Akemi Tsunagawa is the founder & CEO of Bespoke, one of the fastest growing chatbot startups in the travel industry. Prior to Bespoke, Akemi worked at Fidelity Investments, Deloitte, and Macquarie Capital after graduating from UCLA.
Bespoke was founded in 2015 to offer authentic travel experiences beyond guidebooks. Its AI concierge, Bebot, is the most used multilingual chatbot in the hospitality space to enhance guest experiences. Bebot is currently servicing wide range of clients from International Airport to global hotel brands.
Tell us about yourself and what prompted you to start a chatbot company?
I founded “Bespoke” in 2015 to make the product that I’ve always wanted for myself. As a traveler, I’ve always enjoyed discovering little known places rather than major attractions featured in guidebooks. Those are the kind of places only your local friends or family would tell you about, and those are also the kind of places you’d always remember when you tell your travel stories to others. Almost every time I traveled to a new place, I’d try really hard to find those cool “local joints” but didn’t have much luck without knowing locals. It took me about 10 years, going from thinking about “why isn’t there a good (and safe) tool to make that possible?” to “what if I make one?”.
Today, Bebot is a platform that empowers travelers to enjoy the same sort of experiences as if they had people showing them around. That is how I started the company about 3 years ago, which has been the best decision I’ve made in my life so far.
Introduce us to Bebot and what it can do.
Our chatbot “Bebot” virtually guides you in and out of the hotel/airport/train station to help find best local experiences you wouldn’t find elsewhere. Bebot is now the most used multilingual chatbot in the hospitality space that performs concierge tasks on behalf of front desks/info counters. Available at Narita International Airport, Sofitel Hotel, JR Tokyo Station, Hotel New Otani and many other places!
Why did you choose to employ Bebot in the travel and hospitality industry?
I followed my passion rather than opportunities first; however, we did get lucky with getting the timing right eventually. We were focused on consumer facing services rather than business to consumer services for the first 6 months. As we conducted many user interviews, we quickly learned that many of our users were frustrated that locals didn’t understand foreign languages to take reservations at those local restaurants or even to answer very simple questions from travelers. At the same time; staff at hotels, train stations, etc were also experiencing difficulties saying they couldn’t do their job due to the language barrier. Those user interviews helped us connect the dots. We shifted our business from 100% consumer facing to B2B2C. That way, we can help both hosts and guests. Now, Bebot is available through our partner hotels, airports and train stations to ensure that guests have great experiences.
In which industries do you see maximum opportunities for a chatbot of this kind?
Hospitality! Majority of questions that people in the hospitality space receive everyday are the same or similar. And answers to those questions tend to stay the same, which can be automated. Being able to automate these questions allows staff at hospitality facilities to focus on guest requests that provide higher satisfaction.
What security measures does Bespoke take when dealing with a company’s data?
We don’t collect personal information. In addition, we do not share any of our user’s information to third parties. Our services are encrypted, and access to our servers is through a VPN that utilizes unique electronic certificates. Even within our company, only certain team members have access to raw user data.
Are there unique challenges to adapting a chatbot to function in different languages?
Yes. We’ve found that users of different languages also have different requests, so when we built our chatbot from English into Chinese, it’s not just purely translations of the same texts, but we also need to plan for different conversational flows.
Tell us about some of your business partners and how they contribute.
We have strong relationship with JR East group (largest railway company in Japan) and Narita International Airport (Japan’s main airport) since the early days. JR East has been helping us with customer acquisitions. Narita International Airport helps us better train our chatbot by sharing passenger insights. Both of them encourage us to try new ideas that would benefit both parties.
How does Bebot’s interaction change with each guest? What factors does it take into consideration when providing a service or search result?
Interaction changes significantly depending on a guest. The biggest factor we see there is “hotel class”. The most obvious example I can give is the restaurant booking feature. Guests from 5 Star Hotels tend to book restaurants above a certain price range while guests from 3 Star Hotel would prefer places that don’t require booking. We customize Bebot to each client, so the “hotel class” part is already built in. In addition, Bebot’s recommendations also change depending on users’ previous requests.
It’s a bit similar to what you see when you’re shopping at Amazon. Just not that advanced yet.
How does Bebot offer support at airports? How does a traveler interact with the bot?
Bebot for airports helps guide passengers around the airport. They can ask about facilities such as restaurants, shops, gates, flight info, and directions to get into the city, etc. Passengers at Narita Airport access Bebot through the free WiFi available at the airport. Once a user connects to the WiFi, the final landing page is our chat. No additional download is required.
Which enterprise softwares does Bespoke integrate with to make adoption easier?
Bebot used to be available on Facebook and WeChat, but we’ve discontinued that feature. To gain more flexibility, we decided to develop everything in-house from scratch, including all NLP engines, chat framework, database, taxonomies, etc. We currently don’t use any third party interfaces, meaning our chatbot runs on any mobile browser. Users or clients can start chatting without downloading anything as soon as they connect to WiFi at client’s properties (hotels, airports and train station).
What are the challenges in ensuring that the technology is able to understand context, sentiment and tone?
One of the biggest challenges we’ve seen is Bebot trying to differentiate between serious user requests and more playful ones. It is very difficult to train Bebot to be able to understand tone because there are always certain words which can trigger certain responses. To try and improve Bebot’s understanding of tone, we’ve extended Bebot’s small-talk capabilities which have helped a lot in being able to manage playful messages from users.
What’s next on the agenda for Bespoke?
Next on the agenda is to have Bebot available in more countries, in more languages, and to make even more travelers happy!
Thank you Akemi! That was fun and hope to see you back on AIthority soon.