In late 2015, our Chief Customer Officer, Joe Gagnon, and I, met with the IT Director, and the COO of Radisson Blu Edwardian in London. A long-term customer of Aspect’s, we were there to share what we have been so diligently working on over the past year: our vision for re-imagining customer service that would combine the best of all forms of consumer interaction types, and the best of what we and the industry have been able to develop in next generation CX technology. This included:
- The business value of a blend of personal touch and automation
- The response times and accessibility of self-service
- The proven methodologies of Interactive Voice Response
- The consumer appeal of texting/messaging as a communication channel
- The ubiquity of SMS across the world
- The benefits of Natural Language Understanding for free-form dialog
- The value of CRM to show the guest we know them
- The human touch through live service integration where needed
In essence, the vision for re-imagining customer service has at its core how to use Interactive Text Response (ITR), or what is also known as “bots” to provide the ability to let customers self-serve on text channels at blazing-fast speeds with a User Interface that resembles that of a natural conversation with a person.
Needless to say, it didn’t require much convincing that this approach would provide an opportunity to “surprise and delight” in their prestigious hotel chain. With over 2 billion people using texting and messaging solutions extensively today, offering service over these channels just makes sense. Furthermore, it has the promise of saving cost through smart automation while providing a state-of-the-art customer experience, or in the words of the COO: “I want this by Monday.”
From the outset the solution was designed to let guests send SMS text messages to be served throughout their stay, and even before check-in. We were intrigued when we found out how many different questions a guest could have during a hotel stay. 153 to be precise. From answering general questions such as “When is breakfast”, “What cuisine does your restaurant offer”, or “What time do I need to check out?”, to room service needs such as “Can you send more towels?”, or “I’d like a paper delivered to my room” – the system would need to understand plain English inquiries, and even distinguish between “a paper” (newspaper) and “some paper” (stationery).
So off we went and built the team. We brought in experts from various areas of our organization:
- Conversational UI Design
- We got help from our NLU engineers as well as computational linguists to design the front end and develop the Natural Language Understanding scripts
- Aspect CXP Pro app development
- Our app development team took on the job of building out the ITR and IVR flows on Aspect CXP, our self-service application lifecycle management suite
- Proactive Engagement
- The solution integrates with the hotel’s reservation system to kick off an outbound SMS message that welcomes the guest after an opt-in from them
- Telephony, IVR, Call Control, System Integration
- The solution connects guests to staff members when needed (“call me!”)
- We integrated with the hotel’s home-grown staff management system to accomplish this
- Reporting & Analytics
- Both the customer and Aspect wanted data to be able to tell us: will the solution work as planned? How is it being used? Where can it be improved? How often are humans needed?
As we designed the solution, we also came up with a universal template for self-service applications in general:
The generalized model allows us to deal with both the “happy” path and the “reality” of customer interaction that takes the form of complaints (“the neighbor is loud”, “your pillows are too hard”), praise (“love your beds, so comfy!”, “great job by housekeeping”), and even humor: it is probably no secret that people LOVE to play with bots. So we went ahead and built our own corpus (response collection) of humorous, witty remarks. However, we were careful to design the solution such that the humor module would only kick in if all other possible interpretations of an utterance could be ruled out – we carefully want to avoid answering a complaint with a snarky comment after all.
Furthermore, we designed the solution deliberately without a machine-learning component. Machine-learning approaches to Natural Language Understanding work well for large or better yet unrestricted domains, but if the domain is restricted, like in our case to hotel service, a rules-based approach is most effective. Also, recent experience across the market has showed that building a completely unsupervised learning bot can backfire quickly.
An important part of the design of the solution was to give the customer full control over what the system would reply with at any time. So we made use of Aspect CXP’s Business User Interface (BUI). Through this easy-to-use Web portal, Radisson Blu Edwardian staff can change the system’s behavior at runtime, without requiring Aspect’s or even their IT staff’s help.
After weeks of development, QA, and close cooperation with our early adopter customer, we are proud and excited to release our solution to the public. It will be launched as a production pilot in the Sussex location of Radisson Blu Edwardian. To get a glimpse of how the experience will be, take a look at this screenshot:
Or, better yet: go check yourself in to the beautiful Sussex Radisson Blu Edwardian in one to two weeks when we are live and experience it for yourself.
We will report on the outcome and guest feedback as soon as we can, and we will also publish more of our experience around the design and development of the solution over the coming weeks and months. So stay tuned! Meanwhile, we are moving over to working on the “World’s Most Convenient Car Manual” for a large car manufacturer here in the US, as well as various other messaging-based self-service solutions. Curious? Well, talk to us about how we can help YOUR business enter this fascinating space using the industry’s most advanced enterprise bot platform!
Latest posts by Tobias Goebel (see all)
- Do Chatbots Dream of Electric Sheep? - February 21, 2017
- How patient do we need to be with Machine Learning? - January 20, 2017
- 2016 – The Year of the Chatbot - December 27, 2016