The Brick Strikes Back: Traditional Retailers Adopt AI to Level the Field

No one disputes that online retailing has its share of advantages over brick and mortar: Convenience. Personalization. Pricing. But all is not lost for those traditional retailers who are smart enough to co-opt digital leverage rather than deny or compete head-on. An intelligent blend of on-line presence married to an agile and intelligently-instrumented physical presence can create sustainable advantages for the legacy retail industry.

“Smart Retail” is the not-so-secret weapon against UBS’s predictioni that 7% of U.S. retail establishments will vanish by 2026 if online penetration continues at its recent pace. Artificial intelligence and edge computing, embodied in the new Artificial Intelligence (AI) platforms from Blaize, can help make smart retail a reality.

Making Smart Retail a Reality

Many of the advantages that pure e-commerce companies enjoy arise from the data they are positioned to collect and their ability to wield it with instantaneous intelligence. What if traditional brick and mortar retailers could enjoy these same advantages? The key is to combine the data agility of an internet business with the strength of physical storefronts.

All retailers recognize that customers demand outstanding service, loyalty programs with real rewards, and, most importantly, a personalized shopping experience. But these tools are not limited to use by online vendors – they can be successfully employed by legacy retailers and virtuously magnify the power of an in-person shopping experience with an informed sales staff.

Olay’s Skin Advisorii lets users upload a selfie and receive an analysis and recommendations for problem skin areas. While currently an online service, it’s easy to imagine this application as an in-store kiosk experience set into the skincare shelves, especially since Olay maintains that 94% of women agreed that recommended products were right for them. And while Skin Advisor is an unofficial stand-in for a dermatologist visit, it’s not a long leap to see the same application offering color matching and blending tips for cosmetics. All it takes is a miniature camera, some database work, and a bit of augmented reality feeding an attached screen — bonus points for real-time, AI-driven voice interaction.

Amazon Go is the poster child for retail AI. Amazon, ever the company obsessed with optimizing the customer experience, describes Amazon Go as a “checkout-free shopping experience […] made possible by the same types of technologies used in self-driving cars: computer vision, sensor fusion, and deep learning. Our Just Walk Out Technology automatically detects when products are taken from or returned to the shelves and keeps track of them in a virtual cart. When you’re done shopping, you can just leave the store.”iii With 18 locations in four major U.S. cities (as of Q4 2019), Amazon plans to deploy its automated retail model aggressively, opening up to 200 storefronts annually over the next several years. iv And Amazon isn’t alone. See also SmartMart drive-thru stores, v and Dutch bookseller Boekhandles Groep

Ralph Lauren deployed a smart mirror display in its Manhattan Fifth Avenue store.vii Shoppers step into the changing room, see how clothes look on them, and, without stepping away from the touch-sensitive mirror, “have the entire store and [an] associate at your fingertips.” According to Ralph Lauren, 90 percent of those who encounter the smart mirror engage with it. Just imagine where results will go when camera and AI systems allow shoppers to see alternative designs and outfits placed on their reflections in real-time.

Putting Intelligence in Retail

Juniper Research expects retailers to spend $7.3 billion on AI by 2022 — more than three times the amount spent in 2018. viii Naturally, that begs many questions around where and how to place those investments. Juniper notes that spending will focus on “AI tools that allow [retailers] to differentiate and improve the services they offer to customers.”

Improvement is typically incremental; it does not always mean implementing something radically new. For example, nearly every store has cameras. AI at the edge will allow those existing cameras to be used in new and value-added ways. Amazon Go has hundreds of cameras that connect to their backend systems, all running a range of AI algorithms to track inventory and shopper activity in real-time.xi

Imagine how much more responsive and capable these systems will be when the AI computing can be done in the store rather than in the cloud or the data center. In real-time AI applications, long latency can mean lost opportunities and poor satisfaction for both the shopper and business manager. Sophisticated shopper analysis and prognostication algorithms are remarkably complex, currently required data-center grade computing resources. It takes a (relatively) long time to collect video, aggregate it, transport it to the data center and then apply AI algorithms to it. With the currently ‘central office’ approach there may not be time to receive and act on results with recommendations for next best action. That computation and output needs to be local and immediate. This is why global edge computing is compounding at over 27% annually.x

The highly efficient Blaize Graph Streaming Processor(GSP®) architecture solves this latency problem. Conventional compute solutions when applied to AI problems carry over operational inefficiencies from legacy architectures and power budgets unsuited to tiny IoT and edge deployments. Blaize GSP architecture, which combines an innovative system-on-chip (SoC) design with open AI platforms and software, was built for AI Everywhere from the data center to the edge.

Retailers can work with solution integrators to craft the innovative, incremental service improvements they need. From more intelligent camera surveillance to customer heat mapping to smart mirrors, Blaize enabled solutions can drive more capability into more devices. Applications can be easily upgraded through over-the-air networks, just like smartphones, as new features and improvements become available, and software can even jump between hardware solutions if needed.