The woman walked into a clothing store on Poplar Avenue near Ridgeway, picked up a handbag from a display table, and tucked it under her jacket. She moved toward the exit. Before she reached the door, a loss prevention officer appeared from a back hallway and asked to speak with her.
The officer hadn’t been watching on a monitor. He’d been alerted by software. A camera above the display table had tracked the woman’s hand movements, recognized the concealment pattern, and sent a notification to a tablet behind the counter. The whole sequence took less than eight seconds.
That kind of technology used to cost six figures and lived in casinos. Now it’s showing up in strip malls.
The Quiet Rollout
Memphis retailers have been installing AI-enabled camera systems at a pace that would’ve seemed unlikely three years ago. The cost dropped. The accuracy improved. And organized retail crime gave store owners the push they needed to spend the money.
Industry analysts estimate that about 30% of major U.S. retailers adopted some form of AI-powered video analytics by mid-2024. The manufacturing and retail segment of the AI camera market is growing at roughly 19.5% annually, according to a report from Global Market Insights published in September. That growth is hitting Memphis.
Wolfchase Galleria, the largest enclosed mall in the metro area, upgraded its camera infrastructure earlier this year. The new system includes cameras with on-board processing that can flag unusual behavior patterns: someone loitering near an emergency exit for an extended period, a group moving through the mall in a formation consistent with organized theft rings, or a vehicle circling the parking lot repeatedly.
The Germantown retail corridor along Poplar has seen similar adoption, though on a smaller scale. Boutique owners there tend to install two or three AI cameras focused on point-of-sale areas and fitting rooms. The technology they’re buying comes from companies like Verkada, Avigilon (now part of Motorola Solutions), and Rhombus. Prices for a basic AI camera with analytics start around $400 per unit, with cloud storage and software subscriptions running another $100 to $200 monthly.
Oak Court Mall on Perkins has taken a different approach. Rather than individual stores making their own purchases, mall management rolled out a centralized system covering common areas, entrances, and the parking garage. The cameras there use license plate recognition to flag vehicles associated with previous incidents.
What These Cameras Actually Do
The term “AI camera” gets thrown around loosely. It helps to understand what the technology can and can’t do right now.
Object and behavior detection. Modern AI cameras can identify specific objects (a firearm, a crowbar, an unattended backpack) and specific behaviors (running, fighting, concealing merchandise). They do this through computer vision models trained on millions of labeled images. The accuracy varies. A well-lit retail floor with clear sight lines gives good results. A shadowy parking deck at 2 a.m. gives worse ones.
People counting and heat mapping. This is the least controversial application. Cameras track foot traffic patterns, showing retailers which aisles get the most visitors, which displays attract attention, and where bottlenecks form. Stores along the Poplar Avenue corridor between Highland and Perkins have used this data to adjust layouts and staffing schedules.
License plate recognition. Parking lot cameras can read plates and cross-reference them against databases of stolen vehicles or vehicles linked to prior shoplifting incidents. Several Memphis-area auto dealerships along Covington Pike adopted this after a string of thefts in 2023.
Facial recognition. This is where things get contentious. The technology exists. Some retailers use it. Most won’t talk about it publicly.
The Facial Recognition Debate
Memphis has a complicated history with facial recognition. MPD used Clearview AI’s database for investigations before the city council raised concerns about civil liberties in 2020. The Tyre Nichols case intensified scrutiny of police surveillance tools. Public trust is low.
Retail use is different from law enforcement use, legally speaking. No Tennessee state law currently prohibits private businesses from using facial recognition on their own property. A store owner in Memphis can install a system that scans every face walking through the door and checks it against a database of known shoplifters. Whether they should is another question.
Privacy advocates argue the technology disproportionately misidentifies people with darker skin tones. Studies from MIT and the National Institute of Standards and Technology have confirmed higher error rates for Black and female faces in some commercial systems. In a city that’s roughly 65% Black, that’s not an abstract concern.
Retailers respond that the technology has improved significantly since those studies were published. They also point out that organized retail crime cost American retailers an estimated $112 billion in 2022, according to the National Retail Federation. Stores are closing locations in high-theft areas. If technology can reduce losses enough to keep a store open, they argue, the community benefits.
The debate doesn’t have a clean resolution. And in Memphis, it’s mostly happening without public input, since the cameras go up in private businesses with no permitting or disclosure requirements.
What’s Driving Adoption Here
Three factors are pushing Memphis retailers toward AI surveillance faster than the national average.
First, the theft problem. Memphis saw a 12% increase in shoplifting incidents between 2022 and 2023, according to TBI data. Organized retail crime rings have hit big-box stores along Winchester Road and in the Wolfchase area repeatedly. Loss prevention teams at major chains are overwhelmed. Technology fills a gap that hiring alone can’t close, especially when trained LP officers are hard to find.
Second, insurance pressure. Several commercial insurance carriers have started offering premium discounts for businesses that install verified AI surveillance systems. A Midtown restaurant owner told me his insurance broker specifically recommended Verkada cameras during a policy renewal meeting in August. “He said it could knock 8% off my liability premium,” the owner said. “That pays for the cameras in about 18 months.”
Third, the cost curve finally makes sense for smaller businesses. Five years ago, a system with real-time analytics required dedicated servers and a six-figure installation budget. Today, cloud-based platforms process video on the camera itself or stream it to remote servers. A small retailer on South Main can install four AI cameras for under $3,000, with monthly costs comparable to a basic alarm monitoring contract.
The Privacy Gap
Tennessee has no state law specifically governing AI surveillance in retail settings. There’s no requirement to post signs informing customers that AI is analyzing their behavior. There’s no restriction on how long retailers can store the data. There’s no rule about who can access it or whether it can be shared with law enforcement without a warrant.
Memphis city code addresses security cameras in some contexts (banks, certain licensed businesses) and doesn’t mention artificial intelligence at all. The regulatory framework hasn’t kept up with the technology.
Compare that to Illinois, where the Biometric Information Privacy Act requires businesses to get written consent before collecting biometric data, including facial geometry. Or Portland, Oregon, which banned facial recognition by private entities in 2020. Memphis has nothing similar on the books, and no one on the city council has publicly proposed anything.
That gap means the rules are being set by the companies installing the technology and the retailers buying it. For now, the standards are whatever the market will bear.
What Comes Next
The technology will get cheaper and more capable. Cameras with on-board AI processing will become the default rather than the exception. Within two years, a basic security camera without analytics will be the budget option, not the norm.
Memphis retailers are already asking about predictive features: systems that don’t just flag incidents in progress and instead identify patterns that suggest an incident is likely. A group of three people entering a store within 30 seconds of each other, splitting up, and moving toward high-value merchandise? The system would tag that as a potential organized theft attempt before anything is taken.
Whether Memphis residents are comfortable with that level of surveillance is a question nobody has really asked them. The cameras are going up. The software is getting smarter. And the conversation about what’s acceptable is happening years behind the deployment.