What automation means for AI in online shopping

More companies are asking us about automation. In this post, I explain what automation really means when we build AI for online shopping, using a simple example of one shopper and one helpful agent.
Automation isn’t a word I use much. I usually say personalization or optimization. But when I thought about what automation means in terms of AI, it turned out to describe most of what we do.
One shopper, one great agent
Picture an online shopping experience. A shopper, let’s call her Sarah, has a question. She clicks the little chat window in the lower right of the screen and asks it.
Tom, a customer agent, answers. He looks at what she has been browsing for the last few weeks and says, “Based on what you’ve been looking at, I think you’ll like this,” and sends her a link. Sarah likes it, but she asks whether it comes in blue. So Tom does more research. He looks through everything in stock, checks it against Sarah’s browsing history, and sends her a link to a blue item she will probably love.
That’s a wonderful experience. It’s also impossible to give every shopper. If you want to grow your business, you can’t hire a Tom for each customer. It would never scale.
Doing what Tom does, for every shopper
So you look for what can be automated. A system that does what Tom did, looking through someone’s browsing history and recommending the right item, is worth a lot. Many sites do this without a chat window at all. The recommendations simply appear on the page.
That’s where we spend much of our time: building that kind of personalization with AI. A product recommender, or a site that shows each visitor the right experience, both fit under the umbrella of automation. The idea is to take something a person can do very well one-on-one and extend it to tens of thousands, or millions, of people at the same time. Knowing your customers well is the first step. For a Fortune 250 consumer goods company, we run the customer and marketing analytics platform for seven brands.
If you want to see how a recommender earns its place on a page, I wrote a guide to recommendation systems for business.
Why leadership should own it
We see more of these projects run by senior executives, and I think that’s a great fit. The people who see the whole company are best placed to judge which processes are worth scaling beyond what people can do by hand, and where automation pays off.
I’ve always said that AI is there to make products better more than it’s a technology in its own right. That’s why I like to see the AI team report to a product leader or the chief executive. I make that case in why AI teams should report to product leaders.
If you want to talk about a process in your company that could be automated, the first call with one of our senior engineers is free and takes 30 minutes.