'Cool, not creepy': Expedia Group's AI ambitions
It’s all about big targets, big systems and big data
AI is all the rage in 2023, and it's not limited to chatbots; travel giant Expedia is building it into the tech stack from bottom to top.
"We actually look at it across our entire business," says chief architect Rajesh Naidu, naming two examples: Smart Shopping for customers and fraud prevention for suppliers.
Smart Shopping uses machine learning to personalise the booking experience for repeat customers, prioritising your preferences.
"We may know that you always like to have breakfast and you want the WiFi. We will just surface those to you, so you don't need to sift through and find it."
Personalisation is a "big focus" for Expedia's strategy, and not just for customers. Suppliers are included, too.
"On the supplier and the partner side we are doing things around AI and ML to help them better list and sell their property; so, making sure they have the right data that has been published.
"We also give them insights into a guest experience score, which is not just about what the guest review is saying, but also things like about the customer, relocations...and bringing that together.
"We can kind of become like your virtual concierge."
But, Rajesh adds, "The key thing for us is it's got to respect the customer preferences. We want it to be cool, not creepy.
"We don't want it to be like some of these other AI-ML things, which infringe [privacy] and know so much about you."
'Cool, not creepy': Expedia Group's AI ambitions
It’s all about big targets, big systems and big data
AI is all the rage in 2023, and it's not limited to chatbots; travel giant Expedia is building it into the tech stack from bottom to top.
Fighting fraud
It's a worthy goal, but difficult to achieve; AI, especially for recommendations, relies on massive datasets, so by its nature has to know a lot about an individual to be useful.
Somewhere Expedia can leverage that knowledge - without the creep factor - is in its second use case: fighting fraud.
"Most people think of credit card fraud and yes, that is something we look at. But in our case, we also have a lot of fraud that can happen [on the platform]. We've got to protect the travellers, but we also want to protect our suppliers and partners. So, in this case, we sift through that and look at what kind of behaviour is there, is there any early alerting we need to provide.
"Last year we prevented £2 billion worth [of fraud] through AI and ML."
Open architecture
In its new guise as a technology company that also does travel, Expedia want to use its experience in combatting fraud - and more - to help others.
"We are opening up our architecture to our partners and our third-party contributors. We are launching a platform focusing on the developers and the developer experience."
The platform will be "like Shopify," with a developer portal and APIs to integrate services into an existing model. The first pilot, released earlier this year, is fraud-prevention-as-a-service.
"A company who's interested in leveraging our fraud prevention capabilities has access to these APIs that they can then integrate into their process."
Expedia is running an incubator programme with startups under this new model - the aim being that they will build new capabilities on top of the company's existing platform.
"These will be extensions they will offer, which we can plug into our products, as well as open it up to the marketplace.
"This developer experience is a new model for Expedia, which really started in the last year and a half."
'Cool, not creepy': Expedia Group's AI ambitions
It’s all about big targets, big systems and big data
AI is all the rage in 2023, and it's not limited to chatbots; travel giant Expedia is building it into the tech stack from bottom to top.
Plugging in to ChatGPT
The developer experience is not the only way Expedia Group is using extensions and plug-ins.
ChatGPT, this year's killer app, made waves in April when it announced plug-ins for a range of websites, including Expedia.
That means a user can start planning a trip in ChatGPT and then choose the Expedia plugin to "bring that trip to life" - including live information on the availability and price of flights, hotels, holiday rentals, activities and car rentals worldwide.
The function got plenty of attention, but Expedia Group didn't build a plug-in and call it a day. The tech team also developed a product that works in its own iOS app.
"Expedia members can start an open-ended conversation in English and get recommendations on places to go, where to stay, how to get around and what to see and do, based on the chat.
"At the end of the conversation, recommended hotels automatically save onto a new 'trip' in the Expedia app, helping members stay organised and making it easier for them to start choosing dates, checking availability, and adding on flights, cars or activities."
Expedia Group is far from the only firm using ChatGPT, but it is one of the few to have an official partnership - probably helped by the fact that OpenAI CEO Sam Altman sits on the board.
The company gets marks for trying to make its use of ChatGPT more than a gimmick - and then a few more for actually making sure the data is accurate.
Generative AI like ChatGPT is "really good on known information," but also has a tendency to hallucinate. That makes it important to make sure it's getting good data, especially in travel where live information can change rapidly.
"This means focusing on the data quality and the curation of the content. We are constantly evaluating the quality of the training data to ensure relevance and accuracy for our travellers."
Safety and security are key elements to consider with any AI implementation, especially very new systems like ChatGPT. Rajesh emphasises the importance of testing:
"It's important that we carefully scrutinise and evaluate how we use AI and specific models before implementing, since we need to ensure they are not just secure and unbiased but also of real benefit to our travellers and partners."
An automated, curated future
Expedia Group has invested heavily in automation already, with "about 85%" of back-end processes already using some form of the tech.
The company shows no signs of slowing its investment. It sees a future where AI can add contextual information for customer service agents, helping them cross- or upsell; where business teams don't have to rely on technology teams to provide them with "the framework or the guardrails"; and where low- and no code helps build new tools in record time.
The ambitions are lofty, but - the company thinks - reachable, thanks to the existing architecture that Rajesh and his team has spent time preparing.
"When we look at AI and ML, we try to look at it on all the different use cases that we can: traveller, partner and our own internal business. And again, it's all because of our data, as well as our infrastructure."