The FAA is racing to modernize an air traffic control network it admits is "chronically understaffed" and technologically obsolete, but the agency's push to deploy AI tools has drawn resistance from airlines wary of the pace and scale of the rollout.
The tension sits at the center of a broader workforce and infrastructure crisis the agency can no longer paper over. FAA Administrator Bryan Bedford has publicly acknowledged that the national air traffic control system cannot keep up with demand, telling reporters the operation "is designed to be chronically understaffed." His agency's own numbers bear that out: only about 11,000 certified professional controllers are currently deployed across more than 300 facilities nationwide, against a full staffing target of 12,563, a gap of roughly 1,500 controllers that leaves towers and radar rooms stretched thin every day.
The shortfall is not new. But the FAA's proposed remedy, a sweeping three-year workforce and technology plan that leans heavily on artificial intelligence, has forced a question the agency has not answered to the industry's satisfaction: can software safely fill the holes that human beings cannot?
The FAA's 2026, 2028 Air Traffic Controller Workforce Plan, detailed by Fox News, lays out an aggressive hiring schedule: 2,200 new controllers in fiscal year 2026, 2,300 in 2027, and 2,400 in 2028, nearly 6,900 total. The agency is also expanding partnerships with colleges and technical schools to widen its recruitment pipeline.
But hiring alone won't close the gap quickly. Training a fully certified controller takes years, and retirements continue to drain the experienced workforce faster than new hires can replace them. Rising air travel demand compounds the math. Bedford himself conceded the math is daunting.
That is where AI enters the picture. The workforce plan calls for replacing analog systems with fully digital networks and incorporating AI-driven tools into day-to-day operations. The ambition is to give controllers better situational awareness, reduce workload in congested airspace, and eventually automate routine tasks that currently require human attention.
Bedford framed the shift as unavoidable.
"We can't continue to operate the same way and expect better results. We're changing how we hire, train and schedule our controller workforce, and providing them with the state-of-the-art tools they need to succeed."
The gap between the FAA's AI ambitions and its current infrastructure is jarring. Some FAA facilities still rely on floppy disks, hardware that went out of mainstream production more than a decade ago. Workforce scheduling and timekeeping at many locations are still handled manually by local facility managers, a process that belongs in a filing cabinet, not a federal agency responsible for the safety of millions of passengers a day.
That contrast raises a practical concern airlines have been voicing behind closed doors: if the FAA cannot modernize its scheduling spreadsheets, how will it safely deploy artificial intelligence in live airspace? The question is not hypothetical. Controller fatigue and staffing lapses have already contributed to real-world safety incidents. An investigation at LaGuardia revealed controllers left shifts early before a fatal runway crash, exposing the kind of systemic discipline failures that technology alone cannot fix.
Airlines operate on thin margins and tighter schedules than ever. When the air traffic control system hiccups, whether from staffing gaps, weather, or technical failure, the cascading delays hit carriers and passengers hard. Recent FAA ground stops at seven major airports during storms stranded thousands, a reminder of how little slack the system has.
The airline industry is not opposed to modernization. Carriers have spent years calling on the FAA to upgrade its equipment and hire more controllers. But Business Insider reported on the friction surrounding the agency's AI contract push, capturing an industry caught between wanting faster fixes and fearing that the FAA is moving too quickly on unproven technology without adequate consultation.
The concern is straightforward. Air traffic control is a zero-error environment. A bad software update in a bank costs money. A bad software update in a radar room costs lives. Airlines want validation, testing timelines, and fallback procedures spelled out before AI tools go live, not after.
Aviation safety events in recent years have only sharpened that skepticism. An Amazon cargo jet crash that shut down Miami's airport and delayed hundreds of flights showed how a single incident can ripple across the national network. Controllers are the last line of defense in those moments. Replacing or augmenting their judgment with software that hasn't been stress-tested in real-world conditions is a gamble carriers are not eager to take.
Give Bedford credit for one thing: he said the quiet part out loud. His admission that the system "is designed to be chronically understaffed" is not a staffing complaint, it is an indictment of decades of federal workforce management. Congress has known about the controller shortage for years. The FAA has known. Airlines have known. Passengers stuck on tarmacs have lived it.
The agency's plan to hire nearly 6,900 controllers over three years is the most aggressive recruitment push in recent memory. Expanding college and technical school partnerships is a sensible step. But even the FAA acknowledges the shortage will not be resolved quickly, because training pipelines are long and retirements keep coming.
So the agency is layering an AI strategy on top of a workforce strategy on top of an infrastructure that, in some buildings, still reads data off magnetic disks. Each layer depends on the others working. If hiring falls short, as it has before, the pressure to lean on AI grows. If AI tools aren't ready or aren't trusted, the burden falls back on an exhausted controller workforce already logging excessive overtime.
The FAA is also trying to modernize its regulatory posture in other areas simultaneously. The agency has been working to clear regulatory paths for air taxis and supersonic flight, adding to an already crowded reform agenda. Ambition is not the problem. Execution is.
The flying public has no vote on which software the FAA installs in its control towers. Passengers cannot opt out of an AI-managed approach sequence. They trust the system because they have to, and because, for decades, American air traffic control has been the safest in the world despite its aging bones.
That trust is not a blank check. The FAA owes airlines, controllers, and the public a clear, transparent testing and validation process before AI tools go operational. It owes Congress a realistic timeline that accounts for the training bottleneck, the retirement wave, and the infrastructure rot that makes some facilities look like museums.
And it owes taxpayers an honest answer to a simple question: if the agency couldn't keep its staffing levels adequate for twenty years, why should anyone believe it can manage a simultaneous hiring surge and an AI deployment without cutting corners?
Modernization is overdue. But rushing untested technology into a zero-margin safety system to compensate for decades of workforce neglect is not reform, it is a bet placed with other people's lives.