This article is part of a VB Lab Insights series on AI sponsored by Microsoft and Nvidia. Don’t miss additional articles providing new insights, trends and analysis on how AI is transforming organizations and industries. Find them all here.
Retail cannot get enough of AI. Analysts say the sector’s spending in 2023 will outpace all others except banking. A 40% adoption rate is projected to double by 2025, making retail the industry most heavily invested in intelligent technology.
Companies are turning to AI to handle a long list of challenges that would keep any business leader awake at night: brick and mortar revenues impacted by consumer behavior changes, worsening shortages of staff and rising labor costs, supply chain disruptions, heavy pressures on profit and costs (including inflation and double-digit increases for customer acquisition), and massive loss due to theft and organized crime.
Beyond these pressing issues, retailers expect that AI-driven analytics and applications can help them navigate long-term changes. Chief among them are major shifts in buyer demographics (more diverse, digitally savvy, older), consumer values (price and convenience over brand loyalty), sales channels (rapid growth of ecommerce, mobile, and social) and growing demands for global sustainability.
Against this backdrop, here’s a look at two areas of high-value AI innovation in retail.
Intelligent stores: Fighting loss, optimizing business
Despite more than three decades of online growth, physical locations remain important anchors for many retailers. While the two worlds continue to meld, business and technology leaders are focusing on ways to evolve, differentiate and optimize the customer experience (CX) and business performance of stores. For many, loss prevention, also known as asset protection, tops the priority list.
According to the National Retail Federation (NRF), retailers worldwide forfeit more than $100 billion each year to “shrinkage”, the industry term for inventory theft, loss and waste. More than half occurs in North America. Average shrink rates exceed 1.5% of revenues, so for a $20-billion grocer, that is a hefty $300-million yearly hit. COVID and inflation further aggravated the problem: In a recent survey, 89.3% of respondents reported increases in violence, shoplifting (73.2%) along with employee theft and organized retail crime (71.4% each).
To fight back, more retailers are adopting intelligent video analytics (IVA) technologies that can accurately and efficiently reduce shrinkage. New AI-driven systems help prevent losses in real-time, improve asset protection at points-of-sale, reduce shoplifting storewide, and ensure the safety of employees and customers, who bear the cost in higher prices.
From warehouse to checkout, Everseen sees everything
Everseen, an international software company based in Ireland, has developed computer vision AI systems that help retailers see and solve shrinkage problems in real time. Alan O’Herlihy, CEO and founder, says using this AI at the edge effectively transforms an entire physical retail space into actionable data that can drive better decisions. Here’s how it works:
Running on the NVIDIA AI platform on Microsoft Azure, the solution modules integrate with a retailer’s existing cameras, point-of-sale, computers and other business systems. Doing so provides an end-to-end view across their entire supply chain — from warehouse to store to shelf to checkout — that can pinpoint gaps in inventory and other problems requiring immediate attention. The AI then recommends a “next best” action — all in real time.
For example, Evercheck, the company’s point of sale (POS) solution, instantly detects and corrects both deliberate and unintentional errors at staffed and self-checkout lanes. For the latter, an instantaneous “nudge” (delivered within 300 milliseconds) prompts shoppers to correct mistakes such as an unscanned item, reducing the need for staff intervention and potential conflict down to 2%, from 20%. Another AI-powered product, Everdoor, reduces loss and improves process compliance in the stock room.
All told, Everseen each day analyzes in real time a staggering 275 years of diverse and labeled video. The company monitors unstructured data from 22 million customer interactions with 220 million products. O’Herlihy says the insights and actions derived are invaluable in helping retailers reinvent related business processes. That, in turn, can yield a host of benefits, he says, including increased revenues and sales throughput, reduced costs, mitigated risk, better customer experiences and optimized operations in distribution centers.
According to Everseen, more than half of the world’s top 15 retailers have adopted the company’s AI-based loss prevention systems, a total of 6,000 stores and 80,000 checkout lanes. Says O’Herlihy: “The goal of our AI is to reduce the friction for ‘green actors’ and increase friction for ‘red actors’ by dynamically delivering intuitive fixes and split-second decision-making.” More on Everseen’s seamless shopping.
