FAA SMART: Essential Facts on the Risky $875M Air Traffic AI

FAA SMART is the Federal Aviation Administration’s new AI planning system for American airspace, and it has just become the headline in a long-running argument about how to fix air traffic control. The Wall Street Journal reported this week that the agency will soon launch the software, built by a company called Air Space Intelligence, under an $875 million deal spread over 12 years. TechCrunch summed up the plan in one line: the FAA’s fix for air traffic is “$875M worth of AI”.

The short version is accurate but incomplete. The contract was awarded in June, it pays for two systems rather than one, and the software does not direct aircraft. It is a forecasting and scheduling layer that tries to spot congestion before flights leave the gate, so that controllers face fewer conflicts once aircraft are in the air.

This article explains what the system does, what the money buys, why the FAA wants it now, how it fits into the wider modernisation programme, and which risks deserve attention. It also draws out the lessons for any organisation weighing AI decision support in a high-stakes operation, a question at the heart of most AI strategy work.

Table of contents

What FAA SMART Actually Is
Who Is Air Space Intelligence?
The $875M Figure: What It Actually Buys
Why the FAA Wants FAA SMART Now
How FAA SMART Would Work in a Day of Flying
Where FAA SMART Fits in the Modernisation Plan
The Rollout: Washington First
What Airlines and Officials Say About FAA SMART
The Risks and Open Questions
What Other Organisations Can Learn From FAA SMART
FAA SMART FAQ
References

What FAA SMART Actually Is

SMART stands for Strategic Management of Airspace, Routes, and Trajectories. The name is a mouthful, but the FAA’s own two-page fact sheet describes the idea plainly.

The official definition

The agency calls FAA SMART “a cloud-based platform system that enhances existing FAA air traffic management systems”. Using AI, it “analyzes airline schedules, weather, airport capacity, airspace conditions, and operational constraints to predict traffic flows and identify potential conflicts before they occur”. The stated goal is blunt: “zero delays at the start of the day”.

What it is not

The same fact sheet is careful on one point: “Air traffic controllers remain responsible for keeping aircraft safely separated. SMART helps them do that.” FAA SMART is a planning layer. It recommends routes, departure times and adjustments; it does not issue clearances or talk to pilots. That distinction matters for anyone worried about handing safety-critical decisions to software.

Who builds it

Air Space Intelligence, known as ASI, supplies the software. Its chief executive, Phillip Buckendorf, called the award “an exciting moment for the millions of Americans who rely on air travel every day”, and said the FAA is “embracing commercially proven technology already helping everyone from major airlines to the broader aviation community”.

Who Is Air Space Intelligence?

The contractor matters, because FAA SMART will run on its platform for more than a decade. ASI is not a traditional aviation systems integrator; it describes itself as a builder of “industry-proven AI for mission-critical operations”.

The PRESCIENCE platform

ASI’s core product is called PRESCIENCE. The company says it integrates multi-modal data feeds through an open data architecture, builds what it calls “a four-dimensional twin of your operations”, and converts that model into “lookahead simulations” that scan the future outlook and evaluate alternative courses of action. That is a close description of what the FAA wants from FAA SMART: a simulated view of tomorrow’s airspace, tested against options before anyone commits to one.

Defence and energy customers

ASI’s website lists defence work alongside aviation, including a US Air Force partnership, a Defense Innovation Unit contract for an AI-enabled joint sustainment decision tool, and a “logistics kill chain” concept for military supply. It also announced a collaboration with National Grid to deploy operational AI on the electricity grid. A vendor with defence clients will be used to security accreditation, which is an advantage for a federal system.

Every event is logged

One feature deserves more attention than it has had. ASI says PRESCIENCE “logs every geospatial and temporal event” for training, analytics and compliance, “enabling comprehensive retrospective analysis”. For a public system, that audit trail is what would let regulators and independent reviewers check whether the AI’s recommendations actually helped.

The $875M Figure: What It Actually Buys

The number in the headlines is real, but it is not the price of FAA SMART alone.

Two systems, one contract

When the FAA announced the award in June, it described “two complementary, cutting-edge technologies”. The first is Flow Management Data and Services, or FMDS, which becomes “the new technological backbone” of the Air Traffic Control System Command Center. The second is SMART, described as “an enhancement within FMDS”. Government Executive reported the whole award as a 12-year, $875 million contract covering both.

The award came three months before the headline

This week’s coverage reads like new money, but the contract dates from June. What is new is the launch: the FAA plans initial operations “in the Fall of 2026”, starting in the Washington, DC area before other regions.

