Will AI Replace Air Traffic Controllers? NASA Says Not Anytime Soon. Here Is Why.

The Job Where AI Cannot Afford to Be Wrong

At any given moment, there are roughly 5,000 aircraft in the sky over the United States. They are converging on airports, climbing to cruising altitude, crossing paths at different speeds and altitudes, and each one contains anywhere from two to four hundred human lives. According to the U.S. Bureau of Labor Statistics (2024), the people responsible for keeping them all apart earned a median salary of $144,580 in May 2024 — with the top 10% earning more than $210,410 [Fact]. There are only about 24,000 of them in the country [Fact]. And they operate under a simple, non-negotiable rule: zero tolerance for error.

This context explains why air traffic control occupies a unique position in the automation landscape. The technology to automate significant portions of the job exists. AI systems can track aircraft, calculate optimal separations, predict conflicts, and suggest routing changes. NASA has invested millions in research on AI-assisted air traffic management through programs like AATT (Advanced Airspace Technology and Transition) and its successor initiatives.

But our data shows a profession that is being augmented, not replaced. Air traffic controllers face an overall AI exposure of 38% and an automation risk of 26% [Fact]. The BLS projects 1% growth from 2024 to 2034, with about 2,200 annual openings — most created by retirements and transfers rather than new positions [Fact]. The numbers tell a story of stability, not disruption.

This augmentation pattern is exactly what the OECD Employment Outlook 2023 found across member economies: at the early stage of adoption, higher AI exposure among high-skilled workers tends to create new tasks and jobs rather than destroy them, and the wage benefits of AI have so far concentrated among high-income, highly skilled workers [Fact]. Few jobs sit more squarely in that high-skill, judgment-intensive category than controlling the airspace over a continent.

What AI Already Does in the Tower

The task-level data reveals where automation has made real progress.

Monitoring radar and flight data displays shows 62% automation [Fact]. This is the most automated task in air traffic control, and it makes intuitive sense. AI excels at pattern recognition across large data sets. Modern radar systems use algorithms to filter noise, track multiple targets simultaneously, predict trajectories, and flag potential conflicts before they become dangerous. The Traffic Collision Avoidance System (TCAS), installed on all large commercial aircraft, is essentially an AI system that has been saving lives since the 1990s.

Controllers no longer stare at raw radar returns and mentally calculate where each aircraft will be in three minutes. The software does that. What controllers do is interpret the software's output, assess whether its recommendations make sense given weather, traffic flow, runway conditions, and the dozens of other variables that algorithms handle imperfectly.

Calculating separation distances and sequences is at 55% automation [Fact]. Arrival management systems like AMAN (Arrival Manager) and departure sequencing tools calculate optimal spacing between aircraft based on aircraft type, weight category, wind conditions, and runway configuration. These tools are sophisticated and generally reliable.

But "generally reliable" is not the standard in aviation. The standard is "always reliable." When the system suggests a sequence, a controller evaluates it against their knowledge of current conditions, recent pilot communications, weather developments, and the specific capabilities of each aircraft. The algorithm might calculate that a 737 can follow an A380 with standard separation. The controller knows that the specific A380 on approach today reported severe turbulence in its wake, and adds extra spacing.

Where Humans Are Non-Negotiable

Issuing clearances and instructions to pilots sits at just 30% automation [Fact]. This is the communicative core of the job, the actual act of telling a pilot what to do and confirming they understood it correctly. Automated systems can generate draft clearances, and data link communications (CPDLC) can transmit routine messages digitally. But the real-time voice communication between controller and pilot remains essential.

Why? Because context matters in ways algorithms cannot fully capture. A controller hears hesitation in a pilot's voice and asks if everything is okay. A controller knows that the pilot of a small regional jet has less experience with low-visibility approaches and provides extra guidance. A controller detects the early signs of a communication breakdown and switches to simpler language.

Coordinating emergency responses is at just 18% automation [Fact]. When something goes wrong in aviation, the controller is the first responder in the sky. An engine failure, a medical emergency, a bird strike, a security threat, each requires immediate, adaptive, judgment-driven action. The controller must simultaneously clear airspace, coordinate with other sectors, communicate with the pilot, alert emergency services, and maintain separation for all other traffic.

No AI system in operation or development can replicate this kind of multi-domain, real-time, high-consequence decision-making. The FAA has been explicit about this: their NextGen modernization program is designed to give controllers better tools, not to replace them.

It is worth being honest about the counter-argument, though. A 2024 study published on arXiv — "High-skilled Human Workers in Non-Routine Jobs are Susceptible to AI Automation but Wage Benefits Differ between Occupations" — found that even highly skilled workers performing analytical, non-routine tasks are exposed to AI, and that wage outcomes vary sharply by occupation rather than following a single predictable pattern [Claim]. The lesson for controllers is not "you are immune." It is that exposure does not mean replacement, and that the occupations which keep their human edge are precisely those where a wrong answer is catastrophic and accountability cannot be outsourced to software.

The Staffing Crisis Hiding Behind the Numbers

The 1% growth projection masks a more urgent reality. The air traffic control workforce is aging. The FAA has struggled with recruitment and retention for years. The mandatory retirement age is 56. Training takes years. Washout rates are high. The result is that the profession is not facing a surplus of workers that automation might displace. It is facing a shortage of workers that automation might help manage.

This is the opposite dynamic from most professions we analyze. AI in air traffic control is not a threat to employment. It is a potential solution to an understaffing crisis. If AI tools can handle more of the routine monitoring and calculation, each controller can manage more traffic, partially alleviating the workforce shortfall.

What This Means for Air Traffic Controllers

If you are an air traffic controller or considering the career, the automation outlook is among the most secure of any profession we track. The combination of extreme safety requirements, regulatory conservatism, the irreducible importance of human judgment in high-consequence decisions, and an ongoing workforce shortage means this job is not going anywhere.

The tools will get better. The radar displays will get smarter. The sequencing algorithms will become more accurate. But the person in the tower or the radar room, the one who hears the stress in a pilot's voice, who remembers that the taxiway is icy from yesterday's freezing rain, who makes the call to hold all departures when something feels wrong before the data confirms it, that person is not being automated away.

At $144,580 median pay, about 24,000 positions, 26% automation risk, and 1% projected growth [Fact], air traffic control is one of the most automation-resistant high-paying professions in the American economy.

See detailed automation data for Air Traffic Controllers

---

_AI-assisted analysis based on data from Anthropic Economic Research (2026), Eloundou et al. (2023), Brynjolfsson (2025), and BLS Occupational Outlook Handbook. Automation percentages reflect task-level exposure, not wholesale job replacement._

Update History

• 2026-05-22: Added primary-source citations — BLS 2024 wage/outlook data (median pay corrected to $144,580, 2,200 annual openings), OECD Employment Outlook 2023 on AI augmentation of high-skilled workers, and a 2024 arXiv study on AI susceptibility in non-routine jobs.
• 2026-03-24: Initial publication with 2025 data snapshot.

Related: What About Other Jobs?

AI is reshaping many professions:

• Will AI Replace Product managers?
• Will AI Replace Devops engineers?
• Will AI Replace Software Developers?
• Will AI Replace Nurses?

_Explore all 470+ occupation analyses on our blog._