Not so long ago a group of mathematicians and scientists proposed a summer workshop at Dartmouth College in New England to study a newly emerging area of research.
This new field was called “artificial intelligence” (AI) and the brains behind the summer school felt that, within just two months, they could make “significant advances … in making machines use language, to form abstractions … solve problems … and improve themselves.”
These were revolutionary ideas, but the timescale was a little out.
Today, we are still waiting for machines to get this clever. However, we are — at last — starting to see real progress thanks to computer processing power and data storage becoming less expensive. While AI has not reached the point where it can replace humans, it has notched up some impressive achievements, powers voice assistants like Alexa, and helps Amazon deliver smart recommendations.
The travel sector has also started to embrace some aspects of AI, notably in customer service. A number of airlines, including British Airways, have introduced chatbots that allow customers to ask questions or make bookings using natural language. At a smaller scale, Edwardian Hotels launched its Edward chatbot in 2016 to handle room service requests from its guests.
Other travel companies have used AI in different ways. Technology provider Avvio unveiled an AI-powered hotel booking engine which, it has claimed, could increase direct bookings by 25%. Meanwhile, Trainline has been using AI to identify train carriages that are most likely to have empty seats, and IT company SITA has developed a tool that, it claims, can help to predict airline delays before they happen.
New tools, new ideas
American Express Global Business Travel (GBT) has also been making use of artificial intelligence, both within the company itself and within the tools it offers to clients.
One interesting area is the use of AI to power recommendations and searches within GBT’s online booking tool, Neo™, which analyzes a traveler’s previous behavior, and that of similar travelers, to predict where the traveler wants to book. Neo also uses an AI-powered search engine that can “think” more like a person.
Machine learning — a subset of AI in which computers tweak their own algorithms to get better results — is at the heart of Trip Recommender™. If a traveler books through GBT and there is no hotel in the booking, the traveler is automatically emailed a curated list of hotels based on their previous trips.
Daniel Raine, consulting and business intelligence leader for GBT in the UK, says that although there has been something of a hype curve for AI and most uses have not revolutionized the world overnight like the Dartmouth College researchers optimistically hoped, practical uses are indeed starting to appear.
“Businesses have been slow to catch on. There was a disconnect caused by an initial lack of understanding and high setup costs,” he explains. “At GBT, one of our primary uses of AI is for data management, particularly for cleansing data from multiple sources such as suppliers, payment providers, and clients.”
Raine’s team is also using the technology for contract loading and interpretation to help reduce manual processes behind the scenes, including the creation of an AI bot called Kevin. “We are working with clients on airline RFPs where they receive contracts and pricing from various airlines in a multitude of formats and of differing quality,” he says. “Kevin who goes through those contracts, in whichever format they are received, finds and extracts the contractual terms and pricing. He then runs the pricing through our models and identifies anything that doesn’t look right. There is still a manual backstop to fill in any blanks, but Kevin is learning and improving all of the time, and we see the manual residue decreasing the more data Kevin consumes.”
Kevin has helped the company end the laborious task of going through contracts line by line manually, a process that could be complicated by human error. Raine also sees the fruits of the company’s investment in AI within hotel sourcing. Already machine learning is being utilized for the normalizing and matching of data from different sources at the beginning of the process. Over the next year, he believes that it can be widened out further across the hotel sourcing process, in identifying negotiation opportunities and used to help GBT build what is known as “choice architecture” — the way solicitation choices can be presented to clients.
“There is room for AI technology and intermediaries to make the process of choosing better,” Raine says. He also believes that AI will help crunch the increasingly large amount of data being generated when someone travels and gain a deeper understanding of traveler behavior. Hotel sourcing, in particular, is one area where this could work really well.
“There might be a thousand hotels in a city. Traditionally, we might have looked to include hotels in the program where travelers have stayed before, or based on location or brand, or those that typically end up on corporate programs. Going forward, adding in additional data sources, and utilizing AI, we will be able to understand traveler choices better, identify new patterns so that we can make more informed recommendations, and curate a more bespoke hotel program.”
Looking to the future
AI has been with us as a concept for six decades, but the signs are that it is starting to really help those in business travel build more efficient processes, present us with smarter choices, and understand traveler behavior better. Kevin may not be the spectacular creation envisaged by those New England intellectuals, but he is part of the quiet revolution improving the world of business travel from the inside out.