How I Learned to Love Alexa

A few weeks ago we reached a milestone in the development of our first Amazon Alexa integration. I recorded a video demo to show off what we’ve accomplished which is embedded at the end. Here’s the story of how I went from writing about Alexa, to writing an integration with Alexa, to building a vision around Alexa, all while growing a magnificent beard.

It all started when Sam and I got to talking about my first article about Amazon Echo. In this article I write about a fictional smart agent, named Charlotte. Charlotte is loosely based on Amazon Echo but we were just getting to know Echo at that time and I wanted to write something more generic. Sam asked if ImproveCX really could build a Charlotte. And I said: “yeah, uh, I think so… sure we could build that. Ha, ha, we just need a million dollars so we can hire a team and…”.

So he said (British Accent): “Let’s do that then.”.

“… What’s that?” I said.

“Build a Charlotte.” he said.

“Do I get a million dollars?” I asked, sheepishly.

“No.” he said.

Uh oh. I had a good understanding of the Alexa platform’s potential but wasn’t sure yet how to realize that potential in alignment with my vision; Still, I was well groomed, looking good and feeling good, ready to take on a new challenge!

In my original story Charlotte combines a few ideas. It brings together a voice UI, a knowledge base, analytics tool, and web presentation. These combine to create an intelligent agent that can understand voice commands, determine the context and intent of those requests, and then execute an appropriate action like showing a report in some presentation software.

Our perspective on Alexa is a smart agent for business applications; Where much of this space has been focused on consumers, the idea in the story was to create a tool that could serve in a different capacity. We see a lot of potential for purpose built, voice driven, smart agents to tackle a variety of business automation challenges in new ways.

So I start reading Alexa technical documentation. I start spending more time with my Echo. I’m talking to myself All. The. Time. Really I’m talking to Alexa, but that sounds like the same thing from outside my office door. I work from home, my wife starts looking at me funny. I stop shaving and visiting the barber so I can spend more time with Alexa. My wife says she likes the beard.

Our first challenge was to demonstrate both the potential and demand. We needed to build a prototype of something but first we needed a target we could hit. Which piece of Charlotte should we build first?

We decided to start with our partners at Wizdee;  Wizdee aggregates your data sources in a business intelligence tool that lets you use natural language to ask questions. You get back insights in the form of answers, charts, graphs, and tables then drill into that data for even more insights.

If we get it right, it should look unbelievably cool, talking to your data and getting back answers! We set our sights on getting the basic interactions working first: ask a question, speak the response, show the chart in the Alexa app.

In order to get this moving I had to design an interaction model and test it, re-figure it, test it again, over and over, to get it just right. Cue the 80s style movie montage of research and prototyping. Picture me from odd angles typing fast, pacing the floor, furiously drawing diagrams on whiteboards, and asking Alexa the same questions over and over again. A little while later I had a decent working prototype which let you Ask Wizdee a question and Alexa would speak the answer, provided the answer that came back was a single data point. Anything more complex would return an error. Not very useful, but it proved we could make it work.

By now my hair has grown out and likes to stand up at odd angles. My beard does too. I’ve been talking to myself Alexa for more than a month. I filed a feature request with Amazon to add “Wilson” to the list of wake words. My wife gently suggests that I might need a trim.

I polish things up a bit and we tell Wizdee we have charts working, and they say “Great work! We have this fantastic idea, instead of your test data, lets hook this thing up to a real live integration and ask it useful stuff, about sales pipelines and widgets, and trade show revenues! (mumble, mumble, Dreamforce, mumble, mumble).”

So… I did that… they gave us the configuration and a bunch of questions to try out. Which I did, and which promptly showcased every bug I’d written over the last whatever weeks since I’d last had a shave. Ouch! I cued up another montage, and fixed a slew of bugs. I spent a couple of days making sure we could get the data processed from the questions and that our interaction model was solid and that the responses sounded natural.

My wife now complains that I grumble about grandma’s house in my sleep. She says the howling at night is going to wake up the kids. She firmly recommends that I get a haircut and perhaps I could make use of the electric trimmer? I start knocking randomly on neighbors doors to ask folks if they’ve met Alexa. Nobody ever answers…

Things were finally working beautifully within certain constraints and we had a solid prototype. We all felt it was good enough for a short demo! I broke out my tripod and camera, I put on a clean shirt, combed my face, and shot a video to show off what we’d accomplished. In this video you can see my glorious hair, and my whiskery beard, and hear me ask questions.Alexa answers each one flawlessly, displaying the chart data on a screen I strategically placed behind the beautiful black cylinder, Echo, my preciousss.

Working this way, you will be able to quickly find new insights in your data through natural interactions without complex report writing. Just ask questions and listen to the answers. I have to admit, I was skeptical. Before we connected up to some real data I was having some doubt this was truly useful. Once I got to put it through its paces with real data and real questions, it really started to feel like we’d hit on something. Hearing the great reactions to the video really cements our belief in the platform.

Sadly, Amazon rejected my feature request and my wife is making me go to the barber this week… but… I get to keep the beard! 😉

Here’s the video I made, Sam edited it for brevity: