Automation of airline customer service – the predictability model

Automation of airline customer service – the predictability model

In the last post, I had written about price as the one factor that will drive how people accept automation in the air travel industry. The function of price, in this context, is to help compensate for the absence of one characteristic (human element) of a service which passengers are so used to. If, in some period of time from today, it becomes emotionally acceptable for passengers to travel in an aircraft without the presence of a trained cabin crew for support, we need to have some idea of what that will look like.

The enabling keyword is 'predictability' in this context. With current technology, it is reasonable to assign to machines all those tasks with a high level of predictability, as well as those around which we have significant amounts of data, which the machine can read, analyze and detect sufficient patterns/trends from to make predictions about future behavior. A human presence may be required only to manage exceptions, but considering the high level of human interaction on the service delivery side in this particular industry, this may be more common than one might expect. Apart from the weather (which is not entirely unpredictable with today's supercomputers), the biggest unpredictability a properly managed airline's crew may have to face is likely to be customer behavior. Of course, this cannot be avoided completely, but there certainly are ways to address it effectively.

A passenger's interaction with the airline begins at the reservation stage, if we discount the possibilities of AI prior to that in selecting optimal routes and operations planning. OTAs are growing in their share of the travel market and chatbots are reducing the human element in this part of the value chain in quite a promising manner. The (un)predictability of customer behavior in this phase can be managed in two ways – either by covering all possible options a customer might need, or by limiting the options in process flow available in the user interface, to a point where the user experience isn't ruined. For instance, historic and current data may be used to predict what a customer typically asks for. Does this passenger have a history of showing up with oversize baggage and refusing to pay the fee? Is he/she travelling in a group with very old or very young passengers? Clarifying in advance whether his specific likely needs are allowable or not (as opposed to a very long generic list of dos and don'ts) will go a long way towards avoiding conflict at the gate/counter.

Once the tickets are booked, and the passenger arrives at the airport terminal, he/she may scan his tickets (printout or digital) at the automatic gate to gain access. Bags can be put on conveyor belts (one for each type – cabin and checked) from which the machine can flag and divert suspicious pieces, as well as those not meeting size and weight limitations, to a special belt where it may be subjected to further examination or other transactions (excess baggage fees, etc.) as the case may be. In a completely automated system, any excess charges may be deducted directly from the customer's account, or it may be set up as a condition for boarding along with a suitable payment gateway. Security checks on passengers can follow a similar pattern. There is clearly very limited scope for negotiation here, given the absence of a human, so while it may cause initial teething trouble this will improve passenger discipline over the long run.

On a typical day, many tasks of customer service have a high degree of predictability and can be automated. So boarding is likely to be a breeze as long as there are no confusions regarding your baggage and boarding pass. Once inside, passengers can find their own seats, despite missing the welcoming smile and greeting from the cabin crew. Once at their seat, the passenger may notice a small addition to the inflight entertainment system; this may be a dispenser for water and other drinks (depending on flight policy). For larger variety of offerings, a vending machine can be set up in an accessible spot.

I'm sure the point is evident by now. All subsequent touchpoints need to be designed to enable easier self service. Sufficient cooperation from passengers is an essential factor – and one that cannot be taken for granted due to human errors and behavioral inconsistencies. In other words, if everybody agrees to follow the rules, then things will work out perfectly (as far as can be controlled by the airline) but if someone/something falls very far out of line, then AI may not be entirely reliable. Predictability is a necessary condition for currently available technology.

On the other hand, predictive analytics has a superhuman advantage of being able to make predictions about things that could happen, before they actually do. Therefore, instead of having a human presence to solve issues, AI could predict what issues are likely to happen and preempt them to some extent. Is there unusually thick traffic in the city today? Then expect a few people to arrive late for their flight. Is this flight heading from a warm city to a very cold one? Then expect a lot of heavy woolens in the baggage, which could make some of the bags overweight. Might be a good customer service gesture to increase baggage limits slightly, or may be increase tolerance for slightly over-the-limit bags.

Complete automation is therefore neither immediately feasible from a technical standpoint nor advisable from a business perspective in some ways. A human presence – and a well trained one at that – would be required for emergencies and could add a lot of value. Automation would eliminate much of the distractions for this individual (or small group of individuals) and let them focus on the vital aspects of the job which don't lend themselves to predictability as easily. But it is a matter of time, and one more factor. A behavioral change in customers is also essential to facilitate such a move – more to ponder on that in an upcoming post.​

Rate this blog entry:
2
Desire to destination: Why social networks aren’t ...
Flights staffed by AI: The one driver for customer...
 

Comments

No comments made yet. Be the first to submit a comment
Guest
Friday, 22 September 2017

Captcha Image