To call "artificial intelligence" a buzzword is almost an understatement. The market is being flooded with technology in the form of apps, chat bots and software whose creators tout their use of A.I., and the travel industry is no exception


Many of the products are still in beta testing, and experts agree: A.I. is a nascent addition to travel technologies. But it's here, and it's likely here to stay.

A.I. is already a common component in our lives. Apple's Siri, Microsoft's Cortana and Google's Ok Google, all billed as personal assistants built into cellphones, employ A.I. technology behind the user interface.

At home, Amazon's Echo and the recently announced Google Home offer A.I. technology in the form of responsive speaking devices. Even Netflix employs A.I. to determine what a user is likely to want to watch next.

Henry Harteveldt
Henry Harteveldt

As for A.I. in the travel space, "Different things at different levels are usable, but a lot of it is still early on," said Henry Harteveldt, an analyst with Atmosphere Research Group.

Adam Goldstein, CEO of Hipmunk, which offers a virtual assistant powered by A.I., agreed that the technology is still in its infancy.

"As far as innings, at best, we're in the second inning," he said. "Realistically, we're probably in the first."

Harteveldt said integrating new technology into an industry is a complicated process.

"Trying to make these solutions work within a collective travel industry is not easy, because before you can implement something, it has to be developed and configured, and it has to learn the business," he said. "It has to learn the behaviors, and it has to be tested. And, of course, it has to be brought about to our industry in ways that are stable, secure, reliable and, critically, affordable."

This recent rise of A.I. in the travel industry and overall is largely due to evolving technology.

"Every year we are able to do 10 times more than we were able to do the previous year," said Devon Tivona, CEO of Pana, a virtual travel agent and mobile assistant that augments agents' skills with A.I. "I really think that this is just the beginning, not only of the technology itself but also of the applications of that technology."

Travel Weekly senior editor Jamie Biesiada tried a few A.I.-enabled apps for this report. Lola enables users to chat with travel agents, whose knowledge is augmented by artificial intelligence. She used it to plan a fall road trip to destinations in the Northeast.
Travel Weekly senior editor Jamie Biesiada tried a few A.I.-enabled apps for this report. Lola enables users to chat with travel agents, whose knowledge is augmented by artificial intelligence. She used it to plan a fall road trip to destinations in the Northeast.

Defining the term

A.I. has been a star of several Hollywood blockbusters depicting computer-vs.-human struggles. That representation is far from the case today, because the current technology is nowhere near the fictional intelligent computers of the future.

"A.I. would be an independent, rational agent that is able to learn and make its own decisions, and science and technology is not there yet," said Gillian Morris, founder of Hitlist, an app that scans huge swaths of data to alert users when there are cheap flights to destinations they have previously inputted.

While strides are being made in technology development, Morris said, "As a scientific community, no one's really sure when a truly intelligent machine will come into existence."

The artificial intelligence of today, put simply, "is technology that emulates the way the brain works," said Norm Rose, senior technology and corporate travel analyst at Phocuswright. "It's a division of computer science where many different types of A.I. exist, but the common theme is that it's an automated process that emulates the way we think."

A.I. as a whole can be broken into two categories, according to Dennis Mortensen, CEO and founder of x.ai: machine learning and natural language processing (NLP). Mortensen's company has developed a virtual scheduling assistant using A.I.

Mortensen, who has been in the computer science field for more than 20 years, said that NLP enables the human interface with the A.I. agent. Developers typically must employ NLP alongside machine learning when developing A.I. technology so the computer can understand what a human is saying or typing. Machine learning, which has a variety of subfields, enables computers to learn new things without a human having to program those things into it.

"A.I., to me, is something for where there was a human process that only humans could do yesterday, but tomorrow that process can be handed over to a machine," Mortensen said.

Miriam Moscovici
Miriam Moscovici

Further subfields of A.I. are many and varied. Miriam Moscovici, director of emerging technologies at BCD Travel, identified several in addition to machine learning and NLP: perception, i.e. the ability to perceive the environment using A.I.; and planning and decision-making, "which is not only looking at historical data but looking at probability analysis to determine the right decision to make."

"When you join multiple pieces of artificial intelligence, you start approaching what we call super-intelligence, which maybe we'll get to before we all die," she said.

A.I., with a human assist

Several companies in the travel space are employing A.I. to augment services provided by human agents. These include Lola, Pana, Mezi and Tradeshift, which recently acquired what was formerly Hyper Travel.

Pana employs travel agents and artificial intelligence to serve its clients, who communicate with the service via chat in the English app.
Pana employs travel agents and artificial intelligence to serve its clients, who communicate with the service via chat in the English app.

Pana and Lola function in similar ways. Both employ actual travel agents as well as artificial intelligence. Users download apps and communicate with the company in a chat setting. Behind the scenes, the technology uses A.I. to understand what the user wants and, using preferences and knowledge of prior travel, prompts agents with flight choices, hotels and more that make sense to offer that user.

Tivona said Pana was founded with the idea of combining online travel with agents.

"We thought, would it be possible to make the best of both worlds?" he said. "To bring in all the efficiencies that we've been brought through online travel over the last 30 years or so but then also bring back, essentially to a larger audience, the real human touch, the real value of having a live human being who's available there but doing so in a scalable way."

Lola founder Paul English said his service was founded with similar goals.

"The promise of A.I. when it comes to travel is to save time for the traveler and to get them better options, and that's what we're trying to do," he said.

Minqi Jiang, co-founder of Hyper Travel, who stayed on after the sale to Tradeshift, said it's about augmenting what a human can do, not replacing humans.

