Chapter 27: AI in Retail Checkout Systems |
Executive Summary |
Cashierless stores represent one of the most visible and exciting applications of artificial intelligence in the retail sector. By combining RFID tagging, computer vision, sensor fusion, and deep learning algorithms, these systems enable automatic product detection and billing without traditional checkout lines. This chapter provides a comprehensive, accessible overview of how AI-powered checkout systems work and explores their real-world deployment at major American retailers like Amazon, Walmart, and Sam's Club, as well as Chinese leaders like JD.com. We will examine Amazon's pioneering Just Walk Out technology, which has processed over 36.7 million items in a single year across more than 150 locations, and the company's strategic addition of RFID to handle apparel and soft goods. We will explore Sam's Club's ambitious shift to a fully digital, register-free store model using Scan & Go and AI-powered exit arches, and JD.com's early adoption of unmanned stores integrating facial recognition and RFID. The evidence reveals that while cashierless technology is transforming the shopping experience, it also faces significant challenges around cost, privacy, theft prevention, and accessibility that retailers must navigate carefully. |
|
1. Introduction: The End of the Checkout Line |
Think about the last time you went grocery shopping. After walking through the aisles, filling your cart, and making your selections, you arrived at the front of the store. There it was: the queue. A line of shoppers, each waiting to unload their cart, have each item scanned, and complete their payment. For many, this is the most frustrating part of the shopping experience. It is a bottleneck that consumes time, creates friction, and adds nothing to the customer's enjoyment. |
Now imagine a different scenario. You walk into a store, pick up what you need, and simply walk out. No scanning. No cashier. No waiting. The items are automatically detected, your account is charged, and a receipt appears on your phone. This is the promise of cashierless retail, and it is rapidly becoming a reality. |
At the heart of this transformation lies a powerful combination of technologies: Radio Frequency Identification (RFID), computer vision, and Artificial Intelligence. These systems collectively create a 'smart store' that can see what you are taking, track it in a virtual cart, and automatically process payment when you leave. The market for smart checkout systems is growing rapidly, with a global market size estimated at approximately USD 5 billion in 2025 and projections to reach nearly USD 8 billion by 2031 . This growth is driven by rising consumer demand for frictionless shopping, labor cost optimization, and the lasting impact of the COVID-19 pandemic, which accelerated interest in contactless experiences . |
This chapter explores how these technologies work, how they are being deployed by major retailers in the United States and China, and what the future holds for the checkout-free store. We will see that while the technology is impressive, it is not without its challenges---and the path to widespread adoption is as much about managing human behavior and expectations as it is about advancing AI. |
|
2. How the Technology Works: The Invisible Checkout |
To understand the impact of cashierless stores, it helps to first understand how they actually work. The technology behind them is sophisticated, but the core principles are accessible. |
2.1 The Three Pillars: RFID, Computer Vision, and Sensor Fusion |
Cashierless checkout systems typically rely on a combination of three key technologies, each serving a different purpose . |
Radio Frequency Identification (RFID) is a technology that uses radio waves to identify and track objects. Small tags containing a unique identifier are attached to each product in the store. When a customer picks up an item and walks through an exit gate equipped with an RFID reader, the system automatically detects every tagged item leaving the store . RFID is particularly valuable because it does not require a direct line of sight---tags can be read through packaging, bags, and even clothing . |
Computer Vision uses cameras and artificial intelligence to 'see' and interpret the environment. In a cashierless store, ceiling-mounted cameras track the movements of shoppers and the items they pick up. AI algorithms analyze the video feed in real time, detecting when a hand reaches for a product, whether the product is taken or returned to the shelf, and which shopper is holding it . These algorithms use deep learning techniques, including transformer models similar to those used in generative AI, to achieve high accuracy . |
Sensor Fusion is the practice of combining data from multiple sources to achieve a more accurate result than any single source could provide. In a smart checkout system, data from RFID tags, computer vision cameras, and weight sensors on shelves are all fed into a central AI model. The AI cross-references this information to determine exactly what a shopper has taken. For example, if a camera sees a shopper pick up a bottle of juice, the shelf sensor detects a reduction in weight, and the RFID system confirms a juice tag has left the store, the system has high confidence that the juice should be added to the shopper's virtual cart . |
Research has demonstrated the effectiveness of this multimodal approach. A study on intelligent shopping cart product recognition found that a system fusing computer vision, weight sensors, and RFID achieved a recognition accuracy of 98.7%---a 5.2% improvement over vision-only systems and a 3.8% improvement over RFID-only systems . The average processing time per product was just 0.3 seconds, meeting real-time checkout requirements . |
2.2 The Customer Journey: From Entry to Exit |
For the shopper, the experience is designed to be seamless. Here is how it typically works in a fully automated store: |
Entry: The shopper authenticates themselves at the entrance, often by tapping a credit card, scanning a QR code from a mobile app, or using a palm recognition device like Amazon's Amazon One . This links the shopper's identity to a payment method. |
Shopping: As the shopper moves through the store, the system tracks their movements. Cameras identify which items are taken and which are returned. In stores using RFID, the tags on products are continuously read, allowing the system to maintain an up-to-the-second virtual cart . |
Exit: When the shopper is finished, they simply walk out through a designated exit lane. RFID readers scan all tagged items leaving the store. In computer vision-only systems, cameras confirm the final items in the shopper's possession. The system compares the final virtual cart to the shopper's identity, charges the payment method on file, and sends an electronic receipt . |
2.3 Hybrid Models: The Middle Ground |
Not all cashierless experiences require a fully automated store. Many retailers are adopting hybrid models that offer some of the benefits of automation without the full investment . |
Scan & Go Apps are mobile applications that allow shoppers to scan items with their phone as they shop, building a digital cart. At checkout, they pay through the app and show a digital receipt. This approach requires less in-store infrastructure but places the burden of scanning on the customer. It also introduces risks: studies have shown that shoppers with 50 items have a 60% chance of accidentally missing at least one item, and those with 100 items have an 86% chance . |
Smart Shopping Carts are equipped with cameras, weight sensors, and RFID readers. As items are placed in the cart, the system automatically identifies them and adds them to the cart's digital tally. The shopper can pay directly at the cart. This approach offers a cashierless experience without the need to retrofit an entire store with ceiling cameras and exit gates . |
|
3. Amazon: The Pioneer of Just Walk Out |
Amazon is the company most closely associated with cashierless retail, and its Just Walk Out technology has become the benchmark for the industry. |
3.1 The Birth of Just Walk Out |
Amazon first unveiled its Just Walk Out technology in 2016 with the opening of the first Amazon Go store in Seattle. The concept was revolutionary: a convenience store where shoppers could grab what they wanted and simply walk out. The technology combined computer vision, sensor fusion, and deep learning to track shoppers and their selections in real time . |
In 2018, Amazon began deploying the technology in its Go convenience stores, and it quickly expanded to other formats, including Amazon Fresh grocery stores. The technology was initially based purely on computer vision and sensor fusion, with no RFID component. This worked well for packaged goods and food items, where products are typically placed on shelves in a fixed position and can be easily tracked by cameras. |
3.2 The RFID Addition: Solving the Apparel Problem |
However, the original Just Walk Out system had a limitation: it struggled with apparel and other soft goods. Clothing is typically displayed on hangers, can be folded or bunched up, and is often carried in a shopper's hand or bag in ways that obscure it from overhead cameras. The camera-based system could not reliably track these items . |
Amazon solved this problem by adding RFID to the Just Walk Out technology mix . Each apparel item is tagged with a unique RFID tag. When a shopper picks up a garment and walks through the exit gate, a portal-style RFID reader captures the tags and registers the items as purchased. This allows for a seamless experience: shoppers can grab clothes, try them on, even wear their purchases out of the store, and the system will still correctly identify and charge for everything they are taking. |
The RFID-enhanced Just Walk Out technology was first piloted at Seattle's Climate Pledge Arena during the 2023 hockey season. Based on the success of this pilot, Amazon deployed the system at Lumen Field, home of the Seattle Seahawks, where fans can grab Seahawks gear from the Pro Shop Outlet and walk out without waiting in line . This deployment demonstrated the versatility of the technology beyond convenience stores and grocery stores. |
3.3 Scale and Impact |
By 2025, Just Walk Out technology had been deployed in more than 150 locations, including Amazon's own stores and third-party retailers across the United States, the United Kingdom, and Australia . More than 70 company-owned stores and over 85 third-party stores use the system, with new locations launching every month . These third-party locations are concentrated in airports, sports stadiums, convention centers, theme parks, and convenience stores---environments where speed is valued and customers are less likely to have loyalty to a specific brand. |
The impact on customer behavior has been significant. At Lumen Field, the first Just Walk Out store (without RFID) saw a 60% increase in customer throughput and doubled transactions per game compared to the traditional concession stand it replaced. By the end of the 2023 season, transactions per game had increased 85% and total sales per game increased 112% . This demonstrates that the convenience of skipping the checkout line translates directly into higher sales. |
