AI at the retail edge: What’s new, and what’s coming soon

There’s been an absolute explosion of interest in AI, especially generative AI (GenAI), in the last year. Simultaneously, increases in compute power have made it easier to implement AI use cases at the retail edge. That’s a perfect opportunity for some long-awaited retail use cases to turn prime time. Far from just gimmicks, these use cases will usher in a new era of smart stores that boost customer experience while increasing staff efficiency to drive down costs. I know we’ve all heard this before, but let’s walk through some use cases that are finally in the realm of possibility.

They always get you at the drive-through

We’ve all experienced it, and it shapes our willingness to return to the scene of the crime. What is supposed to be a quick meal on the go too often turns into a grim choice between eating something you didn’t order or parking and walking into the restaurant to stand in line to complain … and then wait for your order again.

When you watch the drive-through attendant trying to take orders over a noisy headset, handle payments, and package meals at the same time, you begin to understand why there are so many mistakes. And these mistakes are costly, around $26 million annually for a national chain restaurant.1

Fortunately, we finally have the tools to fix this. Conversational large language models (LLMs) can process spoken language and eliminate errors during order taking to make sure the kitchen gets the right instructions, even if you order “a double burger, without tomatoes, no wait sorry, I meant hold the lettuce, but go light on the ketchup, oh and actually let’s make that a cheeseburger, oh but extra onions.” They can even make context-relevant suggestions for upsells in natural language: “You know if you want the meal deal, I can sub in some rings instead of fries for you.” Then, as the order is being prepared, computer vision AI can verify if the food being prepared matches the order ticket, prompting staff to correct mistakes while at the same time verifying that all the items going into the bag are correct, including the proper utensils, condiments, napkins, straws, ketchup packets, etc. The drive-through attendant can focus on taking payments and doing the final check before handing over the food. The result is fewer mistakes, lower costs, happier customers, and less stressed employees.

It’s not blue, it’s not turquoise … it’s cerulean!

We’ve all done it. You walk into a store to buy a shirt, not really sure where to start. When the staff asks if they can help you find anything, you say, “No thanks, I’m just browsing.” You pick out a couple of different shades of blue and take them to the dressing room. You try them on. One fits well, but you don’t love the color. Another one is too big, but you don’t want to get dressed and go out and find a smaller size. The last piece you try on looks okay and fits okay. Well, I guess that’s the one. It’s in the dressing room that shoppers are converted into customers, often outside the purview of any staff that might be able to help or upsell.

GenAI on its own hasn’t proven adept at solving this problem. Until now, there haven’t been enough data points available at the right time for effective recommendations. You couldn’t tell what someone was buying until they swiped their card, and by then they were done shopping.But when combined with a long-standing technology — radio frequency identification (RFID) tags — smart shopping is finally beginning to deliver on its promise. RFID tags have been around for decades and now cost just pennies. With Walmart®, Target, Macy’s, and Nordstrom all mandating RFIDs from suppliers,2 and RFIDs already in use by 93% of retailers,3 the technology is finally reaching the critical mass the apparel industry needs.

By feeding real-time RFID data into GenAI models, retailers can finally implement smart changing rooms that simultaneously increase conversion ratios and improve customers’ shopping experience. RFID readers can detect what products customers bring into the dressing room while interactive digital signage inside can display product details and allow the customers to find out if there’s a different size or color available and signal staff to bring it to the dressing area. GenAI-powered recommendation engines can suggest accessories and coordinating pieces in real time.

RFID has several other beneficial uses in clothing retail. RFID tags combined with GenAI can be used for inventory tracking, loss prevention, and stocking. They can help employees locate requested products, even if someone has moved them. RFID can speed up checkout times instead of requiring staff to find a barcode tag on a piece of clothing. RFID can even enable self-service checkouts in the apparel industry.

Combining convenience with loss prevention

Surveys show that 40-60% of shoppers prefer self-checkout,4 but many chains have reduced self-checkout due to losses being more than 16 times higher than with human cashiers.5 Combining computer vision AI with self-checkout can stop many common shoplifting tactics so retailers can continue to provide their preferred checkout method. One common shoplifting tactic, referred to as the ‘Switcheroo,’ is to place an expensive item, such as steak or seafood, on the scale but enter the price look-up (PLU) code for a banana instead. Computer vision AI can visually match the code with the item on the scale and prompt the user to re-enter the code or notify staff to assist. Even better still, AI can simply detect and automatically select the item at the point of sale (POS), eliminating the opportunity all altogether, while also speeding up the checkout flow.

Well-trained AI will be able to tell the difference between a Gala and a Honeycrisp, or between London broil and filet mignon. The technology will also help boost the adoption of self-checkout because shoppers won’t have to manually look up and enter their own produce codes — a proposition that currently makes self-checkout less appealing.

Getting from here to there

The technology is there, and the benefits are clear. But for many retailers, legacy systems may be a roadblock to adopting these use cases. Every example in this article requires multiple systems to share data with each other. This can take a significant amount of platform engineering services to accomplish.

Dell Technologies is building a catalog of select partners who deliver the individual functions discussed here in an easy-to-install fashion on our NativeEdge platform. Additionally, our catalog also includes partners such as EPIC iO that specialize in stitching together data from multiple sources for AI and analysis using their EPIC iO DeepInsights platform. Running DeepInsights on the Dell NativeEdge devices allows you to seamlessly integrate, process, and analyze data from an expansive variety of assets and systems on a single platform for enhanced control and oversight. The ability to exchange data from multiple sources and systems on a common pub-sub bus sets the stage for innovation as new solutions become available.

Dell NativeEdge, an edge operations software platform, can be a game changer for edge deployments. With Dell NativeEdge, retailers can securely scale edge operations using automation and zero-touch provisioning to push new servers and applications out to every store without requiring someone to physically install and update servers on-site, especially critical in the rapidly changing landscape of AI providers. Plus, an open design and multicloud connectivity optimize investments by allowing retailers to consolidate new and existing edge applications on the same server.