Virtually Speaking: Airline Chatbots Come in All Types and Guises
Share
APEX Insight: Chatting with an airline bot doesn’t have to feel robotic. In fact, an airline’s bottom line goes up when its passengers feel it isn’t.
Mildred has an impressive profile. Before taking time off to raise her two children, she worked in public relations for a German charity organization. The daughter of diplomats, she’s lived all over the map, from France to India, but she calls Bonn, Germany, home. In November last year, she landed a job with an airline and it’s a perfect fit, especially since she knows its every route – she’s Lufthansa’s new chatbot.
The brainchild of Ivonne Engemann, a developer on Lufthansa’s Factory Sales and Digitalized Customer Experience team who programmed Mildred almost entirely on a three-hour train ride, the chatbot was created to assist customers with flight bookings as well as with general questions in English and German on Facebook Messenger. Using Wit.ai for the natural language processing platform and three APIs to source information, the virtual assistant can “intuit” where a customer is located and connect landmarks with cities. So, if you’re hoping to see the Eiffel Tower in the spring via Tokyo, Mildred can help.
Despite her charm, when it comes to chatting, Mildred is all business. “Some users ask [her] to tell jokes, but she is a serious German,” says Torsten Wingenter, head of Digital Innovations, Lufthansa. Likewise, her name, which was suggested by Engemann’s husband, was chosen because it sounds traditional and on brand, and resonates with German and English customers.
Lufthansa’s choice to create an avatar diverges from the approach taken by KLM and Icelandair, both of which use a tailfin as the profile picture for their Messenger bots. Alaska Airlines’ Jenn, the US airline industry’s first chatbot, developed with Next IT in 2008, takes a more photo-realistic approach. “Back then, adoption was a little bit lower and people were less familiar with the technology,” says Cleat Grumbly, senior vice-president of Next IT. “If you didn’t make it look like a person, they would type in keywords like a search.”
“If you didn’t make it look like a person, they would type in keywords like a search.” – Cleat Grumbly, Next IT
Making the chatbot respond like a person is also key to developing trust, Grumbly says, which is why Jenn has answers to a host of off-topic queries about her favorite food, pet and drink (Copper River salmon, a husky named Denali and coffee, but of course). Good rapport also leads to good sales. “We’ve seen an interesting lift on reservations in the travel sector where people who use the virtual agent will spend more money,” Grumbly says. Julie, Next IT’s avatar for Amtrak, generates a monthly average of 30 percent more revenue per booking than real-life reps, and saves the rail company $1 million in customer service-related costs per year. Many of Next IT’s customers, including Alaska Airlines and Amtrak, find cost savings using the technology internally as well.
Oddly, one of the questions virtual assistants get asked most frequently is whether they’re married – a query Grumbly says female avatars are asked more than male counterparts, and in the airline business, there are only female avatars. Copa Airlines’ bot, Ana, responds to the question with, “I’m engaged with my professional future. Can I help answer any question about Copa Airlines?” Jenn reminds suitors that she is not human, and when told she’s “hot,” deflects by sharing Alaska’s vacation packages to warm and sunny destinations.
“Virtually Speaking” was originally published in the 7.3 August/September issue of APEX Experience magazine.