FAA Started a 90-Day AI Pilot to Predict Airspace Congestion
The FAA has begun a 90-day Washington-area pilot of SMART, predictive software intended to flag airspace congestion before delays spread. It advises human traffic managers and does not control aircraft; no live delay-reduction results have yet been published.
The US Federal Aviation Administration has started using a predictive software system in the Washington, DC region to help traffic managers anticipate congestion and disruption before flights are delayed. The 90-day pilot covers airspace serving Reagan National, Dulles and Baltimore/Washington airports.
The 30-second summary
- What happened? The FAA activated SMART, software that combines schedules, weather, airport capacity and airspace conditions to predict traffic flow.
- Why does it matter? Earlier warning could help managers adjust routes and departure plans before congestion spreads across the network.
- What is the catch? SMART is an advisory tool, not an autonomous air traffic controller, and the pilot has not yet produced public evidence that it cuts delays.
Key Number: The underlying Air Space Intelligence contract is worth up to $875 million over 12 years, but that ceiling covers a broader software programme rather than proven savings from the initial pilot.
What SMART is watching
SMART stands for Strategic Management of Airspace, Routes and Trajectories. Reports about the launch say it draws from roughly 200 data streams, including airline schedules, weather forecasts, runway capacity and current airspace conditions. The software projects how traffic may develop and highlights potential bottlenecks for human managers.
That role is different from separating individual aircraft. The FAA’s command centre balances demand against available capacity, coordinating measures such as reroutes, ground-delay programmes and temporary ground stops. SMART is intended to give that process a more dynamic view of what may happen next.
From contract to operational pilot
The FAA awarded Air Space Intelligence the 12-year contract in June. At that stage, the system was an announced modernization project. Beginning operational use in the Washington area is the material new step. The agency says it plans to expand the software gradually, while officials have pointed to summer 2027 as a period when travellers may begin noticing broader improvements.
The Washington region is a demanding test environment because it combines three major commercial airports, restricted airspace, variable weather and heavy traffic. The pilot can reveal whether predictions arrive early enough to change decisions and whether suggested actions reduce disruption without moving delays elsewhere.
Before we overstate the result
No controlled performance results have been published from the live pilot. Claims about fewer cancellations, lower fuel use or reduced emissions therefore remain expected benefits, not measured outcomes. Humans retain decision authority, and SMART does not replace radar, communications equipment or the controllers responsible for aircraft separation. It also cannot prevent delays caused by failures in those underlying systems.
What evidence to watch next
A meaningful evaluation would report forecast accuracy, the number of recommendations accepted, delay minutes avoided, effects on neighbouring airports and any additional workload for traffic managers. The FAA will also need to explain how it monitors model errors and performance during unusual weather or equipment failures.
For now, the pilot matters because a national aviation authority has moved predictive analytics into day-to-day airspace management. Whether it becomes a genuine modernization milestone depends on measurable operational results, not the size of the contract or the AI label.
Verified topics and entities
Sources and citations5 sources
External references used to support the reporting in this article.
Published by
NewTqnia Artificial Intelligence Desk
An institutional editorial team within NewTqnia
