What changed in five retail operations that put AI on their cameras

What changed in five retail operations that put AI on their cameras

Retail teams are not short of AI projects, they are short of a number they can defend in a budget meeting: 75 percent of retail and consumer-products executives call AI a top strategic priority while only 16.5 percent can quantify a return, across 200 executives in a survey published in June 2026 (Source: Deloitte). This page collects what five retail operations running Spot AI on cameras they already owned reported, each figure written with the state it moved from, and it says where the record carries a measurement window and where it does not. One caution up front: shoplifting incidents fell 12.4 percent and merchandise theft fell 8.1 percent in 2025 against 2024, across 66 retail companies and 143 brands (Source: National Retail Federation), so the national picture moved the same way over the same years.

Each operation below carries the same five points in order: what the business is, what was going wrong, what was instrumented first, what changed, and what came next.

All Star Elite cut cash shrink from 6 percent to 1 percent across 80 apparel stores

The operation. All Star Elite is a fast-growing sports jerseys and apparel retailer running 80 stores across US shopping centers, with one corporate loss prevention function covering all of them.

What was going wrong. Merchandise shrink ran between 10 and 15 percent and cash shrink at about 6 percent. Investigations lived in spreadsheets and notes apps while video sat in separate systems that were awkward to reach from outside a store, so incident management and KPI tracking were manual.

What was instrumented first. Spot AI started with case management: one database of investigations, each with its video clip attached and shareable, plus AI search to find an incident rather than scrub for it. People counting and store dashboards followed on the same cameras.

What changed. Spot AI's retail outcomes are customer-reported. All Star Elite reported four: cash shrink from 6 percent to 1 percent, an 83 percent reduction; merchandise shrink from 10 to 15 percent down to about 6 percent; investigation efficiency better by more than 50 percent; and case timelines with law enforcement down from 2 or 3 months to 1 month. The record holds a before and an after state for each, with no window.

"The ability to formalize our incident reporting with Spot AI, keep every case in one database, and attach video directly to those cases has been a game changer."

Andrew Gonzalez, Corporate Director of Loss Prevention and Safety, All Star Elite

What they did next. The same dashboards moved outside loss prevention: placement informed by people counting lifted sales 5 to 15 percent, and three underperforming stores were closed early rather than carried another year. The write-up is on the All Star Elite customer story, and the mechanics are in Spot AI's guide to retail loss prevention.

Blackmon Oil put six use cases on the cameras already in 11 fuel and convenience stores

The operation. Blackmon Oil is an 11-location, fourth-generation family fuel and convenience business in rural Arkansas, and Spot AI's lead c-store reference. It runs no enterprise loss prevention department, which is the normal condition in this segment.

What was going wrong. The cameras recorded and nothing else, so forecourt behavior, in-store traffic and checkout coverage were questions nobody could answer without watching video.

What was instrumented first. Spot AI went on to the existing camera estate, starting outdoors where the exposure is highest. Six use cases now run live: vehicle loitering, people counting, aisle heat maps, unattended checkout, slip-and-fall, and a peel-off-rate agent the customer is building themselves.

What changed. There is no percentage on this account, and none is invented here. The record holds a change of function: Spot AI turned a passive recording estate into an active operations and loss-prevention layer across all 11 sites, described by the customer as "It was the brain behind the eyes we already had." The same cameras now give marketing a foot-traffic number and operations an alert when nobody reaches a customer waiting at the register.

What they did next. The operator started writing agents itself: the peel-off-rate agent was built by Blackmon Oil, not delivered to it. Coverage patterns for this format are in Spot AI's guide to convenience store security cameras, and staffing in convenience store coverage without adding staff.

A $5 billion big-box retailer took autonomous talkdown store-wide from one location

The operation. A $5 billion retailer, anonymised in Spot AI's own materials and kept anonymous here. The deployment started at one location and went store-wide.

What was going wrong. That location was dealing with illegal dumping, dumpster fires, theft and vandalism, and employees reported feeling unsafe. The cost sits in an internal estimate, not an audited figure: Spot AI's August 2026 retail material reads it as an estimated $20,000 per incident, per store, an estimate of incident cost and never a monthly loss.