Other emerging uses of AI in intelligent stores:
Optimizing layout and experience. Leading firms are exploring how digital twins and simulation can create smoother experiences for customers and employees. Lowe’s uses AI-driven simulation with NVIDIA Omniverse to enhance store layout, optimize merchandising and improve employee productivity. The same technology helps Kroger design the best customer shopping experience, including fast and easy checkout.
In-store ads and promotions. Intelligent targeting delivers live shopping suggestions that can expand cart size through opportunities for upselling and cross-selling. Dynamic digital signage, such as that delivered by Cooler Screens, automatically updates to offer promotions tailored for every shopper and creates additional revenue opportunities tied to dynamic in-store displays. See more here.
Autonomous shopping. Smart “grab-and-go” stores, where customers use their mobile phones to check out, are fast gaining popularity. New approaches include AI-enabled shopping carts, nano stores, smart cabinets and fully autonomous stores. All solution providers, including AiFi and AWM, seek to give customers a more “frictionless” and faster shopping experience that boosts retailers’ revenues and margins.
Generative AI: Pinpoint design for real and digital fashion
Generative AI like DALL·E and ChatGPT can be used to create new designs for products based on customer feedback, sales and market trends and other data. Taking advantage of these tools can help retailers develop new products that are more appealing to buyers and better aligned with market demand.
Startup Fashable is pioneering the use of generative AI to create sustainable fashion designs without the need for fabric.
Unsustainable manufacturing, unsold inventories and long production cycles are common (and costly) problems in fashion. While a high-end designer might take months or years to design and produce a collection, “fast fashion” brands do so with a fraction of the time and cost, thanks to inexpensive materials and labor. But what happens when clothing production goes up while its lifecycle goes down? A growing landfill problem. In the U.S. alone, 21.6 billion pounds of textile waste get trashed every year.
So, in 2021 the Portugual-based startup, led by co-founders experienced in software engineering and AI research and development, envisioned a disruptive AI application. It would create dozens of original and realistic clothing designs and fashion content in minutes without using any material. The pair believed that a smart, all-digital approach would help fashion companies better meet customer demand, get to market faster and reduce fabric and clothing waste, explains co-founder Orlando Ribas Fernandes.
Entire digital collections with a few clicks
The Fashable team created its AI algorithm on Azure AI Infrastructure powered by NVIDIA A100 Tensor Core GPUs, Azure Machine Learning, an enterprise-grade service for the end-to-end machine learning lifecycle and PyTorch, an open-source machine learning framework. The system lets designers quickly generate endless digital options for fashion in the metaverse or the real world, such as the shirt below. More technical detail here.
Fashable AI is composed of different neural networks. These ingest data from multiple sources to learn about trends, styles and clothing types. The models are constantly learning “what’s in” and “out.” Soon, these capabilities will enable co-creation of fashion in real time. Designers could, for example, visually change a digital design to shorten the sleeves of a dress or change a pattern from stripes to polka dots.
In one click, Fashable AI can create an entire collection. Designers can take their pieces to social media to A/B test directly with customers to gauge interest and forecast demand before going into production. Where it used to take months to get a new collection from design to department store, with Fashable it now takes minutes — with far less labor and no guesswork.
The company’s customers use its AI technology across various production phases:
- Creation — from mood boards to iterations and final assortment (design)
- Industrialization — from assortment to tech specs and integrations with tools
- Content creation — from final product to retail-ready content, including imagery for ecommerce, editorial and selling unsold inventory
And the company has moved into metaverse immersive commerce. Brands can now use Fashable to start creating collections for different digital worlds. “Without AI, the process was slow and labor-intensive,” says Fernandes. A recent collaborative collection with Wrong Weather, a casual luxury brand demonstrated Fashable’s potential.
Disrupting the fashion status quo
Today, Fashable bills itself as “Deep Tech for the fashion Industry,” “The ChatGPT for AI Images and Content Generation” and “The Most Powerful Generative AI Toolkit for The Metaverse and Physical Fashion.” The broadened value statements underscore the benefits for designers in both worlds: saving money on research, design and content creation, reducing copyright issues, and freeing users to focus on high-value design tasks, explains Fernandes.
