When KAYAK opened in 2004, a website was a marketer's best friend. Home pages are the new storefront, everybody said, because they were. But by 2017, less than half of the 2 billion requests sent through the travel search engine were via desktop. The others were mobile as the phone became the storefront. And now, chief scientist Matthias Keller says, another shift is coming as KAYAK uses artificial intelligence (AI) to connect.

Specifically, the company launched an Alexa voice app powered by natural language processing -- an area of machine learning that teaches computers to use language like people. Voice assistants not only have to understand what consumers say, but must be able to search KAYAK data for results then answer in a way people can understand. When this conversation's part of the buyer's journey, there's an additional layer: Responses have to be on message and true to brand.

Don't Wait for Things to Be Perfect

Building the app wasn't easy. Keller admits, "It is not perfect at all." There are too many steps to search, he explains, making it harder to use. Whereas customers can get site results by entering their dates and hitting go, the average Alexa search takes five commands. This lengthier process means flight prices could even rise during your request. Also, he adds, "the KAYAK Skill doesn't have [your] calendar," which makes it cumbersome for customers to use the company's flexible travel dates feature. 

He points out that KAYAK's 2004 website wasn't perfect either, and from a marketing perspective, the whole point of AI isn't to build amazing tech -- it's to meet customers where they are and connect.

Related Story: 7 Tips to Better Voice Search Optimization

Delivering an Experience

"It's not just about searching; it's also about traveling or even planning travel," Keller explains. People who use KAYAK don't want data; they want an experience. Fortunately, that's what the company sells, as Keller shared, one day he wants the tech to store a customer's vacation wish list or -- better yet - to offer predictions on where you should go.

Learning Opportunities

"It's not easy right now for a machine to think about it and make the perfect recommendation," he says, adding that to be effective, AI needs the right data. When someone asks the app for family vacation ideas, how big is their family? Are their kids teenagers or toddlers?

Multiple Languages Presents Challenges

Language usage can also differ, presenting problems of its own. A customer might ask for cheap travel tips, but "cheap" doesn't mean the same price point for everybody. Or, he suggests, the customer might say, "I wanna spend 10 days in southeast Asia and I have no idea about southeast Asia and it has to be great." So what does that mean?

Then there are the literal languages we speak. The app works in English, Spanish, Portuguese, Korean, and French. Markets like Germany, Keller says, are tough because of the country's strong focus on data privacy, but Brazilian customers love it. "Right now we see the highest interest in the United States, Latin America, and Asia," he adds.

Related Story: Searching for Brand Success With Voice Assistants

Another Way to Connect

Despite its issues, using voice technology to connect with customers has its advantages. KAYAK doesn't have call centers, so for people who prefer speaking to typing, the app is a voice alternative. And since customers don't touch a keyboard or screen to engage, it also provides a non-physical connection, making search more accessible. "We really believe that travel should be hands-free, seamless, and simple," Keller explains, so the way KAYAK reaches its customer base should be too.