How we deployed Edward the SMS chatbot on Amazon Echo in 30 Minutes


A few weeks ago, I introduced you to Edward. Weeks later, we are still being approached from analysts, customers, partners, other vendors, who all want to hear more about this exciting project. It is still very early, but seeing the first guests interact with it and the things it gets right and the mistakes it makes is fascinating, and very insightful to all of us involved.

Edward is Radisson Blu Edwardian’s new virtual host – it greets guests via SMS and handles questions or needs as diverse as “what cuisine does your restaurant serve”, “send me some ice please”, or “please don’t clean my room today”. (One of the early findings was the guests of the Radisson Blu Edwardian tend to send full sentences. They would, e.g., rather send “Can I make a reservation in your restaurant” vs. “restaurant reservation”.) Edward is fully automated with human backup when needed, understands natural language, and responds within seconds. A whole new level of service excellence in the hotel!

While the system is still being tuned to become better and better, we are also not stopping experimenting and innovating with it. Edward runs on Aspect CXP, a platform that lets us build self-service experiences once and then deploy it on any channel. And by any channel we mean any channel. CXP’s unique design-once-deploy-anywhere approach and its adapter framework for adding new channels lets us take an application as is and deploy it on a different channel. As we had been prototyping an adapter to the Amazon Echo, we gave it a try a couple of days ago. How long did it take us? About 30 minutes, give or take.

And here is how it performs out-of-the-box:

Again, it took us only 30 minutes to point the existing application, which was built for the SMS channel, to the Amazon Echo. This is all due to the architecture of Aspect CXP, that lets us deploy dialog applications on a variety of channels, and add new ones easily. Differences among the channels, such as prompting the SMS user “send X” vs the voice user “say X”, can be accommodated easily using CXP’s Layer concept.

Now, what can we build for you? The sky is the limit.