May 2017, London
AI + NLP: the winning chatbot formula?
Andrew Hennigan has been finding out if artificial intelligence technology will power a new generation of travel chatbots this year
In 2016 there was a lot of talk about chatbots and artificial intelligence (AI). As CarTrawler’s Bob Healey bluntly puts it in an interview last month: “People are now sick of hearing about AI and bots, so I think they’ll actually start doing something in that space rather than just talking about it.”
Putting it a little more gently, is Swapnil Shinde, CEO and co-founder of Mezi, an AI-powered personal travel assistant app which allows users to interact using natural language. "AI-driven chatbots have made massive headway over the last year and we expect this to continue into 2017 as the technology becomes mainstream,” he says.
What many travel businesses applied chatbots to last year was the handling of sales and customer service through messaging apps like Facebook Messenger. The main aim here was to automate customer interactions.
One of the problems of early chatbots is that they were simply based on rules that defined how to respond to specific word combinations. The drawback with this is that even a simple transaction is hard to cover with a predefined set of rules because for every question you anticipate customers making, you will find many more. This is a problem that Travelaer, a tech firm that provides booking engines to airlines, faced with its early rule-based bots.
Chatbot evangelist Mike Slone, Chief Experience Officer at Travelaer, admits: “The first bots that we built in early 2016 did not have machine learning and could not handle all of the various questions that people would ask, even on a narrow topic like booking a flight. After we learned that our rules-based bot could never generate enough rules to cover all questions we began to look for an AI partner.”
In powering chatbots with AI software they can become more effective and help to unlock new revenue streams
Artificial intelligence addresses this problem of rules by creating a system that learns as it works. In short, it doesn’t need to know all the possible questions in advance. In powering chatbots with AI software they can become more effective and help to unlock new revenue streams.
Technology that delivers
The technology already exists in the form of Facebook’s natural language processing technology wit.ai and IBM’s AI engine Watson. Startups like Lola Travel, launched by a former co-founder of Kayak, have also developed their own Watson style AI engine, in the form of Harrison.
The claim is that with the improved performance of chatbots will come significant benefits such as reducing the time needed to plan and book a trip, which could be of particular benefit to business travellers.
Shinde argues that bots have the ability to understand and respond like a human to requests for travel bookings, providing recommendations and completing transactions with greater immediacy and accuracy than a human agent.
Having said that, he doesn’t believe that chatbot technology will lead to job losses, a concern raised at the recent World Economic Forum in Davos. “Automation and machine learning will not necessarily replace humans, but rather help them to do their jobs better, and in turn spur job creation as industries evolve,” he says.
Lola, which is leading the shift away from purely self-serve travel also holds the view that we are a long way human-less travel – hence the firm’s 20-strong agent team that is able to the access its “amazing and efficient tools” to deliver a truly personal experience.
Learning on the go
AI-based bots have the ability to learn and the more work they do the faster they learn. “Chatbots today are learning but it takes practice and training for the bots to understand more intent and questions, which only comes with time and use,” says Slone.
While it might be early days, the potential for disruption is there and Slone even goes as far as to say that firms should consider channeling their app budget into AI- driven bots. One of the biggest advantages is that they can outperform humans in some tasks, simply because they can process huge amounts of data to deliver the most informed choices. Importantly, they can also unlock new revenue streams.
According to Shinde, chatbots can help businesses to quickly scale from serving hundreds of customers to hundreds of thousands. “These advanced capabilities will also help unlock new partnerships across businesses and service industries,” he says.
So while today travel companies are currently focused more on customer-care chatbots, this is likely to change as the potential for upselling and cross-selling is better understood. Where they could prove to be particularly useful is in driving airline ancillary revenue.
Where they [chatbots] could prove to be particularly useful is in driving airline ancillary revenue.
Says Slone: “Once an airline customer purchases a ticket and sends their itinerary to Facebook Messenger, this opens up a smart, personal communication channel between the airline and customer. This channel can then be used to offer the customer options like pre-purchasing food, excess baggage and upgrades.”
In essence then, chatbots are really nothing without AI and NLP. Indeed the future of true conversational commerce will be dependent on these technologies progressing at a rate that allows users of bots to see improvement and smarter results. “Without NLP and AI bots are just a clunky interface that will never be adopted by the mainstream”, says Slone.
Join us at EyeforTravel Europe to hear more about how conversational commerce is evolving in 2017