https://youtu.be/T7OpY6P0cE8

Bespoke is the developer of “Bebot”, the first AI-powered chatbot developed specifically for travel and emergency response. Bebot assists >10 million airport passengers per year to measure and improve the customer experience.

Supporting this experience behind-the-scenes is a team of software engineers, machine learning engineers, and machine learning researchers.

We at Bespoke leverage AI in the travel industry to allow airports to:

a) effectively communicate with travelers in multiple languages

b) improve customer service and increase sales

c) encourage a situation-specific customer dialogue ("did you mean?") to drive insights for our partners

Effectively communicate with travelers:
Bebot is easily accessible from the airport wifi splash page on any device browser and there is no additional download or registration required. Our low friction approach allows for greater user engagement and reach, where we’ve seen up to 38% passenger engagement at airports.

Specifically, Bebot acts as an info hub and emergency response tool to provide answers to questions from passengers in real-time, as well as share key updates with passengers. During the current COVID-19 pandemic for example, Bebot keeps passengers updated on information such as travel restrictions, measures the airport is taking to combat the pandemic, and changes in store hours.

Technically speaking, we use natural language understanding and machine learning to understand users' queries and automatically respond with pertinent information from our knowledge database and external data sources. In this way, travelers get instantaneous answers to their questions without the frustration of finding sparse information booths, making phone calls to call centers in foreign countries, or using up time of airport staff.

Furthermore, the user language is detected for Bebot to be able to automatically respond in the user’s preferred language. For even better accessibility, if accessed on a device with a microphone and screen reader capabilities the bot is designed to be accessed through aural methods.

Improve customer service & increase sales:

Bebot is utilized as a feedback channel to allow passengers to provide feedback to the bot. That includes data on facility services, facility cleanliness, helpfulness of staff or shops/restaurants the passenger would like to see at the airport. 

Bebot’s in-chat survey feature allows partners to survey passengers in real-time to collect specific feedback about their facilities and services. In addition to targeted surveys, we can analyze anonymized chat history with topic extraction, sentiment analysis, and other data analysis techniques, allowing us to extract business insights to assist the airport to improve their marketing strategies or facilities. Finally, Bebot is used as a tool to drive sales by pushing personalized recommendations, for example, to help passengers find their desired souvenirs at the airport.

An example of how we helped our partner Narita Airport improve services was when, through chat history topic analysis, we discovered a large volume of complaints about wifi connectivity. We decided to investigate further by sending exit surveys within the chat conversation and reported our findings to the airport. Once we shared the insights about wifi complaints, the airport updated the wifi routers in their facilities. After the changes were made, we reviewed the results of the exit survey and found that passengers satisfied with the wifi went up from 66% to 88%.

Encourage a situation-specific customer dialogue:

Our AI allows us to understand the intent of what the traveler is asking for, represented as an extensive machine learning classification problem. Furthermore, when the traveler's intent is ambiguous, we ask clarifying questions to find the right direction for the conversation. For example, when there are two airports that are close in proximity and a user asks for directions to the airport, our AI recognizes the need for the user to clarify which airport they want access information for. Technically, this case is represented by two classes of intents having similar probability of being correct. By utilizing intent disambiguation, Bebot quickly provides the correct answer.

Conclusion

The above is just a glimpse of how by using AI, and more specifically natural language understanding, machine learning, and data analytics, we are able to achieve improvements simultaneously in traveler experience as well as the operational efficiency of airports.