Hyper was sold to Tradeshift in June, and it has been incorporated into the company's Tradeshift Go product, a travel-purchasing personal assistant designed for small- to midsize businesses and their corporate travelers. Tradeshift Go also generates virtual credit cards for an employee's purchases, cutting back on the risk of fraud from having a physical company credit card.

Mezi is a message-driven travel service that utilizes A.I. and agents but attempts to have A.I. answer users' queries first, said co-founder Swapnil Shinde. It also includes a shopping service.

"We say that our platform is human-assisted artificial intelligence," he said. 

Hitlist gave Biesiada a heads up on flights to Spain under $500 and also connected her with some traveling Facebook friends.
Hitlist gave Biesiada a heads up on flights to Spain under $500 and also connected her with some traveling Facebook friends.

If A.I. cannot fulfill a request, a human steps in. That currently happens about 50% of the time, Shinde said, but he expects that percentage will go down as the A.I. learns more.

Right now, the human/A.I. combination is viewed as a positive partnership by many.

Moscovici said A.I. can pick up the slack, enabling companies to "reserve our talented human assets -- our most important assets, our biggest investment as a company -- for the highest value, highest impact work." She called humans and A.I. together a "winning combination."

In addition to working well, however, the partnership remains necessary because, Rose said, the technology is simply not advanced enough to work on its own.

"We're not at the point where artificial intelligence could replace a human being," he said. "It's not that smart yet. Maybe in a few years."

A.I.-only technology

On the flip side of the coin, though, some companies have gone completely virtual with their A.I. efforts.

HelloGbye is a message-based app that invites users to message travel plans via a chat interface, and it creates an itinerary for up to nine people in an initial search. No agents are used to fulfill travel requests.

"That's a key differentiator," said Greg Apple, head of marketing. "HelloGbye is fully 100% digital. There are no people."

HelloGbye is currently in beta testing, and Apple said that avoiding the use of human agents will make it more scalable.

By emailing hello@hipmunk.com, Biesiada happily and speedily had her questions answered about flight prices to Florida.
By emailing [email protected], Biesiada happily and speedily had her questions answered about flight prices to Florida.

"When you think about how scalable [the company is], we could literally have thousands of people asking us thousands of questions," he said. "We'll respond within seconds, which is impossible to do when you have people. There's only a certain amount of scalability you have when you have people."

Hipmunk also runs a travel-planning service, Hello Hipmunk, that answers users' questions about travel, including data about flights, hotels and more. It can be accessed via email, the Slack messaging system and Facebook Messenger.

No humans are involved in the answers, Goldstein said, but humans do review the answers the computer sends back to users after the fact. If a mistake has been made, the humans then train the A.I. how to handle similar requests in the future.

"We have aspirations for this to become an omnipresent virtual assistant, and if it's going to be everywhere, you simply can't scale up an army of 10,000 people from zero and have it work well," Goldstein said.

The bot craze

Messaging bots have been increasing in prevalence recently. Companies are using NLP and other A.I. technologies to create bots that can chat with users either on their own platform or through Facebook Messenger.

Rose pointed out that bots have been around for some time, but he said he believes the current surge in their development has to do with the popularity of message-based communication. He pointed to the growth of Facebook Messenger as well as other services, such as WeChat in China.

For example, Expedia recently launched a Facebook Messenger bot enabling users to book hotels, and Hello Hipmunk is available on Messenger as well. Technology company Zumata is developing a Messenger bot that will enable users to book hotels, and the travel-search website Skyscanner has launched its own bot on Messenger.

Samantha Otter
Samantha Otter

Cheapflights, a website that publishes flight comparisons and fare deals, has a flight and hotel search bot on Messenger, as well.

 Global marketing director Samantha Otter said Cheapflights launched the bot, "because we wanted to have a conversation with customers in an environment that they choose," and since Messenger has so many users, it was a natural choice.

Other companies are choosing to develop their own chat bots. Caravelo, a technology provider that focuses on ancillary-sales solutions for airlines, recently developed a white-label chat solution called Nina, geared toward travel management companies (TMCs) and airlines.

Nina could be integrated into any existing messenger platform, according to commercial director Jonathan Newman, making it easy for clients to communicate with either an airline or TMC, and vice versa.

Similarly, Tripoto, a social community of user-generated trip reports and photos, has created a chat bot that answers travel questions using both the content currently on the site and information it receives from third parties. Co-founder Anirudh Gupta said he hopes to expand the bot into a virtual travel assistant in the future.

Outside of the bot realm, some travel technology companies, including Zumata, WayBlazer and Trisept Solutions (via a WayBlazer partnership) are working with IBM Watson to generate highly personalized search results.

For example, Trisept is developing a product called Xcelerator, which works with VAX VacationAccess. A version of Xcelerator is available now, but when a new version launches later this year, agents will have access to its research and booking technology powered by Watson, enabling them to generate personalized, customizable itineraries based on preferences from a client's profile.

Travel technology powered by A.I. is also finding its way to personal assistants such as Amazon's Echo. Skyscanner has developed a skill for the Echo that enables users to search for flights via Alexa (the name by which users address the device when querying it).

As A.I. develops further, experts predict it will bring to the trade an increased level of personalization while simultaneously reducing selling time.

"I think what A.I. can do is, especially for complex decisions, it can be used to shorten the selling time and produce more meaningful results," Harteveldt said.

A.I. might not yet offer all that its science fiction brethren do, but Phocuswright's Rose predicted it will become more and more useful to agents as it is further developed.

"Don't be fearful," he said. "Embrace new technology. Find out how it can make you better, and don't dismiss it as science fiction, because it's here today."

Comments
JDS Travel News JDS Viewpoints JDS Africa/MI