American Airlines is using AI to predict missed flights

Few things in travel sting quite like sprinting across a terminal only to watch your connecting flight pull away from the gate.

American Airlines is now leaning on artificial intelligence to stop that from happening. In a recent company podcast, Chief Digital and Information Officer Ganesh Jayaram revealed that the airline is deploying AI to predict whether passengers are going to miss their connections at major hubs.

Holding planes instead of stranding passengers

The real surprise is how the airline responds to those predictions. Instead of simply letting passengers miss the connection, American uses the system to intentionally delay departing flights.

Jayaram noted that holding certain flights at hubs allows incoming travelers to actually make their connection, keeping their arrival at the final destination as close to on-time as possible.

Timing is brutal in aviation. Data from platform OAG shows nearly a quarter of all commercial flights in the U.S. were delayed by at least 15 minutes last year.

OAG pointed out that weather is only part of the story. About 30% of delays stem from airline and airport inefficiencies, while another 30% come from aviation infrastructure issues like aging air traffic control systems.

Reworking the tech stack from phone to tarmac

To support these new systems, American bumped its technology budget by 20% last year and brought in operational experts from other industries to overhaul legacy workflows.

For travelers, American plugged generative AI directly into its existing chatbot. When bad weather derails travel plans, customers can now use the bot to find alternate routes and rebook flights immediately without waiting in terminal lines.

The airline also refreshed its mobile app, adding Live Activities support on iPhone and Apple Watch for real-time gate updates, seat assignments, and landing times. On the ground, new hardware cut airport check-in times to a couple of minutes.

Behind the scenes, internal developers are leaning on AI coding tools to accelerate software upgrades. Jayaram noted that before rolling any of this out, the airline spent considerable time establishing an AI governance framework to handle privacy and data protections.

Why it matters

For anyone who flies regularly, missed connections are the single biggest point of friction. Being able to rebook through a genuinely capable chatbot saves hours of standing in customer service lines, but holding a flight for you changes the journey entirely.

We think using predictive AI to delay a departure is a remarkably clever move on paper. Keeping thirty connecting passengers on their journey often costs an airline far less in vouchers, hotel rooms, and rescheduled flights than letting them get stranded.

However, the strategy carries an obvious operational risk. Holding a plane for connecting travelers is great for the people sprinting down the concourse, but it tests the patience of everyone already strapped into their seats. It also risks triggering cascading gate delays down the line.

Furthermore, AI cannot fix aging government air traffic control systems, which OAG notes cause nearly a third of all flight holdups. American can optimize its internal clockwork, but the broader airspace remains beyond its algorithmic control.