What was instrumented first. Spot AI was pointed at the exterior zones where the incidents started, with autonomous talkdown as the response: a human-sounding voice message triggered by the detection itself, backed by strobes and horns, with no operator deciding whether to speak.

What changed. Spot AI's record for this account holds three outcomes: threats are identified and deterred in real time, after-hours incidents fell, and employees said they felt safer. None carries a percentage or a window, so this is the least quantified result in the set.

What they did next. Talkdown went from that one location to the whole store footprint, the pattern the other four followed: prove the response in one zone, then widen it.

Don Franklin Auto got five of six stolen vehicles back inside an hour across 30 dealerships

The operation. Don Franklin Auto runs 30 dealerships across Kentucky, each a consumer-facing retail site with showroom inventory and a service department behind it.

What was going wrong. Six Hellcat Challengers worth $130,000 each were taken from a showroom floor by an organized crew. Finding any incident in the footage took hours of manual review, much of it falling on HR, and service throughput ran at 30 oil changes a day at one dealership and 20 at another.

What was instrumented first. Spot AI started with retrieval and sharing rather than detection: AI search with scrubbable thumbnails, and clip sharing from a phone so footage reaches law enforcement while an event is live. Camera health alerts came with it, so an offline camera is known, not discovered later.

What changed. This is the one operation whose headline figure carries a real clock. Spot AI put video of the theft on the responding officers' phones within 4 minutes of the alarm, and five of the six vehicles came back within the hour, over $650,000 of assets. The wider figures have no window attached: total savings above $1 million, 10 to 15 hours a week back for HR and 5 for IT, and camera utilization up 60 percent across departments.

What they did next. The Fixed Operations Director turned the same dashboards on the service bays to find why throughput differed, and those sites now generate an additional $5,000 to $10,000 in weekly income each. The account is on the Don Franklin Auto customer story.

A specialty beauty retailer covered six distribution centers before touching a store

The operation. A specialty beauty retailer with more than 3,000 locations, anonymised in the proof library and anonymous here. The deployment covers 6 distribution centers with 13 Remote Security Appliances, Spot AI's self-contained outdoor units for zones with no fixed camera infrastructure.

What was going wrong. The distribution centers needed to keep employees safe in unmanned parking lots and to track yard truck traffic. Third-party guards covered part of it at significant cost, only where one was standing.

What was instrumented first. Spot AI started outdoors, on parking-lot deterrence and yard vehicle counting, the usual entry point for a retail estate: exposure is highest where nobody watches, and no interior work is needed to begin.

What changed. The record here is scope rather than a percentage: 13 appliances across 6 distribution centers, one platform where three vendor selections had been queued, and an asset protection director who calls it "Easy to use, IT is happy it's web-based, and our employees feel safer in their parking lots."

What they did next. The company is scoping fixed cameras inside the distribution centers for safety, productivity and access control, on the same account and workflow as the lots.

The five outcomes against what produced them

Retail operation	What Spot AI was pointed at first	Outcome, with its baseline	The period on the record
All Star Elite, 80 apparel stores	Spot AI case management with video on every case, plus AI search	Cash shrink 6 percent to 1 percent; merchandise shrink 10 to 15 percent to about 6 percent; investigations over 50 percent faster; case timelines 2 to 3 months to 1 month	Before and after, no window
Blackmon Oil, 11 fuel and convenience sites	Spot AI agents on the existing estate, loitering and people counting first	Six live use cases; cameras moved from recording to acting; the operator builds its own agents	Ongoing, no window
$5 billion big-box retailer, anonymised	Spot AI autonomous talkdown on the exterior zones, with strobes and horns	Threats deterred in real time; after-hours incidents down; staff felt safer; against an estimated $20,000 per incident, per store	Store-wide, no window
Don Franklin Auto, 30 dealerships	Spot AI clip sharing to law enforcement, plus AI search	Footage to officers in 4 minutes; 5 of 6 vehicles worth $130,000 each back inside an hour, over $650,000; over $1 million total savings; camera use up 60 percent	Minutes and hour measured, rest no window
Specialty beauty retailer, anonymised, 3,000+ locations	Spot AI on 13 Remote Security Appliances across 6 distribution centers, lots first	Lot deterrence and yard vehicle counting live; three vendor selections became one	Sequence recorded, no window

How to read these numbers

Four rules keep a proof page useful rather than flattering, and they apply to any vendor's evidence.