Over a one-year period from June 2024 to May 2025, more than 36.7 million items were sold through Just Walk Out technology-enabled stores . This is a testament to both the scale of deployment and the growing consumer acceptance of cashierless shopping. |
3.4 The Strategic Shift: From Grocery to Third-Party |
In 2024, Amazon made a strategic decision to remove Just Walk Out technology from its Amazon Fresh grocery stores in the United States, replacing it with Dash Carts (smart shopping carts) . Amazon spokesperson Jessica Martin explained that while customers enjoyed the benefit of skipping the checkout line, they also wanted the ability to easily find nearby products and deals, view their receipt as they shopped, and know how much money they saved . |
This was not a failure of the technology, but rather a recognition that different retail formats have different customer needs. Just Walk Out continues to be offered to third-party retailers, where it is thriving in high-traffic environments like stadiums and airports . In these settings, speed and convenience are the paramount concerns, and the lack of a traditional receipt or in-app shopping list is less of an issue. |
Rajiv Chopra, Vice President of Amazon Just Walk Out, noted that the technology is being deployed in an expanding range of environments, including hospitals, universities, and corporate campuses. He described the system as 'AI at the edge,' with custom edge compute devices optimized for running vision algorithms, backed by cloud processing . This architecture allows the system to operate with high accuracy even in chaotic, crowded environments with many people moving in and out. |
|
4. Walmart and Sam's Club: The Hybrid Approach |
While Amazon has focused on fully automated stores, Walmart and its warehouse club subsidiary Sam's Club have taken a different path, developing a hybrid model that combines mobile apps, computer vision, and AI-powered verification. |
4.1 Sam's Club: The Register-Free Future |
Sam's Club, a division of Walmart, is making one of the most aggressive bets on cashierless technology in retail. In 2024, the company announced that it was opening a new store in Dallas that would be completely free of checkout lanes . CEO Chris Nicholas described it as 'a fully digital experience,' where every customer is expected to use the Scan & Go app . |
Scan & Go is a mobile application that allows Sam's Club members to scan items with their phone as they shop. The app builds a digital cart, and when the customer is finished, they pay through the app. Rather than having a traditional checkout lane, the Dallas store features AI-powered exit arches equipped with computer vision technology . |
Here is how the exit works: As a shopper walks through the arch, multiple 4K cameras capture images of everything in the cart. The computer vision system compares these images to the digital receipt generated by the Scan & Go app. If everything matches, the shopper is cleared to leave. There is no need for an employee to check receipts or count items . The system is designed to be privacy-conscious: it does not use facial recognition or biometrics. It simply looks at items in the basket and compares them to outstanding receipts . |
Sam's Club has already rolled out Scan & Go to more than 120 stores, with plans to expand to all 600 locations in the coming years . Early results show exit times cut by nearly a quarter, and shoppers have reported positive experiences . Nicholas called it 'one of the fastest, most scalable transformations happening in retail today' . The company has tied Scan & Go usage to higher visit frequency, higher spending, and stronger membership renewal rates . |
4.2 Balancing Automation with Accessibility |
Sam's Club's aggressive shift to a mobile-first model is not without challenges. There is a meaningful minority of members who either do not own a modern smartphone, do not install retail apps, or are not comfortable managing payments on a phone . To address this, Sam's Club has indicated that store associates with mobile devices can scan and check out members who do not use Scan & Go, effectively turning staff tablets into roving point-of-sale terminals . |
Privacy is another concern. The app ties purchases, visit patterns, and in-store behavior to a named membership account, and the computer vision system captures images of carts and movement through the exit. Sam's Club has explicitly stated that the exit system does not use facial recognition, and the company's privacy notice allows for biometric data only in specific use cases like virtual try-on features . |
4.3 Walmart's Hybrid Strategy |
Walmart, Sam's Club's parent company, has taken a more cautious and flexible approach. The company continues to invest in self-checkout kiosks but is also selectively reducing or removing self-checkout in some stores where shrink and congestion are problematic . Walmart is one of the most active experimenters with AI-enhanced kiosks, computer vision, and mobile Scan & Go, including pilots that blend traditional barcode scanning with cameras and sensors for real-time validation . |
Walmart's strategy is to create a 'hybrid front end' where self-checkout kiosks, staffed lanes, and mobile self-scanning coexist, tuned store by store based on risk and demographics. This allows the company to balance the benefits of automation with the need for human oversight and the realities of different customer bases. |
|
5. Chinese Innovators: JD.com's Unmanned Stores |
