10 things we’ve learned from building airline bots in 12 months

We built our first chatbot over 4 years ago. It’s name is Tato, it lives on Hipchat and its mission is to help our developers by creating bookings in the test environments of our airline partners with just one line of text. Thanks to Tato we realized that a conversational interface was a great way to book airline tickets. We then set ourselves a mission: to turn messenger platforms into channels for airline retail and servicing.

In May 2016, we entered an IATA NDC hackathon in Berlin (and won!). This gave us momentum to not only design but commercialize the concept.

After launching 6 airline chatbots in the last 12 months, we thought we’d share some of the secret source behind getting bots in the airline world right.

Manage expectations

What happens when you don’t deliver on you claims, you disappoint your user. What we’ve seen is that users who are not served properly on the first experience won’t try the bot again.

Your bot cannot do everything: despite some airlines and their creators claiming so very loudly.

But the reality was that to actually do anything remotely helpful the bot handed back over to the call center agent or gave a link to the website. Users were left disappointed.

Now what we see are bots having specific, clear purpose. Often this is scaled back to single use-cases. Which again, leaves customers underwhelmed unless the purpose is very well articulated.

Make your bot part of your team

Bots are not here to replace humans.

The two most asked questions across our bots are about baggage and how to check-in online. Now, imagine being a customer service agent and answering those questions 20 times a day, 5 days a week and keeping a smile on your face.

Instead, that human capital should be used for the stuff that makes a difference for both the airline and the customer: the complex, difficult and sensitive issues that chatbots just aren’t equipped for.

Make payments easy

Make it easy for your customers to book and they will use it as a retail channel. Why shouldn’t the purchase of flight segments or associated ancillaries be as easy as amazon prime?

We believe that if consumers trust Uber or Amazon to tokenize their credit card details, why wouldn’t they trust airlines? And they do, and it works.

Be helpful

Simply being a conversational interface and providing links to airline.com is not really helpful.

Being helpful is getting things done in the background: like looking for award inventory or telling you when the price of a favorite city pair has fallen or risen.

An airline customer service representative is not there to give currency conversions or to give advice about the best pizza restaurants at your destination. But bots can do that. Airline reps are there to help with the complex and the sensitive, where bots aren’t. Bots, in turn, can broaden the capability of service provision by the airline: not just checking flight arrival information, but helping with the broader traveller needs that can’t, and shouldn’t, be delivered by customer service representatives.

Be a polyglot

English is the most spoken second language in the world. But who wants to be second?

We realized early on that in order to be successful, our bots needed to be fully fluent in the core home languages of the majority of its customers.

NLU such as wit.ai, dialogflow & Watson between them now cover over 50 languages and further regional differences. There is no excuse now for a chatbot not to recognize that ‘brizzy, straya’ is a place or that Boleto, meaning ticket, is very different in Spain and South America.

The best bots have broad vocabularies and well tuned recognition.

Connect your bot

It’s amazing how quickly a chatbot can be forgotten about after the initial hype and PR.

Where airlines have an ambition to bring bots to life as part of a digital marketing strategy, it is very clear. With the rise of bot discovery within facebook, using a chatbot as a call to action is a phenomenally powerful tool. Clicks turn into conversations, not just impressions.

Making a bot the smiling first responder to customer service enquiries can only happen if customers know your bot exists. We’ve seen big increases in bot engagement when airlines link their direct customers from initial PNR confirmations.

Be ready to train and learn

What happens at most bot launches is that the expectation managed, team playing helpful chatbot doesn’t succeed on day 1.

Good bots and their businesses know that every fallback, every unrecognized intent, every mistaken utterance is an opportunity for improvement. The teams of people that manage bots in the background (we call them trainers) are the superstars that live for improving bots, day in day out.

Good trainers recognize and create bot content proactively. Being ready to deploy meaningful content as weather events happen and maintaining relevent marketing content to keep bots fresh and useful is important work.

External customers aren’t the only users

Chatbot technology enables a single source of the truth to be available quickly, retrievable with just a sentence of text. We’ve seen call center agents and airport associates use bots to get the answers for customers instead of existing portals. Why? Its quick, its easy and the answers are right there in easy and intuitive platforms.

In our conversations with airline leaders around the world we hear of a need for digitalization of certain internal processes. Let’s see how those are approached in 2018.

Be available where your customers are

The first airline chatbots were just about Facebook Messenger. Messenger remains globally a leader but irrelevant in certain global markets.

Being available in the most important messenger platforms for your customers is critical. Bots are now possible and powerful in Viber, LINE, Telegram, Kakao talk and elsewhere. Knowing where your customers are and then building the bot for it is the right place to start: don’t expect your customers to migrate to a new messenger platform for you.

Let the data determine iterations

When chatbots were the number one conversation in the airline industry 18 months ago, the primary function was answering frequently asked questions.

Who would have thought that conversations make for great places to learn? Being able to have a one on one conversation with thousands of actual and potential customers, every day, enables a rich data foundation that provides for huge amounts of understanding. That understanding helps drive the next steps in bot utility. People ask about booking seats? Build that use case. Customers ask for updates and notifications about retail options at their departure airport? Build that use case.

Overall, what we’ve seen that when used as part of a well thought out digital strategy, where deployed with real purpose and utility, bots cut through the noise and become credible assistants.