Component	What it does	Who uses it
FMDS (Flow Management Data and Services)	Analyses flight plans, schedules and live positions to estimate traffic flows and capacity limits; plans local reroutes around severe weather	Managers at the FAA Command Center
FAA SMART (within FMDS)	Uses FMDS data to predict conflicts and coordinate schedules and trajectories before departure	FAA, airlines and other operators, sharing one view
Contract	$875 million over 12 years, awarded to ASI in June 2026	Department of Transportation and FAA

The yearly arithmetic

Spread evenly, $875 million over 12 years is about $73 million a year. Against the $12.5 billion Congress provided in 2025 for the wider air traffic overhaul, the whole contract is equal to about 7% of that sum. The FAA has said it needs roughly $20 billion more; against a combined $32.5 billion, the FAA SMART and FMDS deal is under 3%.

The ASI contract against the size of the overhaul (our arithmetic on published figures)

$12.5B already provided by Congress 100%

$875M ASI contract (FMDS and SMART) as a share of $12.5B 7.0%

$875M as a share of $32.5B ($12.5B plus the $20B still sought) 2.7%

One year of the contract ($72.9M) as a share of $12.5B 0.6%

A comparison, not an accounting claim

Neither the FAA nor ASI has said which appropriation pays for the contract, so the chart compares scale only. The point it makes is simple: the AI piece of the plan is a small line item next to radars, radios, fibre and new facilities.

Why the FAA Wants FAA SMART Now

The agency is attacking delays from two directions at once: more controllers, and better tools for the controllers it has.

The staffing gap

The FAA’s 2026 workforce plan sets a full staffing target of 12,563 certified professional controllers. As of April 2026 about 11,000 were deployed across more than 300 facilities, with another 4,000 in the training pipeline. It can take more than two years to certify a new controller. Even with aggressive hiring, the gap will not close quickly.

Certified controllers against the FAA’s target (FAA 2026 workforce plan)

Target 12,563

Deployed, April 2026 (about) 11,000 (87.6%)

Shortfall 1,563 (12.4%)

Hiring alone is slow

The plan targets 2,200, 2,300 and 2,400 new hires in fiscal years 2026, 2027 and 2028, a total of 6,900. Many of those will leave training or replace retirements, which is why the same plan leans on technology. One of its pillars is to “use artificial intelligence and machine learning tools to better simulate and manage NAS performance before the day of departure”. That sentence is, in effect, the job description for FAA SMART.

Conflicts are known in advance

FAA Administrator Bryan Bedford put the problem directly: “Every day, our air traffic professionals knowingly manage thousands of scheduling conflicts across the National Airspace System, which ultimately end up as delays for the traveling public.” Airlines schedule independently, each against its own data. The conflicts only surface on the day, when the cheapest fix, moving a departure time, is no longer available.

Airlines pad their schedules

Bedford told Reuters in July that airlines build significant extra time into their timetables because delays are expected. “If we pull all of that cost out of the system, inconvenience comes out of the system,” he said. If FAA SMART makes the day more predictable, the payoff is not only fewer delays but shorter scheduled journey times.

How FAA SMART Would Work in a Day of Flying

The FAA’s descriptions allow a reasonable sketch of the workflow, from long-range planning to the day of operation.

Weeks and months ahead

According to the award announcement, the combined platform lets the FAA “proactively identify delays and available airspace to mitigate them days, weeks, and even months in advance”. Airline schedules, airport capacity and known constraints are compared early, while there is still room to move flights.

The night before and the morning of

Forecasts for weather and capacity are refreshed. FAA SMART predicts where demand will exceed capacity and proposes routes and departure times. Because the FAA and airlines see the same picture, they can agree changes before passengers board.

During the day

As conditions change, the system “continuously updates forecasts and planning recommendations”. FMDS handles tactical reroutes around severe weather. Controllers still separate aircraft; the software aims to hand them a less congested sky.

After the day

The fact sheet lists faster recovery from weather and congestion events among the expected benefits. Every disruption becomes data for the next plan, which is where machine learning systems usually earn their keep.

Room for new kinds of traffic

The fact sheet also argues that FAA SMART “creates capacity for new airlines, cargo operators, and advanced air mobility aircraft”. Demand is the long-term pressure: Reuters reported that policymakers expect US air traffic to double over the next two decades, and “planning the way we always have will not keep up”, in the FAA’s words. Squeezing more flights through the same airspace is only safe if the plan is better before take-off.

Where FAA SMART Fits in the Modernisation Plan

The AI software is one strand of a much larger programme, the Brand New Air Traffic Control System, launched by Transportation Secretary Sean Duffy in May 2025.

Strand	Published scope	Status
Telecommunications	5,170 new fibre, satellite and wireless connections	57% of copper replaced by July 2026, per Reuters
Surveillance	New surface radars at 44 airports; surface awareness at 200	Radar and telecom compressed to a three-year timeline
Automation and displays	Tools at 89 airports; new displays in 435 towers	Peraton selected as prime integrator in December 2025
Facilities	First new en-route centre since the 1960s	Target: whole system by end of 2028
Traffic flow software	FMDS and FAA SMART from ASI	Initial operations fall 2026

Equipment is the bigger cause of pain

The FAA’s own fact sheet says flight-delay minutes caused by equipment problems in 2025 were about 300% higher than the 2010 to 2024 average. Better planning software cannot fix a failed radar or a cut copper line. That is why most of the money goes to hardware, and why the AI layer’s value depends on the rest of the programme landing on time.

