American retailers are staring down a theft and fraud problem measured in the tens of billions of dollars, and the old mix of guards, cameras, and locked cases is no longer keeping up. Now Amazon is pitching a new generation of AI-driven systems that it says can help crack a $100 billion shoplifting and “shrink” crisis as major chains prepare to roll out the technology in 5,000 stores across the United States. The stakes are simple and stark: if this wave of automation works, it could reset the economics of brick-and-mortar retail, and if it fails, the industry’s most painful losses will keep climbing.
At the center of this shift is a bet that smarter software can see what human staff miss, from subtle sleight-of-hand at self-checkout to organized theft rings testing store defenses. I see Amazon’s push as both a defensive move to protect its own formats and a bid to become the default security brain for American retail, just as its cloud arm already powers much of the internet. The question now is not whether stores will adopt AI, but how far they are willing to let algorithms police the aisles.
The $100B theft wave that forced retailers to rethink security
Retail crime has quietly become one of the most destabilizing forces in physical commerce, with losses that rival the profit margins of entire chains. Industry data shows that Shrink, the catch-all term for theft, fraud, and inventory errors, accounted for more than $112 billion in losses for retailers in 2022, a figure that helps explain why executives now talk about shoplifting in the same breath as labor costs and rent. When a single line item grows that large, it stops being a nuisance and becomes an existential threat, especially for grocers and big-box stores that operate on thin margins.
Against that backdrop, Amazon is positioning its new systems as a direct response to what one report describes as a $100 billion theft crisis that has pushed American retailers to look beyond traditional cameras and guards. The same reporting notes that major chains are preparing to install Amazon’s technology in 5,000 locations, a scale that signals this is not a pilot or a gadget but a structural change in how stores will be monitored. I see that number as a tipping point: once thousands of outlets are wired into a shared AI security layer, the industry’s baseline expectations for loss prevention will shift almost overnight.
Why 5,000 American stores are betting on Amazon’s new system
Retailers do not retrofit 5,000 stores on a whim, especially when it means rewiring checkout flows and retraining staff. The decision by American chains to adopt Amazon’s system at that scale reflects a belief that incremental tweaks to existing cameras and guards will not be enough to tame a problem measured in tens of billions of dollars. By committing to thousands of deployments at once, these companies are effectively standardizing on Amazon as a core security vendor, much as they once standardized on barcode scanners and point-of-sale terminals.
The reporting on Amazon’s push to help crack a 100B theft crisis makes clear that the company is not just selling hardware, it is offering a full stack of AI, sensors, and cloud analytics that promises to spot suspicious behavior in real time. For retailers, the appeal is twofold: they get access to Amazon’s scale and data science without having to build their own teams, and they can roll out a consistent security posture across thousands of sites instead of stitching together local solutions. In my view, that kind of uniformity is exactly what organized theft rings have lacked a counterweight to, and it is why this wave of installations could reshape how shoplifters assess risk.
From Just Walk Out to a new AI playbook
Amazon’s latest security push does not come out of nowhere, it builds on years of experimentation with frictionless checkout and computer vision in its own stores. The company’s Just Walk Out technology was marketed as a way to let shoppers grab items and leave without stopping at a register, relying on a dense mesh of cameras and sensors to track what went into each basket. That system promised to Increase customer throughput and Speed up checkout, and Amazon has long highlighted those Benefits as a way to boost sales in formats like cafes and campus dining halls.
Yet the story has been more complicated inside Amazon’s own walls. Earlier this year, the company began removing Just Walk Out from its Fresh grocery stores, a shift that NPR reported as a move toward more traditional self-checkout and cart-based tracking in Fresh locations. Separate coverage noted that the technology, which first rolled out in Amazon Go and Amazon Fresh stores, has drawn scrutiny as accusations are flying in media about its cost and complexity, a tension captured in Feb reports on Amazon stores removing their own Just Walk Out systems. I read that pivot not as a retreat from AI, but as a recalibration: the same sensor and vision stack that once tried to replace checkout entirely is now being repurposed to watch over it.
How AI is becoming retail’s new loss-prevention partner
What has changed in the past few years is not retailers’ desire to stop theft, but the tools available to do it. Artificial intelligence has moved from a buzzword to a practical engine for scanning video feeds, transaction logs, and sensor data at a scale no human team could match. In loss prevention, that means algorithms can flag patterns like items consistently leaving shelves without corresponding scans, or customers who repeatedly abandon full carts when approached by staff, long before those behaviors show up in end-of-year shrink reports.
Industry analysis describes The Rise of AI as a new kind of Powered Security partner for stores, with Artificial intelligence taking on the grunt work of monitoring cameras and alerts without burning out human staff or straining budgets. Separate research on loss prevention notes that AI at the checkout has already been shown to cut losses by 50% or more, and that when similar tools are extended across the store, overall shrink can be reduced by up to 45%, according to a Jan analysis. Those figures help explain why retailers are willing to endure the disruption of new systems: if the technology performs anywhere near those benchmarks, it pays for itself quickly.
