One AI application techies haven’t been talking about as much yet: How to make shopping, and selling, smarter than it is today.
As a former Google engineer who worked (among other projects) on creating detailed profiles of Gmail users—partly to better serve them through AI—I see many of the same opportunities to analyze and learn from online and offline shopping patterns, plus reams of other data, to offer retailers and consumers better experiences.
A retailer backed by AI algorithms could function as a neural network thousands of times more powerful than a merchandiser relying on more rudimentary, back-office data to stock and market products. This network would understand the tens of millions of SKUs, or specific products, in a clothing or toy company’s virtual catalog, for instance, and inherently know which ones have been successful in the past, and which ones will sell best in the future. It could predict which specific consumers, where, would buy certain products: The parent in Montana might want an Elsa doll for his toddler in December, while board shorts might be selling briskly in certain regions during a hot spell in April.
How? This hypothetical network, what I call a Super Merchandiser, would be a voracious reader. It would quickly digest trends and tastes as reflected on Facebook and Instagram, and in fashion-magazine websites and news stories. It could track shoppers’ Web-browsing habits and, by internalizing and digesting all this information, smartly predict shoppers’ intent. This would allow retailers to present the right products to shoppers at the right time, boosting customer satisfaction—and of course sales.
We are already starting to see some companies dabble in early forms of AI. StitchFix, for example, uses algorithms to pick out custom clothing to mail to each customer, learning over time about what will appeal to them. Essentially, StitchFix is trying to figure out what the customer actually needs, and not just what she wants—and is betting that this value to customers overrides any privacy concerns. But most retailers are still behind.
Here are three other specific ways AI could change the future of retail.
Failure to launch
What if smart retailers could predict the success of product launches? This would result in new revenue and fewer costly inventory problems.
As an example, I recently tried to buy an iPhone 7. It was sold out. I was disappointed, but in many ways I shouldn’t have been surprised; retailers have long been bedeviled by the complicated process of predicting product demand.
But with the recent advances in AI, it doesn’t have to be this way. Apple shouldn’t just assume that users will upgrade their phones every two years, or depend on user surveys (I never answer them) to figure out if people are craving new features in a new model.
Why not listen the data generated by specific customers? As a gadget freak, I had been searching for news about the iPhone 7 for months. Apple could harness this information, combined with other data about my tech preferences and past purchasing history, to figure out that I was a likely buyer, and stock appropriate inventory in stores in my area.
The lack of specific buyer information can be even worse for other retailers in more competitive markets. If the FitBit I wanted was out of stock, I could easily decide to buy a competing Misfit, Jawbone or Xiaomi Mi device. AI could prevent this.
You’re a fraud
Fraud is one of the biggest problems in retail and one of the most costly. Credit-card fraud, which cost retailers $32 billion in 2014, is just part of the equation. The other is fraudulent behavior by consumers, such as returning used goods and claiming they’re new.
Efficient, sometimes “no questions asked” return policies have become the gold standard in e-commerce today. And for most people, product returns are legitimate. But AI could help retailers target, and tax, the minority of consumers who cause most return problems.
Data exists for which customers have a history of returning which products, and where—online or in physical stores. Retailers should be centralizing this data and using it to their advantage. By smartly tracking this activity, and predicting it, retailers could change return policies for certain customers—implementing a return or re-stocking fee for repeat offenders, for example. One Silicon Valley electronics chain implemented a re-stocking fee for digital cameras about 10 years ago, for instance, after it discovered people were buying the devices for single events and then returning them. These fee programs are similar to how car insurance works today: People who have more accidents, or engage in riskier behaviors, pay more than safe drivers.
Sometimes, it’s necessary—even fun!—to go to a brick-and-mortar store. Maybe you need to try on pants for work, or find the right kind of piping or paint for a home-repair project. But often, finding specific items in department stores or big-box stores can be a real pain, resulting in lost time and customer dissatisfaction.
Robots powered by machine learning can help solve this problem. Imagine entering a Target or Macy’s and simply taking a seat on a comfy chair near the entrance, sipping a cup of coffee while a robot listens to your request and scoots off to find your product on crowded shelves. Perhaps you would like to find three specific dresses in your size and, if they don’t fit, get recommendations for new styles based on your preferences, sizing and past purchases. You could even order the robot around through a smart app on your phone.
Clever retailers like eyeglass maker Warby Parker are using data and AI to open “smart” brick-and-mortar stores, moving from online to offline the same way most retailers—from Macy’s to Wal-Mart to Toys R Us—have moved from offline to online and established two main distribution channels. At a smart store, managers pore over data to figure out which specific products to offer where, and to whom.
Similarly, consumers can get smarter—and more robot-like—themselves through data delivered through smart phones, smart watches and virtual/augmented reality devices. Standing in a Target or Best Buy, they wouldn’t have to ask someone why a certain high-definition TV is better than another; a virtual assistant could give them an overview and run through specific features, even delivering a final purchase recommendation.
In the end, artificial intelligence can be good for buyers and sellers in the retail world, and will likely help many retailers scale faster. As Kleiner Perkins Partner Mary Meeker pointed out in her annual “Internet Trends” report earlier this year, some retailers are now achieving $100 million in annual revenue in five years or less; it took Nike 14 years to reach that milestone, and Lululemon eight. The use of AI by retailers will only accelerate this trend. Really, it’s only a matter of time before the industry smartens up.
This post originally appeared on Forbes.
For a full list of all Battery investments and exits, please click here.