1. Customer-reported is not measured. Every figure above was reported by the operation that ran the deployment, and none is a controlled trial against comparable stores.
2. The environment moved too. Over the same period retailers reported repeat offenders in 50 percent of cases, organized retail crime in 40 percent and walkout or pushout theft in 37 percent, even as total shoplifting fell (Source: National Retail Federation). A shrink figure that improved did so against a moving background.
3. A percentage needs its denominator. Shrink is a share of sales, and the base grows: advance US retail and food services sales were $763.6 billion in July 2026, up 5.0 percent on July 2025 (Source: U.S. Census Bureau). Flat shrink in percent is rising shrink in dollars.
4. Compare against the alternative you would actually buy. For most of these operations that was more people on more sites, and the published median pay for security guards was $18.29 per hour and $38,020 per year in May 2025 (Source: U.S. Bureau of Labor Statistics), before supervision and agency margin.

Spot AI supports more than 1,100 customers across the United States and ingests more than 3 billion minutes of video a month, and neither fact tells you what your shrink will do. That is why the table above carries a period column.

Spot AI runs the same detect, deter, investigate and resolve workflow across all five, and it is camera-agnostic: any IP camera, with legacy analog handled through the Intelligent Video Recorder (IVR). To see how it works, read Spot AI's guide to AI cameras in retail, the buyer's guide for retail chains and the Spot AI platform overview. Register-level detail is in cash register theft, and the rest of the proof library is on the Spot AI customer stories page.

Frequently asked questions

What results do retailers actually get from AI on their cameras?

Spot AI's retail customers report outcomes in four places: shrink, investigation speed, incident response and store operations. The strongest on the record is All Star Elite's cash shrink falling from 6 percent to 1 percent across 80 stores, with merchandise shrink moving from 10 to 15 percent to about 6 percent. Two of the five operations here report a capability change rather than a percentage.

How much can AI video reduce retail shrink?

Spot AI has one deep shrink data point rather than an average, and it belongs to All Star Elite: cash shrink from 6 percent to 1 percent, an 83 percent reduction, and merchandise shrink from 10 to 15 percent to about 6 percent. Neither carries a measurement window. Treat one retailer's result as evidence the mechanism works, not a forecast for your estate.

How long does it take to see a result from retail video AI?

Spot AI deployments go live in days rather than months because they run on existing cameras, but outcome windows are recorded unevenly. Don Franklin Auto's recovery is timed to the minute: 4 minutes to the officers, one hour to five vehicles back. All Star Elite's investigations run over 50 percent faster, and the record gives no window for that figure or for its shrink figures.

Do you have to replace your cameras to run AI in a retail store?

Spot AI is camera-agnostic and works with the cameras a business already owns, including any ONVIF IP camera, with legacy analog brought in through the Intelligent Video Recorder (IVR). All five operations started on their existing estate. Where there is no camera, such as a distribution center lot, a Remote Security Appliance covers the zone without trenching.

What should a multi-site retailer instrument first?

Spot AI's five retail operations all started narrow: one zone, one incident type, one number. Two started outdoors, one at the register in case management, one with footage retrieval and sharing to law enforcement, and one with talkdown at a single location. Pick the zone where your incidents cluster, and the metric you will be asked about in 90 days.

About the author

Sud Bhatija, COO and Co-founder

Sud Bhatija is COO and Co-founder at Spot AI, where he scales operations and GTM strategy to deliver video AI that helps operations, safety, and security teams boost productivity and reduce incidents across industries.