The Federal Aviation Administration is preparing to introduce its new AI-assisted traffic-management tool at Washington-area airports, putting a high-profile test of automation into one of the country’s most crowded and sensitive airspaces.¹

SMART—short for Strategic Management of Airspace, Routes, and Trajectories—is designed to forecast traffic flows and flag possible conflicts using airline schedules, weather, airport capacity and airspace conditions. The FAA’s plan is to begin in the Washington region before extending the system across the 29 million square miles of airspace it oversees.²

The rollout follows the agency’s June award of a 12-year, $875 million contract to Boston-based Air Space Intelligence. SMART is intended to give the FAA, carriers and aircraft operators a common operating picture, with the promised payoff being more efficient routes, lower fuel use, stronger on-time performance and faster recovery after disruption.²

Supporters say the narrow debut is not a retreat but a necessary guardrail. Philip Mann, a former FAA official and aviation-safety consultant, called limiting the initial scope the “right call,” arguing that “every cut in scope shrinks the unknowns.”² His concern is not one faulty forecast, but the risks of introducing AI at national scale.

Airlines, meanwhile, were reportedly unsettled by weeks of uncertainty over how the system would alter operations. Their concerns eased only after the FAA said SMART would not change controller or airline procedures; it would supply alternative-route information through existing systems.²

That distinction matters. The FAA is pitching SMART as an enhancement to—not a replacement for—human control, while it also confronts a longstanding controller shortage and aging infrastructure. Another account frames the project as part of a broader modernization and staffing push, describing the software as a way to help controllers manage workloads and navigate routes more safely.³

The next test is harder than launch day: Mann says the FAA should show performance under real workloads and degraded data, then answer in writing who owns the outcome when a prediction is wrong.² In Washington’s airspace, that is the question no algorithm can sidestep.