“With social media, the metaverse and Web3, the need for new content is exploding for fashion brands,” Fernandes says. “The war for new content has never been so intense. Only AI can generate very realistic images to solve that need.”
Fashable is convinced of the disruptive power of AI more than ever. Beyond keeping today’s trends out of tomorrow’s trash, he believes personalized and exclusive garment designs are the key to business success in physical and virtual worlds. “Generative AI,” he says, “will completely change the status-quo in the fashion industry.”
Generative AI is helping retailers in other important ways:
Merchandising and product onboarding. Generative AI can generate images, music, fonts, videos, 3D models and more for advertising and marketing. Custom images can show how a product looks in different settings. For ecommerce, generative AI and computer vision can create product descriptions, attributes, meta tags and cataloging based on product images.
Service chatbots and conversational AI. For customers and agents, AI helpers can provide virtual assistance, language translation, order status, search, product recommendations, email and chat answers. Brand avatars are delivering consistent omnichannel customer service whether on a kiosk, mobile app, ecommerce or in the drive-thru. Employees can get answers to FAQs via voice, text, videos, and images in multiple languages.
The right infrastructure is crucial for retail AI
The plentiful opportunities for transformational use of AI also brings retailers new challenges.
As in other industries, many companies will discover they lack the powerful infrastructure needed to develop and deploy AI-enabled applications. Requirements here are typically far more demanding than for conventional computing, especially with large model sizes and high complexity. Optimized, accelerated environments that are “purpose-built” for AI are needed to deliver real-time speed, predictability and accuracy.
Here is what retail experts say is needed for AI success: The versatility to support diverse models with one end-to-end application that can deliver the desired user experience. High performance and scalability to optimize time-to-solution and deployment costs. An end-to-end solution stack that supports the entire workflow — including data prep, model building, training and deployment within an AI-powered service. And a uniform stack to flexibly train in the cloud and deploy at the edge in stores and other locations.
For many, meeting these criteria will mean adding AI to the growing list of critical workloads shifting to the cloud as a way of gaining high-performance processing, servers, networking, storage, development platforms and environments and software without heavy capital costs.
Getting infrastructure right is crucial, agrees Everseen’s O’Herlihy. He says a high-performing cloud environment has been crucial for his company’s AI success in several ways. It enables scaling across thousands of locations, easy “lifting and shifting” of technology building-blocks from one part of a store to another and delivers high performance that lets AI models understand a moving scene, including analyzing how humans interact with objects, in real time.
Fashable’s Fernandes concurs. “We talk a lot about on-demand cloud infrastructure and services to accelerate product innovation and build competitive advantage. ‘Doing more with less’ with a small team is important, so we can invest in our IP and leverage Azure Machine Learning and PyTorch with the NVIDIA AI platform to achieve state-of-the-art results. This is crucial for minimizing upfront investments and risks so that our applied research team can fail and adapt quicker.”
A related important consideration, he says, is the data used for AI innovation. Fashable opted to build its own datasets with internal tools to get “complete control” of its innovation. But, Fernandes acknowledges, “the risks and cost can be prohibitive for building AI innovations, so business leaders can minimize those by using existing environments and tools and not ‘reinventing the wheel.’”
Laggards risk sharing $1.7 trillion of new industry value
Today’s retailers live at the intersection of commerce, consumers and technology transformation. Some 87% are planning to increase investments in AI/ML in 2023. Yet researchers say many retailers have not gotten started with smart technologies. Laggards risk missing out on the $1.7 trillion in business value, roughly 12% of all sales, that McKinsey estimates AI and analytics could generate for the retail industry.
For many economists and forecasters, retail is an important canary in the coal mine. Sales and investments are seen as key indicators not just for the sector, but for the overall economy. If so, retailing’s surging embrace and emerging success using AI for innovation and transformation amidst multiple headwinds bodes well far beyond store walls.
VB Lab Insights content is created in collaboration with a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact email@example.com.