The cloud is coming to air traffic control

FAA SMART is cloud-based from the start, and Reuters reported that the FAA will later move all 313 of its facilities from individual computers to the cloud. For an agency that until recently relied on floppy disks in some systems, that is a sharp change, and it mirrors the cloud adoption journey many large organisations have taken, with the same benefits and the same new dependencies.

The Rollout: Washington First

The deployment plan is cautious in geography, if not in timing.

Why start in DC

The software will roll out in the Washington, DC metropolitan area first, according to the Journal’s reporting. The region’s airspace is complex, crowded and politically visible. It is also where a January 2025 collision between an American Airlines regional jet and an Army Black Hawk helicopter killed 67 people, which gave the modernisation push much of its urgency.

Timelines from both sides

The FAA says initial operations begin this fall. ASI has said both systems will roll out over the next 12 to 24 months. Bedford told Reuters in July he wanted to launch the new scheduling system in September, so the programme appears close to its own schedule.

What success should look like

Neither the FAA nor ASI has published target numbers for how far FAA SMART should cut delays. Useful public measures would include on-time departure rates in the DC region before and after launch, ground-delay programme minutes, and fuel burn. Publishing them would let taxpayers judge whether FAA SMART is working.

What Airlines and Officials Say About FAA SMART

The announcement drew strong support from the airline industry, which has lobbied for exactly this kind of shared planning tool for years.

The airlines

Chris Sununu, chief executive of the trade group Airlines for America, said FAA SMART “leverages the latest technologies to make air traffic more efficient and timely while maintaining our gold standard of safety”. He added that “a program of this kind has been talked about for decades, and this administration is finally taking action”, and that airlines are “coordinating closely with the FAA” on routings and capacity information.

The Transportation Secretary

Sean Duffy framed the software as the step after hardware. “We’re already making remarkable progress on upgrading the radars, radios, and telecom wires air traffic controllers rely on,” he said. “But to really improve the flying experience for the traveling public, we must change how flights are managed.”

The controllers

The controllers’ union has not commented publicly on FAA SMART in the sources we reviewed. The FAA’s workforce plan commits the agency to work with the National Air Traffic Controllers Association “to identify and resolve bottlenecks across hiring, training, and operations”. How controllers experience the new recommendations in the DC pilot will shape whether the tool spreads smoothly.

The Risks and Open Questions

AI decision support in aviation is sensible in principle. The details decide whether it works.

Automation trust and workload

If FAA SMART recommendations are good, controllers and airline dispatchers will come to rely on them. If they are erratic, staff will learn to ignore them, and the investment is wasted. The FAA says the system will ease controller workload; that claim needs measuring, because new tools often add screens before they remove any.

Cybersecurity and resilience

A cloud platform that sees every airline schedule and every constraint is a valuable target. The programme will need strong cybersecurity, tested fallbacks and a clear plan for the day the platform is unavailable. The FAA’s history of outages, including a 2023 system failure that briefly grounded flights nationwide, makes resilience a first-order requirement rather than a detail.

A 12-year dependency

Twelve years is a long commitment in software. The public documents do not say who owns the models, how they are validated, or how the FAA could switch supplier. Those terms matter more over a decade than the headline price.

Funding uncertainty

The FAA said in December 2025 that it needed $20 billion beyond the first $12.5 billion; Bedford told Reuters in July that the agency needed another $10 billion. The figures may describe different scopes, but the gap shows that the budget for the wider overhaul is still unsettled.

What Other Organisations Can Learn From FAA SMART

The air traffic case is extreme, but its design choices apply to any operation where software advises and people remain accountable.

Keep the human in the loop by design

The FAA drew a clear line: controllers separate aircraft; the software plans. Writing that boundary into the system’s definition, not just its policy, is good practice for any AI decision-support tool.

Fix the shared data first

Much of the promised value comes from putting everyone on one picture of the day. Controllers today, the FAA says, use “numerous screens and spreadsheets”. Unifying data often delivers more than the model on top of it, and it is the part of intelligent automation projects most often underfunded.

Start where the problem is visible

Launching in one region, with a public baseline, lets an organisation measure real results before scaling. That is as true for a warehouse, a hospital or a service desk as it is for airspace.

FAA SMART FAQ

What does SMART stand for?

Strategic Management of Airspace, Routes, and Trajectories. It is an AI planning layer within the FAA’s new Flow Management Data and Services platform.