Alexa Shopping With Skills

Alexa Shopping with Skills

May 19, 2016

In my last post on the SPI blog, I wrote about how a small team here at SPI successfully built an Amazon Echo using Amazon provided code and a Raspberry Pi computer. I also wrote that we had a plan for round 2 of experimenting with our DIY Echo, building an Alexa Skill, or a voice enabled application for use with the Echo.

As we played with our Echo, we knew we needed to build an application that involved a shopping experience given SPI’s focus on retailers. And that’s what we did. The requirements for the application were:

Must involve an inquiry about a product or service
Must allow a “retailer” to post items that can be read and presented by our application
Must allow a shopper to “interact” with the voice service, accepting some items and not others
Must be “real”. The application should be one that could be posted to the Amazon Skill store

We did it! We built an application that met all of our requirements. Will it attract millennials in droves? It won’t in its current form. But, with this groundwork laid, we see lots of opportunities in the voice space creating technologies that allow retailers to quickly delivery voice applications in all the places people dwell. Phones, Echos, TV remotes.

Our application is called the “Deal of the Day” Alexa Skill. Once you request the app from an Echo, you can then make your request. “Alexa, get Deal of the Day”. She then replies “Would you like to hear the deal of the day?”. From there you can either hear a list of deals or you can request specific deals from specific retailers. Each time, our application is requesting the appropriate URL containing the deal of the day at the retailer requested.

OK, was this really hard? It really was a fun and reasonable effort. It was though far simpler than building a mobile app given no concern had to be given to a visual user interface. The user interface for this app is a persona, Alexa, who has already been given a gender, a tone, appropriate intonation and a number of actions and reactions. Further, that persona is driven by structured JavaScript running on our own Amazon Web Services (AWS) instance. Sure, we had to write the JavaScript controlling the back and forth discussion between a user and Alexa. But, what we missed or ignored was already defined by Amazon, ensuring we had some real nice training wheels on our new racing bike.

We did it though. And we tested it on our Raspberry Pi Echo and on my Amazon one. Are voice personas going to take our retail orders in the future? I bet they will. I just hope they don’t answer back with a “I’m sorry Steve, I cannot do that.”