Skip to Main Content

Jan 30, 2017 | 5 minute read

eCommerce chatbots are here

written by Clegg

  We chatted to Pavel Veller, who has taken buying ability to the next level with his Elastic Path-powered commerce chatbot.

We’re in an age where convenience is key. Retailers strive to reach customers in as many ways and places as possible making it easier and easier to shop from wherever you are. 2016 introduced the likes of Amazon Go, Drone deliveries, Amazon Echo, and Google Home, which leaves the question, what’s next? We chatted to Pavel Veller, a CTO of Digital Engagement Practice, web developer and technologist, who has taken buying ability to the next level with his Elastic Path-powered commerce chatbot.

You can find out more about the chatbot in the video below, followed by a Q & A.

Why did you choose to explore the relationship between chatbots and commerce? And why was Elastic Path the chosen eCommerce API?

From early 2016, I was interested in machine learning generally and started researching the topic. Whilst looking at this, ML APIs were also on the rise with Watson, Google and Microsoft all publishing theirs and so I looked into how they can work together. During the researching, I applied to speak at a conference in Boston about building smart apps with cognitive APIs. As part of my application, I decided I would suggest talking about commerce chatbots. Even though I hadn’t actually built one at this stage, I knew the technology was available so when I got accepted to speak, I just had to build it.

When I was first accepted for the talk, I hadn’t heard about Moltin (acquired by Elastic Path) but I knew I needed a flexible, well-documented API with commerce functionality and had an idea about some on the market but the ones I had heard of didn’t suit my needs. When researching what was out there, I came across Moltin’s website and it fitted the purpose of the chatbot perfectly. It looked simple to use and provided me with all the functionality I needed.

How important do you think AI is to commerce experiences?

Where we are with chatbots right now is exactly where we were with mobile apps back in 2008. They were something new, still being tested and used for small experiments as no one knew whether they would take off. Fast-forward to now and 82% of consumers’ time browsing mobile media is through these apps. Mobile apps have now become a channel in their own right and a medium people think about on their own.

The majority of our time on mobile is spent on messaging apps, so it makes sense to build shopping experiences where people already spend their time, as, why leave something that people are already familiar with.

In your work, you’ve called chatbots ‘API Orchestrators’, can you explain this?

Yes, if you think about it, the bot itself is just a web app, an HTTP endpoint basically. In order for it to be useful, you have to firstly, connect it to an actual messaging platform (I used Skype for the purpose of this). The next thing is to understand what the users are saying and that’s where NLU (natural language understanding) comes in. The last thing is to perform the intent expressed by the user. And how else can a bot do it if not by orchestrating other APIs like search, commerce, booking, calendar, and more.

Do you think there’s a danger for chatbots as they become more sophisticated?

I don’t think so. Technology has allowed us to conquer many things including image recognition and speech detection and chatbots are next to be improved.

Right now a chatbot can do as much as any other web app. You don’t think there’s a danger in web apps becoming more sophisticated, right? Messaging platforms are guarding user profiles details. All my bot on skype can see, for example, is the name of your handle. I get a unique ID of a user and that’s about it. It may change over time but I am sure it will be small steps. The AI part is more interesting and clearly has a lot of potential, but right now we’re distilling users’ utterances to intents and entities and it will take some time for the technology to mature beyond that.

What do you think commerce will look like in 5 years time?

Although I do see the world becoming more automated, I don’t think the future can be predicted so far in advance these days.

Humans have a tendency to have a linear thought process and think they can predict what will happen in the future, based on what they’ve seen happen in the past and base everything on how long things have taken before now. As the world advances, I don’t think there’s any way of predicting what will happen in 5 or 10 years time. Somebody said that we, humans, tend to overestimate what will happen a year from now and underestimate the progress we will make in ten years.

Forrester makes an interesting observation as they cover the rise of the API-first headless digital platforms both in commerce and content. They see a new group of technologists driving the evolution of these products and they call them “millennial architects”. Who knows what we can do in the next five years :)

Having said that, I would see the idea that companies have of ‘we know who you are without saying who you are’ expanding on and improved in the near future to enhance those consumer experiences.

I would also expect Alexa to become more skillful and maybe include visualization (Alexa plus Hololens anyone?) and not just voice. At this point, Pavel’s own Alexa joins the conversation “I do not understand the question you’re asking”. I think we all agree, Alexa needs some improvement!

If you’d like to learn more about Pavel’s experience creating the eCommerce chatbot check it out on Github and follow@pveller on Twitter to find out what he’s up to next. You can also catch him at Syntax and probably a few more conferences in 2017.


Stay on top of industry trends with the latest blogs and articles from our eCommerce experts.