In China, JD.com has been a pioneer in cashierless retail, developing its own fully integrated solutions that combine RFID, facial recognition, and image recognition. |
5.1 JD.com's Two-Model Approach |
In 2017, JD.com unveiled two models of unmanned stores, ahead of China's Singles' Day shopping festival . This demonstrated JD's commitment to transforming offline retail through technology. |
The first model, the JD Unmanned Convenience Store, was a full-service solution integrating multiple smart technologies. It used facial recognition for entry and payment, so customers did not have to wait in line at checkout. Cameras on the ceilings recognized customers' movements and generated heat maps to track traffic flow, product selection, and customer preferences, helping store owners stock efficiently . |
The second model, the D-Mart Smart Store Solution, was a low-cost option designed for existing store owners to upgrade their operations. It included smart shelving with JD Smart Vision technology that could recognize products and in-store behavior, a smart counter for product recognition and checkout, and a smart advertisement screen that used facial recognition to show customized ads based on shopping habits and demographics . This modular approach made the technology accessible to a wider range of retailers. |
5.2 The Technology Behind JD's Unmanned Stores |
JD's unmanned stores relied on a multi-layered technology stack : |
Facial Recognition for Entry: Customers linked their face to a payment method on first entry. At subsequent visits, the system recognized them automatically, allowing them to enter and shop without any additional authentication. |
RFID for Product Tracking: Every product in the store was tagged with an RFID tag. When a customer picked up an item, the RFID system tracked it, and when the customer passed through the exit gate, RFID readers identified all items being taken. |
In-Store Cameras for Tracking: Ceiling cameras tracked customer movement, analyzed where shoppers spent time, and generated heat maps of store activity. This data helped store owners optimize product placement and inventory. |
Image Recognition for Verification: The system used computer vision to verify that the items a customer was taking matched the items in their virtual cart, reducing errors. |
The shopping experience was designed to be seamless. Customers entered by facial recognition, picked up items freely, and walked through a settlement channel where RFID automatically identified the products and charged the account . No phone or wallet needed to be taken out after the initial setup. |
JD's unmanned stores were initially tested with the company's 10,000+ employees at its headquarters in Beijing, before being rolled out more broadly . The company also announced plans to open the world's first fully automated Business-to-Consumer warehouse, demonstrating the breadth of its automation ambitions . |
5.3 The Broader Chinese Context |
JD was not alone in China's cashierless retail boom. The country's widespread adoption of mobile payments (via Alipay and WeChat Pay) and the high comfort level of Chinese consumers with technology created fertile ground for innovation. However, as later analysis noted, early unmanned stores also faced challenges around limited space, insufficient product variety, and technical issues that became apparent when multiple customers entered simultaneously . |
JD's response was to emphasize its complete solution and modular approach, making the technology adaptable to different store sizes and formats. By offering both a full-service solution and a low-cost upgrade path, JD positioned itself as a technology provider for the entire retail ecosystem. |
|
6. Challenges and Considerations |
Despite the excitement and investment in cashierless retail, significant challenges remain. These are not just technical problems but also human, economic, and regulatory issues. |
6.1 Theft and Shrinkage |
The risk of theft---both accidental and intentional---is a major concern for retailers adopting cashierless systems. Studies have shown that in Scan & Go models, shoppers with 50 items have a 60% chance of missing at least one item during scanning, and those with 100 items have an 86% chance . While AI-powered verification can catch some errors, the system cannot prevent a shopper from intentionally leaving without scanning an item. |
Retailers are exploring various approaches to manage this risk. Sam's Club's AI exit arches provide real-time verification, and the company has indicated that early results show success in reducing shrinkage. Walmart has experimented with TruScan technology that can identify unscanned items and repeat offenders . However, as one analysis noted, retailers who use Scan & Go face an increased risk of product loss, and false accusations of theft can damage customer trust . |
6.2 Cost and Accessibility |
Building a cashierless store requires substantial upfront investment. The cost includes cameras, sensors, RFID tags, computing hardware, software development, and ongoing AI model training. For smaller retailers, these costs can be prohibitive . |
Accessibility is another concern. Not all customers own smartphones, are comfortable with mobile payments, or want to download retail apps. Even in advanced markets, a meaningful minority of shoppers may be excluded from a fully mobile-first experience . Retailers must find ways to serve these customers without making them feel like second-class shoppers. |
6.3 Privacy Concerns |
