Ifs and Bots: Demystifying artificial intelligence and its applications in travel
In recent years, artificial intelligence (AI) has become one of the most talked about trends in corporate travel. But while travel managers are showing increased interest and curiosity in AI, there is still much confusion about what the term actually means and how it can be applied to business travel.
In a survey of travel buyers conducted by Business Travel News (BTN) in 2018, only 16% of respondents said they had “good” or “excellent” knowledge about AI. The majority admitted to having limited awareness about the subject. The same survey also found minimal adoption of AI solutions in corporate travel programs thus far.
What is AI?
“Artificial Intelligence” is a marketing term rather than a single technology suite, according to Dr. Eric Tyree, Chief Data Scientist at CWT. “Broadly speaking, AI is computers doing things that people thought only humans could do, and so that definition evolves over time as computing capabilities become more advanced.”
When English computer pioneer Alan Turing developed his test for AI, he was looking for a machine that behaved in a way that was indistinguishable from a human. The infamous Turing test, which is seen as a hallmark of AI, investigates the ability of a machine to execute tasks such that a person cannot tell if the task is being completed by a human or a machine.
There are various technologies that are talked about as “AI”, with differing methodologies and levels of sophistication used to execute tasks.
Towards the less sophisticated end of the spectrum there is “robotic process automation” (RPA), which uses technology to automate simple tasks. Because processes repeated by a machine are faster and more accurate, RPA can be used to automate tedious and repetitive tasks, allowing people to be more productive. Whether this constitutes ‘true’ artificial intelligence on the part of a machine is a philosophical question, but RPA is undoubtedly a powerful task automation technology.
Then you have “machine learning”, which uses examples to learn the underlying patterns and drivers in data or a task. It improves processes by referencing previous or example interactions. When there is a decision that the technology needs to make, it makes it based on the patterns it has seen and its ability to extrapolate from the patterns to new ones it experiences. For example, the past ten times you called a travel consultant, you booked with airline X. On your eleventh call, even though you do not have this specific information in your profile, machine learning will infer you will most likely be booking airline X. However, if you travel to a new destination not served by airline X, if the training data is rich enough, the machine learning algorithm may correctly infer your preference for airline Z which is similar or is an alliance member with airline X.
More sophisticated machine learning techniques can be even be overlaid on RPA to ‘teach’ machines the more subtle or complex elements of a task, rather than explicitly programming it.
But while RPA and machine learning do enable computers to display some human-like characteristics, there is of course much more to AI.
Really well-implemented AI is the ability for a computer to apply human-like intelligence to a task – or series of tasks – to ensure the best outcome. This is often accomplished by combining process automation, machine learning, expert logic and other techniques to solve problems and complete tasks that previously required human intervention.
The aim here is to use a variety of technologies to give the system the ability to learn and adjust its processes to improve outcomes over time. This enables AI to redefine and re-align processes by developing an understanding of the business at hand. It is able to apply this understanding to a particular process to make decisions, even if it has not undertaken this specific process before. Consider facial recognition as an example. When beginning to recognize faces, an AI-enabled engine might have a low success rate at first. However, with larger, richer data, explicitly logical enhancements and other modifications (both human and computer implemented), it will become more and more accurate, quick and sophisticated in recognizing faces – like learning to ignore glasses, beards and makeup to more accurately recognize the true underlying facial characteristics of people.
How is it being used in travel?
In the travel context, a wide range of AI techniques and technologies are being leveraged to enhance or automate traditionally human-executed tasks, to the point that we are unaware whether it is a human or computer conducting it. Travel AI is starting to pass the Turing test.
“Most AI today is very observational in that it looks at patterns and either tries to see those patterns in data it hasn’t seen before or tries to extrapolate from what it has seen in data in the past,” Dr. Tyree explains. “However, this is changing rapidly with the growing sophistication of AI applications in travel, where ensembles of technologies are being used to create human-like automation of tasks. Coupled with the growing power, increasing speed and decreasing cost of computing, the amount of data available is also growing exponentially, leading to the wider application of AI.”
There are myriad ways that AI could potentially be applied to change the way business travel is viewed, managed and experienced. Here are some of the exciting applications we’re seeing today, and a few that are around the corner:
Facial recognition to speed up airport check-ins: Several airlines and airports around the world are testing and implementing facial recognition software that can verify travelers with a “quick photo capture”, allowing passengers to board using a biometric self-boarding gate. This is expected to improve safety and security, while creating a faster and more efficient check-in and boarding process.