Inside Amazon’s AI security stack, from AWS to the sales floor
Behind the cameras and sensors that shoppers can see, Amazon’s security ambitions are increasingly rooted in its cloud arm. Amazon Web Services has been rolling out a suite of AI-enhanced security innovations that promise to ingest vast streams of signals, from video to transaction logs, and surface anomalies in real time. At its Invent conference, AWS highlighted new tools for analyzing security events and orchestrating automated responses, positioning the cloud as the central nervous system for both digital and physical defenses.
One AWS Security Blog post details how AWS is launching AI-enhanced security innovations at re:Invent 2025, with a focus on correlating diverse signals and securing agentic access across complex environments. Separate coverage of re:Invent notes that Amazon Web Services used the event to unveil new products and capabilities, with Amazon Web Services and other vendors emphasizing AI-driven detection, exposure summaries, and customizable widgets that can be embedded into security dashboards. When I connect those dots to the retail deployments, the picture that emerges is of store cameras and sensors feeding directly into AWS, where machine learning models trained on global data can help local managers decide when to intervene.
What happens at the checkout when AI watches every move
The front line of this transformation is the checkout, where the tension between convenience and control is most visible. Self-checkout lanes and mobile scan-and-go apps have given shoppers more autonomy, but they have also opened new avenues for theft, from “banana trick” barcode swaps to simply skipping items. AI systems now being deployed in stores are designed to watch those interactions closely, comparing what the camera sees with what the register records and flagging discrepancies in real time.
Research on loss prevention AI points out that we already know AI at the checkout cuts losses by 50% or more, a claim that underpins the business case for these deployments in the Jan report. At the same time, the broader shift toward grab-and-go formats has been shaped by the same forces, with one analysis noting that Shrink accounted for more than $112 billion in losses even as retailers experimented with new checkout models, a tension explored in detail in the grab-and-go discussion. I see Amazon’s latest systems as an attempt to square that circle: keep the speed and autonomy that shoppers like, but quietly layer in a digital security guard that never blinks.
Bias, privacy, and the promise that “THE ALGORITHM DOESN’T CARE”
Any time retailers point more cameras at customers, questions about bias and privacy follow quickly. Civil liberties advocates have warned that AI systems trained on skewed data can unfairly target certain groups, and that constant surveillance can chill legitimate behavior as well as crime. Retailers adopting Amazon’s tools know they will have to answer not just whether the systems work, but whether they work fairly and transparently enough to maintain public trust.
One vendor already active in this space has tried to get ahead of those concerns by stressing that “THE ALGORITHM DOESN’T CARE WHAT PEOPLE LOOK LIKE,” arguing that its system focuses on identifying suspicious movements rather than profiling individuals, a claim highlighted in Nov coverage of stores adding AI to security systems. That framing underscores the research methods behind these tools, which rely on patterns of behavior rather than demographic traits, but it does not fully resolve the privacy debate. In my view, the real test will be whether retailers can explain, in plain language, what is being recorded, how long it is stored, and how customers can challenge mistakes when the ALGORITHM flags them incorrectly.
Amazon’s influence, from AWS to the broader business climate
Amazon’s move into large scale retail security is happening against a backdrop of broader economic and political volatility that shapes how companies think about risk. Markets are whipsawed by everything from lottery jackpots to courtroom drama, with headlines about a soaring Powerball prize sitting alongside coverage of Trump’s economic agenda and the latest corporate earnings. For retailers, that noise reinforces a simple lesson: they cannot control macro shocks, but they can try to control what happens inside their own four walls.
Within that context, Amazon’s dual role as a retailer and as the operator of AWS gives it unusual leverage. The same cloud infrastructure that secures banks and media companies is now being pitched as the backbone for in-store surveillance and loss prevention, with Lise Feng describing new AI-enhanced offerings for Customer Solutions and Sec teams. I see that convergence as a strategic bet: if Amazon can become the default security layer for both digital and physical assets, it not only locks in cloud customers, it also shapes the standards by which future retail systems are judged.
Will AI security fix retail theft or just change its shape?
The open question is whether this wave of AI deployments will meaningfully reduce theft or simply push it into new forms. History suggests that when one avenue of fraud is closed, determined actors look for another, and organized retail crime groups are already adept at probing for weak spots. If cameras and algorithms make self-checkout riskier, some thieves may pivot to return fraud, online scams, or targeting stores that lag behind in adopting new tools.
Still, the scale of the current investment signals that retailers believe the balance of power can shift. With 5,000 American stores preparing to install Amazon’s systems to confront a 100B theft problem, and with AI tools already shown to cut checkout losses by 50% or more, the industry is clearly betting that smarter surveillance will at least narrow the gap. From my vantage point, the most likely outcome is not a clean victory over shoplifting but a new equilibrium, one in which algorithms quietly patrol the aisles, honest customers move faster through the store, and would-be thieves find that the easy opportunities of the past few years have started to dry up.
More From TheDailyOverview
Grant Mercer covers market dynamics, business trends, and the economic forces driving growth across industries. His analysis connects macro movements with real-world implications for investors, entrepreneurs, and professionals. Through his work at The Daily Overview, Grant helps readers understand how markets function and where opportunities may emerge.
