The entire airline industry is at first glance highly technical and, at the same time, highly dependent on personnel to handle tough issues. The big story this week for United Airlines -- which is quickly turning into a public relations nightmare and even creating conflicts with China -- has to do with a passenger who was forcibly removed due to overbooking the seats.
There are many variables to the story -- yet, it’s a wake-up call because it could have been prevented using artificial intelligence. Even if the overbooking is partly a way to ensure that flights are always full, A.I. could handle that problem.
Here’s how it would work. Because there is already a wealth of data related to passenger counts, who needs to get to their destination and when, and even which employees are on stand-by for a given flight, machine learning could determine who should be allowed to board and when. Not allowing someone to board is a different issue for airlines than asking someone to disembark. It has a different set of security parameters, and a different set of passenger relation issues.
Say the A.I. model determines that there is a flight to Miami with 471 seats open on a Boeing 777. Flight attendants can’t really manage all of the data involved, and there are systems today that show models for who is going to board and where they will sit. What airlines are not doing today is determining, in mere seconds, who is actually going to board based on up-to-the-minute flight changes and delays. This is what adds to the complexity for agents. I remember on one flight being shuffled to a later route due to weather. I met an agent, she made a change, and I walked over to another counter and boarded that flight a few minutes later. No human could keep up with that and deal with the interface that shows all of these changes in real-time.
In fact, it is entirely up to the agent. The agent simply looks for an open seat. There is some statistical modeling today, A.I., and machine learning that makes this smooth, but nothing that assists gate agents as easily as Amazon Alexa ordering flowers.
An A.I. could keep up, though. There could be one moment, based on all the last minute changes and delays, when the model determines that -- “OK, there’s a good statistical chance that you will need to ask four passengers to deboard and that could create problems. It is better to wait or not allow those passengers to board until there’s a better plan.” The A.I. could gauge, using predictive analytics, when it is OK to allow boarding or when it doesn’t make sense. A really good A.I. would intervene -- maybe with a simple green or red flag -- and make sure that no passenger is ever told they have to leave a plane, based on all of the incoming data for the flight.
You might think -- there are still rare circumstances, such as a decision to allow the employees to board. The A.I. would be smart enough to know about that situation, especially since even the seemingly rare circumstances are probably not that rare. People are shuffled around, but an A.I. can keep pace. It can use data as a way to make sure there are no biases. And, it can look out into the future in ways a human can’t. For example, it can determine when one route is the last one for the day and to make sure no one boards until the seats are all perfectly assigned.
Maybe this will cause delays -- in my view, it will speed up the process. Passengers will be less frustrated if they know what is going on. This A.I. can think much faster than a human, applying logic to every route for every person to make sure the seats are all properly assigned. And, we’re not talking about turning over the actual boarding process. Humans are still there to deal with someone who might trip in the aisle or has a health issue.
Seat assignments are a surprisingly random endeavor today. An agent is obviously picking your seat, but an A.I. would be able to do a better job. No one would be forcibly removed again.
This article is published as part of the IDG Contributor Network. Want to Join?