Robot hotel concierges creating a better guest experience: A number of hotels have been experimenting with AI-powered robots to manage guest services. A robot concierge named Connie, developed by IBM, has been deployed at the Hilton McLean in Virginia in the United States. It has been introduced in a pilot program designed to help guests figure out what to visit, where to dine, and how to find things they need. The idea is that it will reduce the need for guests to wait in line to ask a question, while freeing up employees’ time to focus on other tasks.
More relevant booking search results: Just as Amazon and Netflix suggest new products and movies based on your preferences and past selections, AI is being used to show business travelers more relevant search results – while ensuring they remain within their organization’s travel policy, of course. This helps reduce the time they spend searching and booking, allowing them to focus on their work and be more productive.
Quicker support with chatbots and messaging: If you’ve ever opened a chat box when contacting your bank or talking to a retailer online, you’ve likely experienced AI-enabled messaging. This is being applied to business travel too.
CWT is rolling out a hybrid messaging model, which has an AI-powered chatbot supported by an experienced human agent. Travelers can ask for assistance by sending text messages via the mobile app, where CWT’s hybrid travel counselor, Reece, responds immediately. Reece answers queries quickly because she already knows an employee’s name and travel information from the app. More complicated questions are seamlessly transitioned to a live travel counselor.
Easier and more actionable data and reporting: There is a lot of data in corporate travel. The sheer volume and complexity makes it hard to use effectively, especially if you’re not a data scientist. AI can help travel managers, business unit leaders and other stakeholders who may not be data experts, make sense of all this information. For example, CWT has a platform that makes searching data questions easier. You type in what you’re looking for into the search bar – similar to if you were using a search engine like Google – and the tool quickly visualizes the data and builds the reports that you need. It also has an underlying AI engine that improves its search capabilities, getting smarter and more personalized through use.
Text message alerts for missing hotel bookings: Systems can be designed to automatically identify trips with no hotel booked and proactively offer travelers program-compliant hotel options, even with last-minute bookings. This helps ensure they book in-channel and their accommodation data gets captured, allowing for them to be easily located in the event of an incident or emergency.
Predicting trip disruptions: Various solutions which use predictive analytics to identify delay patterns across flights to forecast disruptions are being developed. They factor in things like the historical on-time performance of an airline, seasonality, flight timings, flight paths and air traffic, congestion at various airports, and weather forecasts. With this information they can predict the probability of a delay occurring, as well as the length of the delay.
Enabling traveler wellness – CWT is using machine learning to analyze travel patterns and other data to provide analysis and feedback on traveler wellness and productivity. The aim to help travelers plan better and be more effective when travelling, and to help travel managers ensure their programs have the appropriate processes and feedback loops to ensure travelers are healthy and productive.
Are we ready for a machine takeover?
While experts agree that AI will deliver cost savings, a streamlined booking process, an improved travel experience on the road, enhanced duty of care capabilities, and many other benefits, it’s not going to completely replace travel counselors or travel managers any time soon.
“Business travel is riddled with exceptions and complexity and it is surprisingly hard to automate a lot of functions,” says Dr. Tyree.
Andrew Jordan, CWT’s Chief Technology Officer, believes people think AI can do a lot more than it really can. “At the moment, AI mostly consists of pattern matching,” he notes. “Computers can be very helpful in supporting humans in some tasks; but we can’t simply put a robot in a call center to replace humans.”
To what extent and how quickly these technologies will make their way into corporate travel programs will depend on several factors. Amongst these are the appetite of travel programs to try something new, concerns around data security, as well as an organization’s culture, demographics and booking patterns. For instance, an organization whose travelers expect very high-touch service or book a lot of complex itineraries might see fewer opportunities to deploy AI-enabled solutions than a company with a self-servicing culture and predominantly point-to-point trips.
“What travelers are looking for is ease and comfort in the travel experience,” says Anikesh Patel, CWT’s Director of Customer Management for India. “There is an expectation of personalization and intuitiveness – and while organizations are happy to see technology used to reduce costs, they are not prepared to compromise on the travel experience, whether it is supported by travel counselors or booking tools.”