FAA Prepares to Launch $875M AI System for Air Traffic Management
The skies above Washington, DC are among the most demanding in the world to manage. Three major commercial airports — Reagan National, Dulles International, and Baltimore/Washington Thurgood Marshall — feed into a compressed corridor of airspace that controllers must navigate with precision around the clock. Against that backdrop, the Federal Aviation Administration is preparing to deploy a significant new tool: an AI-powered system with an expected price tag of $875 million, designed to help manage the complexity of traffic moving through that congested region.
Known as SMART, the system is expected to begin advising air traffic controllers in the Washington, DC area as soon as late September 2026, according to US government and industry officials who spoke with The Wall Street Journal. This initial deployment is designed as the first phase of a planned nationwide rollout — one that would eventually extend FAA AI air traffic control capabilities across all 29 million square miles of US national airspace that the agency oversees.
That figure deserves a moment's reflection. Twenty-nine million square miles encompasses commercial flight corridors, military operations zones, general aviation paths, and increasingly dense drone traffic. Managing all of it manually, in real time, with growing demand from airlines and private operators, has pushed existing systems to their limits for years. SMART is the FAA's most ambitious answer yet to a problem that has compounded steadily over decades.
How AI Will Predict and Prevent Air Traffic Conflicts
SMART functions as a decision-support layer sitting between raw data and the human controllers who make final calls. The FAA describes the system as using AI models to analyze and predict air traffic flows while identifying potential conflicts before they develop into operational crises. To do that, SMART ingests a range of real-world inputs: airline schedules, current and forecast weather, airport capacity constraints, and live airspace conditions.
Read next Laika's Wildwood: Stop-Motion Fantasy at TIFF 2026The emphasis on prediction rather than reaction is what sets this approach apart from legacy tools. Traditional air traffic management systems are largely reactive, responding to conflicts that are already forming. SMART's design is built around anticipating those conflicts earlier — flagging a scheduling bottleneck before planes are circling, or identifying a weather-driven choke point before it cascades across multiple airports.
Crucially, the system advises controllers rather than replacing their judgment. This distinction matters both practically and politically. Air traffic control remains one of the most consequential and legally sensitive occupations in the federal workforce. Any tool that automated final decisions rather than informing them would face significant regulatory and labor obstacles. By positioning SMART as advisory, the FAA keeps human controllers in the decision loop — a design choice that reflects both the current state of AI reliability and the realities of deploying such technology in a safety-critical environment.
FAA AI air traffic control tools like SMART belong to a growing class of systems that augment expert judgment rather than substitute for it. The same architecture appears in medical imaging support tools, financial risk platforms, and military logistics systems. The common thread is a recognition that AI can process data at speeds and scales humans cannot match, while humans retain contextual judgment, ethical accountability, and the ability to handle genuinely novel situations.
The Road to a Nationwide Rollout Across US Airspace
Starting in Washington is a deliberate strategic choice. The DC corridor's density and complexity make it a rigorous proving ground. If SMART can handle the layered challenges of that airspace — commercial traffic from three major airports, military flight restrictions, and high general aviation volumes — it will have demonstrated performance under near-maximum stress.
The planned expansion to cover the full scope of US national airspace is the longer-term objective, but that timeline remains contingent on how the Washington deployment performs. $875 million is a substantial procurement figure. For context, the FAA has historically struggled to modernize its technology infrastructure on budget and schedule — the agency's NextGen modernization program, launched in the early 2000s, ran well past initial projections before reaching full operational capability. SMART represents a more focused, AI-specific investment, but institutional pressure to deliver concrete results is real.
Federal AI procurement at this scale is still relatively uncommon, which makes the SMART program notable beyond aviation circles. Agencies have been cautious about large-scale AI deployments in safety-critical systems, and the FAA's commitment signals growing institutional confidence that AI is mature enough for high-stakes operational use. It also establishes a precedent for phased deployment architecture — an approach with implications for similar programs in energy grid management, freight logistics, and emergency services.
Aviation Experts Weigh In on AI Governance and Safety
Philip Mann brings a specific and credible perspective to the question of how SMART should be introduced. Mann spent 17 years working at the FAA in multiple roles before moving to the private sector, where he now serves as principal consultant at Vector Strategic Consulting LLC, advising clients on aviation safety and AI governance. His experience spans both the regulatory culture inside the FAA and the outside advisory perspective that consultants provide.
Mann's assessment of the DC-first deployment strategy is clear: starting in limited scope before going nationwide is the "right call." That framing reflects a core principle in safety-critical AI deployment — controlled exposure, real-world validation, and iterative learning before full-scale commitment. It also draws on hard lessons from past technology rollouts in aviation, where systems that performed well in simulation sometimes revealed unexpected edge cases at operational scale.
The governance questions surrounding FAA AI air traffic control systems extend beyond technical performance. Who bears accountability when an AI advisory recommendation is followed and something goes wrong? How are the underlying models updated, validated, and audited over time? What happens when conditions fall outside the distribution the models were trained on? These are not hypothetical concerns. They are the questions that safety professionals, regulators, and legal frameworks are actively working through as AI becomes embedded in operational infrastructure.
The phased approach Mann endorses addresses both technical validation and the institutional learning required to answer those accountability questions responsibly before full national deployment.
Implications for Airlines, Passengers, and the Future of Air Travel
For travelers, the most tangible potential benefit of SMART is reduced congestion-related delay. Air traffic congestion in the United States costs the industry billions of dollars annually and produces cascading disruptions that ripple through airline networks for hours or days. A system capable of predicting and mitigating conflicts earlier in the flow management process could meaningfully reduce the frequency of ground stops, holding patterns, and compounding downstream delays.
Airlines stand to benefit from more predictable operations, which translates into better fuel planning, crew scheduling efficiency, and improved passenger outcomes. Efficiency gains from even marginal improvements in traffic flow management — across tens of thousands of daily flights — compound into substantial operational value over time.
For the broader trajectory of FAA AI air traffic control development, SMART's Washington deployment marks a threshold: the transition from AI as a research and planning tool to AI as an active participant in daily national airspace management. Whether that threshold becomes the foundation for transforming how the US manages its skies — or a cautionary lesson in the gap between promise and operational reality — depends on how the system performs in the months ahead.
The controllers watching their SMART advisory displays in the coming weeks will be navigating that future in real time. They probably won't think of it that way. That may be the best possible sign.
Source: Ars Technica - All content
