You’ve probably noticed how, by comparison to other products and services, buying travel online can be complicated. This is because of the huge number of product variations. Research suggests that while there are around 700 billion different variations of products on sale at Amazon, Expedia customers have to pick out what they’re looking for from 1.26 quadrillion (that’s a one followed by 24 zeroes) options!
In other words, they come crashing up against what is probably the biggest limitation of search engines when it comes to using them for picking out one option from a truly vast range of possibilities.
Luckily, this is just the sort of problem that technology like ChatGPT – generative artificial intelligence (AI) chatbots built on large language models (LLMs) are great for solving. Rather than a search engine, which presents users with a sometimes overwhelming number of results that often have to be filtered to find the best options.
One way to picture the difference is to think of it as having two expert helpers on hand. One of which will give you ten possible answers to any question you ask it and let you work out which one you think is right. The other will just get straight to the point and tell us what we need to know!
So it’s interesting to see that Expedia Group has become one of the first major e-commerce operators to build ChatGPT into its app. Users of the feature (which is currently in beta for iOS users) can now plan their travel arrangements by having an open-ended “chat” with the application. This means discussing travel plans, the best options for accommodation, the sights they'd like to see, and the activities they'd like to experience – just as if they were planning the trip while sitting with a human travel advisor.
How Does Expedia Use AI to Simplify Travel Planning
Generative AI hit the headlines late last year with the public launch of ChatGPT – a chatbot built on top of the GPT-3 LLM created by OpenAI. There have, of course, been AI chatbots before; Amazon's Alexa and Apple's Siri both fall into the category. But ChatGPT amazed users by being capable of holding far more sophisticated conversations. It can also generate text, including articles, stories, scripts, speeches, poetry, and even computer code, to a standard that can often pass as human.
Because the technology is accessible via API, the developers of many services and applications are now racing to be the first to build it into their own services. For example, Microsoft (a key investor in OpenAI) added its capabilities to its Bing search engine, and the social messaging app Snapchat recently unveiled its own GPT integration.
Although they are all based around the same language model developed by OpenAI, developers are able to add their own features and functionality to their ChatGPT deployments. Bing enabled it to search the web, augmenting the training data it uses to respond to queries with more up-to-date information. And Snapchat’s integration is able to learn about its user over multiple conversations, developing its personality as it goes.
Expedia’s innovation enables it to compile handy lists of the options that are discussed during its conversations, known as Trips. Users can then refer back to these to get an overview of everything that was discussed. Vanilla ChatGPT, for comparison, tends to forget what you’ve spoken to it about between sessions.
According to Expedia, it will also let users know when there are exclusive rewards and discounts they can take advantage of directly via the ChatGPT interface.
Aside from this latest deployment of generative AI, Expedia uses AI in a number of other ways across its services. This includes providing personalized recommendations and results via its search engine.
It also analyzes historical trends in the pricing of flights in order to track price fluctuations and predict the best time to make bookings.
What Are The Challenges of Using Generative AI in Travel Planning?
One limitation that’s become apparent when it comes to generative AI chatbots is that they are sometimes prone to making mistakes!
Anyone who has used ChatGPT, in particular, for any length of time is likely to have noticed that it can fairly frequently provide incorrect information. This propensity – which affects other LLMs as well as OpenAI’s – has been referred to as “hallucination” because it demonstrates that ChatGPT can sometimes be out-of-touch with reality!
Expedia has not given specific details of what steps have been taken to address this issue. I can’t help but feel it could cause problems if it turns out that it can give incorrect or inaccurate information about travel bookings, for example stating that accommodation or venues are wheelchair-accessible or dog-friendly when in fact they aren’t!
It does, however, state that “measures have been taken to limit inaccurate results and inappropriate responses; at times, the experience may not work exactly as expected."
Time will tell how effective Experia Group has been at mitigating these issues, and reports of their success (or failure) are likely to emerge soon.
How Might Generative AI Chatbots be Used by Expedia and Other E-Commerce Platforms in the Future?
Expedia Group might be the first major e-commerce operator to incorporate generative chatbots into its core platforms, but it’s highly unlikely to be the last.
Numerous ways exist in which the technology could be used to streamline operations and improve customer experience. These include:
AI Personal shoppers that can analyze a customer's behavior and preferences and interact with them in order to provide personalized services.
It can provide enhanced customer support that can answer customer queries and complaints more accurately and deal with routine tasks such as managing returns.
It can also create personalized product descriptions so shoppers arriving at a site will see products and services described in a way that addresses how they are specifically relevant to them.
It could be used to create interactive marketing material such as personalized marketing emails or social media advertising that customers can talk to and get to answer their questions.
It can even conduct market research – collecting data from customers it interacts with as well as providing natural-language interpretation of behavioral data that can give businesses valuable insights.
Whatever useful innovative applications for generative AI emerge in e-commerce in the near future, it’s clear that the technology has huge potential for improving user experience.
Like most people, I increasingly shop for goods and services online simply because I appreciate the convenience. On the other hand, there have been times when I have found myself frustrated because choices and results are presented in an overwhelming way (this is particularly true with travel), and it’s hard to find answers to specific questions. If generative AI can help solve these problems, then it has the potential to be a game-changer for the future of digital retail and commerce.