When we launched Travelocity in 1996, it had state-of-the-art design and, along with Expedia and others, it created the largest category of e-commerce.
Sad to say, the travel user interface has hardly progressed since those days … until now.
I can say from experience, and with solid evidence, that ChatGPT will radically improve sales conversion and customer satisfaction, if implemented correctly.
How do I know?
In 2013 I got a call from the office of Ginni Rometty, then CEO of IBM. She asked if I could come to IBM and “teach IBM Watson about travel.” That led to me first learning about artificial intelligence and then evangelizing it for IBM.
Along the way I got to know the general manager of IBM Watson. When he left IBM (he was a serial entrepreneur), he decided to form several AI companies and asked me to co-found one in travel.
That led to Wayblazer, a company whose mission was to change the travel experience using AI.
Subscribe to our newsletter below
After looking at the tools available way back then, we decided to use natural language search, image analysis and text analysis to revolutionize the travel user interface.
Early on we determined that given our limited money, computing power and training ability we wouldn’t try to build a worldwide B2C app, but a more limited product that could deal with the data of one hotel chain or one convention bureau.
We built an interface that allowed the user to make requests in natural language like, “I want to go to the Caribbean in January with my family, stay in a deluxe hotel that has great golf and a wonderful spa.”
The beauty of using natural language search is that you immediately capture user intent. And intent data is gold.
Most travel sites know nothing about intent other than the usual suspects; city, date and number in party. But we knew: location (Caribbean), who they were going with (family) and that they wanted a “deluxe” hotel with golf and spa.
By having intent, we could then instantly provide a tailored list of deluxe hotels for families with great golf and spa.
But more than that, by using image analysis the first photo we showed was not of the front of the hotel, but of the spa. And by using text analysis we could select the perfect review that gushed over the great golf course.
And of course, the chat format allows you to further qualify the customer with questions about budget and time frame to hone the results just like I did 50 years ago when I was a front-line travel agent.
In the typical current travel search process the user must select the city first (and may have no idea of the best one) and dates. When they receive the result, if they are lucky, there would be a filter for spa. Clicking that, they would have to search through links or images to determine if the spa was a building full of amazing experiences or just a hot tub and a massage table. Finally, they’d read endless reviews to find comments on the golf course.
But our implementations at Wayblazer proved to us, and our clients, that a natural language interaction improved conversion substantially and our user feedback was exceptional.
We developed many use cases.
Think of the user playing “mileage roulette” on an airline site, making dozens of entries trying to find where they can use their miles to go to a beach. Why not just let them say, “Beach vacation using miles in coach in March” and give them the flights available (and if you are a savvy carrier, select and offer the best hotel for them and make money off that free flight).
Or a “Scottish hotel with falconry available June 6.” (Try that on an online travel agency or hotel website).
Building AI interfaces was significantly harder back then. We were working with stone tools, and the training was tedious and backbreaking. With GPT the "PT" in the name says it all: “pre-trained.” And while that doesn’t mean there is no training, it makes the job so much simpler. Add new graphics processing unit (GPU) chips and cloud support, and it’s a whole new world.
Wayblazer failed as a company as major brands were simply not willing to change their UI and doubted user acceptance. Many hotels outsource their web design, and the designers were stuck on the traditional ways of doing things.
Others, of course, said they could do it themselves (but never did). We also erred in trying to sell to IT, and they have endless lists of things to build. We should have (once we proved conversion increases) been selling to the CFO or CEO who could then direct the change from the top down.
Today with 100 million users and announcements of GPT plugins from Expedia and Kayak, I think that the barriers to adoption we faced back then should fall away quickly.
The major OTAs are already launching GPT interfaces and are talking about different landing pages depending on user intent (hooray!). I think we will see strong user adoption as these implementations evolve and mature.
And don’t forget AIs are “learning” systems. They grow and improve with each interaction. If your competitor has a learning AI-based interface and you don’t, they will soon have an AI Ph.D., and you will still be in kindergarten. I doubt you will ever catch up.
ChatGPT and its cousins will permanently disrupt the travel UI, and I for one will cheer the destruction of those interfaces we built back in 1996!