April 2018, San Francisco
TUI gives artificial intelligence a big tick, but what next?
Machine vision, says Utrip, TUI's chosen partner in AI, natural language processing and machine learning
From being one of the first to dabble in blockchain technology to driving forward with AI-fuelled partnerships, the TUI Group is on a mission to remain a travel heavy weight. In a recent move, the world’s biggest travel company has signed a deal with Seattle-based AI firm Utrip, to up its game in offering a deeply personalised experience throughout the customer journey.
The idea, David Schelp, managing director of TUI Destination Services (DS), said in a press release, is to combine its “inventory of unique destination experiences with Utrip’s artificial intelligence solution to provide our guests with the most personalised travel planning experience available today”.
TUI passed on an opportunity to hop on a call to discuss the deal further, the ever enthusiastic Gilad Berenstein, CEO and co-founder of Utrip, was keen to talk about how far his company, which launched in 2012, has come.
“The power of AI today is that we can match up people’s preferences with the inventory that has already been curated by TUI or by any of the partners we work with,” he says.
Although Utrip counts Hilton and JetBlue among its partnerships, TUI is the biggest to date, and the first in continental Europe. According to Berenstein: “TUI is integrating the data capabilities that Utrip has in a much deeper and more meaningful way that our other partners have to date”.
Right time investment
TUI and Utrip started negotiating a little over six months ago, and Berenstein has high praise for the household name in travel. Not only has TUI done a “beautiful job” in curating inventory well suited to its own customer base, for a large company, it is also “highly sophisticated” in thinking about tech and data.
This latest move could be seen as TUI investing in AI technology at the right time; timing is something brands often get wrong. For years, there has been much hype about what artificial intelligence can deliver the travel industry, but only recently has it entered the mainstream. You only have to look at how far Utrip’s technology has come since it launched back in 2014 as a B2C AI trip-planning app to help road warriors plan their itineraries. “UtripIA is [still] our core AI but back then it was the only piece that was working,” says Berenstein.
Today, however, there are three core pieces of technology that make up Utrip’s AI engine:
This is a ‘machine learning’ ecosystem that shakes things up by global aggregating content from a huge range of sources including the big boys like TripAdvisor and Yelp, but also from media outlets and Mom & Pop shops. It then merges this content with unique inventory from a partner like TUI, and any reviews they might have.
SnowGlobe uses machine learning and natural language processing (NLP) to verify accuracy of this information which can include everything from hours of operation to addresses, phone numbers and so on. NLP is also used to read through reviews and descriptions to establish relevance to a particular customer – for example, is this a good place for lunch based on what type of people frequent the place, the geography and so on.
There is also a human element to SnowGlobe. “I am a strong believer that, in the future, we will continue to see a hybrid AI develop, where human experts continue to play a role. In Seattle, for example, there is a famous chef called Tom Douglas, who can recommend a great dinner in a way that 1 million reviews cannot,” Berenstein says.
In Seattle, for example, there is a famous chef called Tom Douglas, who can recommend a great dinner in a way that 1 million reviews cannot
Yes, certain roles will become obsolete because of AI, but Berenstein wants to see companies using human experts where judgement, creativity and communication is needed, and using AI in a way that it really excels.
Getting slightly more technical is the next piece of Utrip’s AI - TravelGraph. According to Berenstein, this “is a little bit different from a SQL database that most people are familiar with”. The argument goes that graph databases perform better than relational (SQL) and non-relational (NoSQL) databases, because although data queries increase exponentially, the performance of graph databases doesn’t.
“Graph databases are a new type of technology that are specifically built for machine learning,” he explains. By building machine learning into the database, it is able to make connections between people, places and things independently. So, if an Utrip user dislikes an experience and removes it, not only does the system get to know the individual traveller better, but also the entire ecosystem, including the likes and dislikes of a particular cohort.
According to Berenstein, the recommendation engine is highly sensitive. Citing his own travel behaviour as an example, he says: “When I travel with friends, I’m a sort of a foodie millennial, but if I travel with my wife, it’s about relaxation and romance. This can be updated in real time.”
It can also be amended to fit different customer types. In the case of Hilton, for example, a Waldorf guest is very different from a guest staying at a Hilton Astoria hotel.
As mentioned earlier, this is still the company’s orginal AI and is about matching people and places. “It is the piece that helps to create dynamic packaging in our partnerships with firms like JetBlue,” says Berenstein. It works by taking feedback from users and analysing traveller behaviour to understand what similar people have enjoyed. SnowGlobe then helps to define the right experience looking not just at interests and budgets, but opening hours, geography, diversity and more.
Trendsetting and spotting
Owning the ‘over ground ‘experience is a trend that all travel brands are looking to tap into, and partnerships like this are increasingly viewed as the way to deliver commercial value. In what has emerged as a common theme, white-label technology offered by firms like Utrip, GetyourGuide and mtrip, is directly integrated into a partner’s website but that is not where it stops. The recommendation engine can also be included in emailed marketing, social media, ad retargeting and more.
But what else should firms have an eye on, in this fast-moving space?
Although, Utrip isn’t using blockchain, something that TUI has shown an interest in, in the longer term Berenstein sees some interesting opportunities with respect to transaction and loyalty management. In the shorter term, however, he believes AI photo and video recognition will be a trend to watch.
Today AI is largely blind to photo and video, so we are seeing a lot of people spending time on what is called machine vision
“Today AI is largely blind to photo and video, so we are seeing a lot of people spending time on what is called machine vision,” says Berenstein,
This is a huge area of opportunity in the travel space, where visuals are so important to a firm’s success.
Going forward, Berenstein also expects “much more advanced cohort analysis”. Using psychology and over 100 million traveller preference data points, Utrip knows that there are many more distinctions to be made between people than simply where they live or what they earn.
“Because of that, a lot of the cohort capabilities of the past, are pretty weak. Ultimately, as we look at graph databases and apply advances in machine learning to do cohort analysis based on more factors that just income and geography, we will see the industry becoming far more personalised,” he says.
Utrip is a sponsor and speaker at EyeforTravel’s upcoming San Francisco North America Summit where Berenstein will be speaking about working creatively with personalisation