Everyone seems to be “using” artificial intelligence these days. So is retail. Big players like Amazon and Target are pouring huge amounts of resources into machine learning, and many companies sell “Artificial intelligence” tools for the retail industry.
There’s just one problem. Most of what the retail industry refers to as artificial intelligence isn’t AI. Furthermore, it’s bad for both customers and retailers. Using the “AI” that worked online in physical stores risks making physical stores look increasingly like websites amid a larger trend towards automation and reducing human presence in stores. This is altogether a very poor idea.
I teach at MIT and have worked with artificial intelligence tools to solve problems across several domains for years. AI is a field, but also an aspiration, we’re sold on (or fear) the aspiration, but it’s important for retailers to understand what AI –the field– can deliver today and how it can help (and hinder).
Despite all the hype, there is a place for artificial intelligence in retail for the future.
AI myths AI hype
Many retailers have forgotten what really helps customers: In-store assistance from human workers that’s relevant to their shopping needs. Artificial intelligence can help human employees with that task, as well as with the less glamorous but quite important realm of supply chain and logistics management that has given Amazon its edge.
In order to understand how AI tools can help, it’s important to understand what AI is not.
In the retail world, products like Amazon Alexa and chatbots are commonly referred to as AIs, but they’re just sophisticated programs built on top of machine learning, natural language processing, and statistical algorithms.
There’s also a lot of hyperbole around machine learning and analytics that conceals a people problem. Used that way, machine learning needs vast amounts of data that needs to be formatted and cleaned for use. Computers aren’t good at automatically cleaning data; humans are.
The technologies we use today are hobbled by this. In order to receive clean data for machine learning, for instance, many retailers send customers questionnaires which are easier for computers to process.
Customers aren’t AIs. Most never answer the questionnaires, and many fill whatever they remember. This leaves retailers with faulty but Big data.
Meanwhile, Amazon and other online-first retailers understand every bit of customers’ purchase and shopping habits. Their machine learning models, however imperfect, see all the data. However, brick and mortar retailers know little about the items you picked up, glanced at, and put back on the shelf or what you looked at next.
Amazon’s success is tantalizing. It is easy to forget that online-retail needed machine learning to seize their only advantage and overcome their enormous disadvantage—you get to shop from home with nothing but a picture and comments from strangers.
In brick-and-mortar stores you simply can’t get data to use machine learning wholesale; but you do get to meet and greet the customer (that’s if you, the retailer, are still there). You need better questions to take advantage of the same “AI” tools. Your advantage is that you are not a website.
Less Algorithms, More Psychology
In retail, many of the most vaunted algorithmic success stories have as much to do with human psychology as with engineering know-how.
For instance, Amazon’s recommendation engine has been described as artificial intelligence by some. The engine’s success, however, is based also on simple human psychology.
When you shop on Amazon, pay attention to how no page has a single set of recommendations or a short list of related items. Instead, there are many recommendation sets scattered around the page.
The success of Amazon’s recommendations isn’t just due to how they use the tools we’ve developed to figure out artificial intelligence: It’s the novelty of being bombarded with more-or-less relevant recommendations, delivered by sophisticated algorithms that guarantee you continue browsing.
How Retail AI Can Work Together
There are two areas where a more genuine use of those same tools holds potential for retail: Logistics and turning customer assistance into expert advice and information gathering.
AI technologies help interpret information differently from humans. This means that they hold great potential for priming supply chains, potentially all the way to your physical store.
Using the same “AI” tools that work for online-retailers any retailer should be able to make sure items travel from point A to point B timely; maybe, even, so customers don’t have to haul their purchases.
And, by the same token, the physical store is where humans first interact with physical products and can generate “product intelligence” in ways online retail simply can’t.
In the future, retail employees could use AI tools that help them become instant experts about a product, and as they do, feed data back to those AI tools about how customers respond to products. In that ideal future the AI would help forecast product improvements to sell back to producers, generate recommendations on where to stock and display products, and share insights about inventory that boost store profits and customer satisfaction.
Today, however, retailers confuse analyzing large amounts of data and profiling customers for artificial intelligence. Throwing data at machines doesn’t make machines (or anyone) smarter.
Just as customers are encouraged to scan and bag their own items at self-checkouts, the need for data has moed retailers to ask customers to infinitely browse online or use smartphone apps in-store instead of receiving assistance from paid employees inside the store.
Increasingly customers perform unpaid labor once performed by paid employees. And retailers themselves are set on making physical stores look like websites, with fancy decorations and no other humans to talk to. This may be artificial, but it is not intelligent.
Over the years, I’ve learned that “AI” is at its best when we program it to address problems that are hard for humans; when not used to upskill humans, however, all it does is shift work from employees to customers. AI is simply different.
A happy relationship between AI retail requires people, not apps. In retail, artificial intelligence tools should make employees more knowledgeable, producers better informed, and customers happier. This means putting machine learning and AI tools towards asking better questions, upskilling employees, same-day delivery and installation, and other things that encourage customers to shop in stores as much as they do from their homes.
Remember: Machine learning gets better with humans.
Luiz is the author of Innovating: A Doer’s Manifesto for Starting from a Hunch, Prototyping Problems, Scaling Up, and Learning to Be Productively Wrong. Learn how to write for Quartz Ideas. We welcome your comments at email@example.com.