Virginia. In June 2026, the agency announced the selection of Air Space Intelligence to deploy advanced air traffic control software tied to the SMART program under a long-term contract described in published reporting as an $875 million, 12-year award.

In that FAA announcement, the agency emphasized the goal of reducing delays by improving how airspace is managed before flights depart, easing congestion and lowering controller workload. The software is described as continuously analyzing multiple data streams to predict traffic flows and flag potential conflicts.

Separate reporting in September 2026 described the FAA engaging airline leadership on a broader plan to cut delays, with advanced software and scheduling reforms a key element. That context matters for travelers because it suggests the AI testing is not an isolated pilot but part of an operational push aimed at the day-to-day reliability of the national airspace system.

How AI recommendations could change what passengers experience

For travelers, the biggest potential impact is earlier intervention. When weather, runway constraints, or en route congestion build, delays can cascade: a late departure at one hub can disrupt inbound aircraft rotations, crew duty limits, and connecting itineraries. The FAA’s approach aims to detect those pressures earlier, creating more time for tactical adjustments such as revised departure metering, alternative routing, or managing arrival demand before airports become saturated.

It is also designed to make traffic management more consistent across regions. When multiple facilities must coordinate around the same constraints, better shared forecasts can support smoother flow decisions, potentially reducing last-minute ground stops or prolonged holding patterns that can lead to missed connections and cancellations.

The FAA and NASA have a documented history of testing predictive scheduling and surface-management concepts that target delay and congestion on the ground, not just in the air. For example, FAA and NASA previously completed research and testing of software capabilities aimed at timing gate pushbacks so aircraft can taxi more efficiently at busy hubs, an idea integrated into the TFDM program. While that work is separate from SMART, it illustrates the broader modernization direction: using prediction and data-sharing to manage queues more intelligently.

Limits, safeguards, and what “AI in ATC” does not mean

The FAA’s current framing of SMART is careful: it is positioned as decision support for traffic managers rather than a system that issues direct control instructions to pilots or replaces human separation responsibilities. Published coverage of the new tests emphasizes that initial deployment is expected to provide recommendations, with humans retaining authority over operational decisions.

That distinction matters because “AI” in aviation often raises questions about safety and accountability. In practice, many NextGen capabilities already rely on automation and advanced decision support, and the AI effort can be viewed as an extension of that evolution: using more sophisticated models to improve prediction performance and planning, while maintaining human control over execution.

There are also implementation challenges that do not vanish with better prediction. Congestion can be driven by structural capacity limits at major hubs, weather patterns that compress arrival and departure rates, and staffing constraints that reduce the usable volume of airspace or runway throughput. An AI tool may help optimize within those limits, but it cannot create additional runways or instantly expand sector capacity.

What to watch next: test results, scaling, and the NextGen timeline

The immediate question is whether early testing shows measurable operational value: fewer minutes of delay per flight in affected flows, fewer severe cascades after thunderstorms or low-visibility conditions, and more stable schedules during peak travel periods. Published reporting suggests the Washington-area command and flow-management ecosystem is a focal point for early deployment, reflecting the role of the FAA’s command center in national traffic management.

Another key watch item is integration with existing FAA systems and data pipelines. FAA materials on NextGen describe how time-based management depends on multiple automated decision support systems, and TFDM surface metering emphasizes integration with TBFM, TFMS, and surface surveillance inputs. An AI forecasting layer will likely be judged on whether it can plug into those operational workflows without adding complexity for controllers and traffic managers.

Finally, travelers should expect any benefits to appear unevenly at first. NextGen deployments often expand in stages, and FAA materials describe continuing deployment timelines for certain capabilities through 2029 across dozens of airports. If SMART performs well, the long-term vision described in FAA communications and published reporting suggests broader scaling, but the path from test to nationwide impact will depend on operational validation, integration work, and sustained funding.