AI could make personalisation less creepy, not more

Everybody loves the personal touch, up to a point. We like it when the guy at the lunch place remembers our order or when a saleswoman knows what shoes we’ll like just by looking at us. No one likes being totally anonymous and we all appreciate being treated like we’re an actual individual, not just a faceless member of a giant group.

That is, we like it offline. Online, we’re constantly on guard for personalisation that feels, to put it frankly, creepy. Users enjoy getting personal emails—according to eMarketer, 80 percent find emails with personally recommended products helpful—but can be turned off by display ads that follow them around the internet as they read the news or check Facebook.

Users like it when they feel like you’re helping them, not when you know where they live or publicise their purchases. And they’re especially uneasy when their online personalisation overlaps to the brick-and-mortar world. 

Essentially, it’s important to find that sweet spot where personalisation doesn’t feel too personal. This is especially true for retailers.

So how can they and other sites provide the benefits of personalisation without turning users off? What’s the sweet spot?

First, we need to look at the sorts of personalisation users enjoy. We mentioned that users like tailored emails where they receive recommendations based on their buying patterns, so let’s use that as a guide.

This personalisation works because it adds value to their experience. It helps them find what they want. It’s streamlined and doesn’t follow them around as the browse elsewhere. It’s personal, but the user itself controls the action. If that email converts, it’s because the user themselves opened it and decided they wanted to buy. It’s a reminder as opposed to a salesman following them around a store, asking, "Do you want this now? How about now?"

But there’s an issue. This type of personalisation requires a ton of data. Recommendations are based on buying patterns across large swaths of purchases by "users like you." And it needs the user to be logged in for it to work in the first place.

Which leads us to artificial intelligence. See, with AI, we can actually avoid all those roadblocks. We can provide increasingly personal experiences to users without intruding. In many cases, they won’t even know their experience is being personalised at all.

Here’s an example of how that might work: 

Say a user hits a retail site for the first time. They’re not logged in, so there’s no buying history to lean on for personalisation. Now, say they’re curious about a line of brand new products, so there’s no product history to lean on either. Normally, that would mean a generic experience. But with AI, it’s different. Once the user clicks the first image, an AI can actually analyse the image itself, looking at hundreds of vectors, and actually recommend products based on visual similarity, not just keywords and tagging and browsing history.

The next click is where things get personal. Because once a user keeps browsing and clicks on a suggested product, the AI starts to understand what they’re really after. Is it the shape of the product? The colour? Some subtle, hard to describe, characteristic? Some combination of these or other attributes? AIs can actually pick up the patterns here, effectively intuiting a user’s sense of style in the moment. All without them being logged in. 

That’s personalisation that isn’t too personal. 

There are other examples too. AI can help optimise websites, personalising them for users based on audiences the AI itself has uncovered, providing messages and design that resonate with specific users. That means websites that evolve and change, based on both user demographics and user behaviour in the moment.

The best part about all this is that these kind of AI-enabled experiences are great for users and businesses. Users see better products and smarter websites, getting them to the things they want quickly. Businesses get to spend less time doing archaic web practices like product-tagging or slow A/B testing. And businesses receive real insights about their customers they can use to try new messaging and experiences, forecast upcoming trends, and, really, do whatever they’d like with that information.

Just keep it from being creepy. Nobody likes that. 

Gurmeet Lamba is chief operating officer at Sentient Technologies.