Cashierless stores are inherently camera-heavy environments. The same sensors that enable automatic checkout can also collect detailed data about shopping behavior, movement patterns, and even physical characteristics. While companies like Sam's Club have stated that their exit arches do not use facial recognition, privacy advocates have raised concerns about how long video and sensor data are retained, how they are linked to customer profiles, and whether they might be repurposed for more intrusive uses . |
As one analysis noted, 'even without biometrics, this is still a dense data environment: the app ties purchases, visit patterns, and in-store behavior to a named membership account, and the computer vision system records carts and movement through the exit' . Ensuring transparency and clear limits on data use and retention will be essential for maintaining consumer trust. |
6.4 The Role of the Human Worker |
A common concern about cashierless stores is that they will eliminate jobs. While the technology does reduce the need for cashiers, it creates new roles in store operations, maintenance, and customer service. Sam's Club CEO Chris Nicholas has argued that the technology takes 'mundane tasks' out of associates' workload, allowing them to focus on higher-value activities like helping customers . However, the transition requires investment in training and support for displaced workers. |
|
7. The Future of Cashierless Retail |
The technologies powering cashierless stores continue to evolve, and the business models are maturing. What does the future hold |
7.1 Hybrid Models Will Dominate |
The most successful retailers will likely adopt hybrid models that combine elements of full automation, mobile apps, and traditional checkout, tailored to their specific format and customer base . Walmart's 'configuration as a variable' strategy, where self-checkout kiosks, staffed lanes, and mobile self-scanning coexist with adjustments store by store, is likely to become the norm . |
7.2 Expansion to New Verticals |
As the technology becomes more affordable and reliable, cashierless systems will expand beyond convenience stores and grocery stores. Hospitals, university campuses, corporate offices, and hospitality venues are all potential use cases . Amazon has already noted that Just Walk Out is being deployed in hospital cafeterias where medical staff and family members need quick access to food . |
7.3 The Continued Role of RFID |
RFID is likely to become an increasingly important component of cashierless systems. Its ability to handle apparel, soft goods, and items that are difficult for cameras to track makes it a valuable complement to computer vision. The fusing of computer vision, RFID, and weight sensors has already demonstrated higher accuracy than any single technology alone . |
7.4 AI at the Edge |
As AWS Vice President Rajiv Chopra noted, cashierless retail is a prime example of 'AI at the edge.' Edge computing allows for real-time processing of video and sensor data without the latency of cloud transmission . This makes the systems more responsive and reliable, even in crowded environments. As edge AI hardware becomes more powerful and cost-effective, we can expect to see more retailers deploying their own on-premise systems. |
|
8. Conclusion |
AI-powered cashierless checkout systems represent one of the most significant innovations in retail since the introduction of the barcode. By combining RFID, computer vision, sensor fusion, and deep learning, these systems deliver a shopping experience that is faster, more convenient, and less friction-filled than anything that came before. |
Amazon's Just Walk Out technology, which has processed over 36.7 million items in a single year, stands as a testament to the viability of the concept . The company's strategic addition of RFID to handle apparel and its expansion to third-party retailers in stadiums and airports demonstrate the technology's versatility and its ability to create value in high-throughput environments . Sam's Club's bold bet on a register-free, Scan & Go future, supported by AI-powered exit arches, suggests that cashierless retail is moving from niche innovation to mainstream expectation . Meanwhile, JD.com's early investments in unmanned stores with facial recognition and RFID show that the ambition to create frictionless retail is a global phenomenon, not limited to any single market . |
However, the path forward is not without obstacles. Cost remains a barrier for smaller retailers . Privacy concerns around camera-heavy environments and data collection need to be addressed transparently . The risk of theft and the challenge of serving customers who are not comfortable with mobile technology require careful management . And the human dimension---both the workers whose roles are changing and the customers who must adapt to new norms---cannot be ignored. |
The future of cashierless retail is not about eliminating the human element entirely. It is about using AI and automation to handle the mundane, repetitive tasks of checkout, freeing up both customers and employees for more valuable interactions. It is about offering choice: the choice to scan and go, the choice to use a smart cart, the choice to interact with a human cashier. The most successful retailers will be those that balance the power of AI with empathy for their customers and their workforce. |
The checkout line, that perennial frustration of the shopping experience, is on its way to becoming a relic of the past. The question is not whether cashierless retail will become the norm, but how quickly---and how thoughtfully---retailers will navigate the transition. |
cs@easiersoft.com