So far this year we’ve seen the launch of many a chatbot, but considerations of user experience and practicality are sometimes left by the wayside as chatbot excitement takes over. We spoke to Steven Gutierrez, Lead, Social Media and Content Editor at Transport for London, to hear more about TfL’s latest innovation – the TfL TravelBot. By focusing on usability and accessibility, and empowering customers to self-serve, Tfl is successfully using AI to increase its efficiency and improve customer experience.
Empowering Customers To Self-Serve For A More Efficient Service
The TravelBot’s main function is to answer simple travel enquiries. “We have a lot of customers on social media asking quite simple questions; the bot makes it easy to self-serve, and allows customers to find out the information for themselves” Gutierrez explains. This is a win-win, because it provides customers with a faster service, and frees up the time of customer service agents to answer more complex issues such as accessibility related queries or questions about contactless payment. “At the moment, if there’s disruption on the Circle line for example, there will be hundreds of tweets asking about the latest status. This is unsustainable because agents can’t reply on a one to one basis in such a short space of time, whereas the bot can help thousands of people at once if it needs to.” Allowing people to self-serve on a platform that they are already very active on was another key element of the bot’s development. “We know the majority of people in London use the Facebook platform and spend a huge amount of time in that space, and we wanted to make the information available in a really friendly way, in a space where customers are already spending a lot of their time.”
Self-service is something TfL promotes more widely across its services, and it encourages customers to use tools such as Journey Planner. Future plans for a Twitter bot that will allow bus passengers to ask queries over private messaging will also provide this self-service opportunity. “Whenever we see an opportunity to promote self-service we are working hard to increase awareness – we want our agents more focused on the problems that require greater thought process than just travel disruption”, says Gutierrez.
Breaking Barriers: A Conversational Bot
The TfL TravelBot has been designed to suit the modern audience, who prefer to use conversational language when engaging with a bot, rather than formal text patterns. “We learnt from a year of user interaction on our Twitter bots that people like to use their own language, and don’t want to use specific commands. So we incorporated a layer of AI in the TravelBot that does natural language processing. It can learn as people use colloquial terms, shortcuts and nicknames – so it’s about allowing the technology to figure that out itself, and understand the intent from keywords.” This, Gutierrez explains, is a big jump for TfL, because it means the barrier is massively reduced, particularly if English isn’t a customer’s first language. Looking to the future, TfL is also looking to make the bot easier to use in everyday life, by allowing people to add it to their group conversations on Facebook Messenger. “A group chat can have a passive bot participating in the conversation when it’s useful such as providing travel updates – this will work really well when friends are trying to meet up, or with sports or music communities.”
Responding To Customer Behaviour: Information When They Want It
TfL ensured they responded to how customers wanted their travel updates; from experiences with the Twitter bot, TfL learnt important lessons about user behaviour. “We learnt that customers really like checking for information when they want it” Gutierrez explains. “Therefore, we learnt request response functionality had a good chance of being popular.”
Although the TravelBot is currently available to use, the launch has been a deliberately quiet one. “Putting the TravelBot out into the wild has allowed us to refine it, fix bugs and really understand what our next priorities need to be; the ultimate test is to have real customers using it, which is what we’re paying attention to at the moment. We made hundreds of refinements and changes to the bot within the first few weeks.” This approach was taken in order to ensure the user experience is as good as it can be ahead of further promotion – “one of the things we were clear about was that if users get stung by a bad experience, they don’t tend to go back. So we wanted to roll the bot out gently initially, to ensure that the bulk of users we receive over the next few months when we promote it properly will have a better experience, and will be more likely to be repeat users.”
By rolling out the TravelBot slowly, and taking time to iron out problems, TfL are creating a service that is user-focused and will only improve with time. This is a tool that will help people with daily travel issues, without them having to leave a space that they are likely to already be active in. In one click, customers can move between chatting to a friend to asking the bot about a bus timetable; this ease of use is reflected in the natural language processing of the technology that allows people to use conversational language when engaging with it. All in all, this bot looks set to reduce commuter headaches and make London travel a more seamless, informative experience.
Steven Gutierrez, Lead Social Media and Content at Transport for London, will be appearing at the Festival of Marketing, 4-5 October 2017, Tobacco Dock, London: http://www.festivalofmarketing.com