The gesture itself is the evidence, not the person.
Each year, a quiet and largely invisible tax falls on the retail world: roughly one and a half percent of inventory simply walks out the door. Veesion, a Paris-based startup founded by two French engineers, has built an artificial intelligence system that watches for the physical grammar of theft — the pocket, the conceal, the stroller — without ever needing to know a face. Deployed across more than 350 Spanish stores since early 2022, the technology represents a broader human negotiation between surveillance and privacy, between loss and accountability.
- Supermarkets silently absorb a 1.5% annual inventory drain, and the problem compounds because eight in ten undetected shoplifters return to steal again.
- Veesion's gesture-recognition AI flags suspicious movements in real time — pocketing items, hiding goods in strollers, consuming products without paying — without relying on facial recognition.
- The system integrates directly into existing store camera infrastructure, generating ten to fifteen theft alerts per week in newly installed locations, though only 65–70% of those alerts reflect genuine theft.
- Despite the accuracy margin, pilot stores have cut unaccounted losses by over 60%, a dramatic improvement over an industry where normally only 5% of theft is ever caught.
- Having already secured €10 million in first-round funding and surpassed €1 million in 2021 revenue, Veesion is now preparing a second funding round with ambitions to triple its customer base and enter the US market.
Shoplifting costs supermarkets roughly one and a half percent of their inventory each year — a slow, invisible drain. What compounds the problem is impunity: about eighty percent of undetected thieves return to steal again. This is the wound that Veesion, a Paris-based startup founded in 2018, set out to close.
The company's solution watches security camera feeds in real time, using AI to recognize the physical gestures that signal theft — the reach, the conceal, the walk. It doesn't use facial recognition, which is restricted in Europe, and makes no attempt to identify individuals. It simply reads movement. When a customer pockets an item, hides goods in a stroller, or consumes a product without paying, the system sends an alert.
After early success with major French retailers like Carrefour and Intermarché, Veesion expanded across Europe and arrived in Spain at the start of 2022. Within months, the software was running in more than 350 stores — supermarkets, pharmacies, cosmetics shops, and jewelers. Installation is simple: a software package connects to the store's router and links to existing cameras.
New installations typically generate ten to fifteen alerts per week, with a genuine theft rate of 65–70%. Even accounting for false positives, stores have reduced unaccounted losses by more than sixty percent — a significant leap in an industry where only about five percent of theft is normally caught.
Having generated over one million euros in revenue in 2021 and secured a first funding round of ten million euros, Veesion is now preparing a second round with its sights on the United States — aiming to triple its customer base and carry a French engineering project into a global market.
Shoplifting costs supermarkets roughly one and a half percent of their inventory each year—a steady drain that most customers never see. What makes the problem worse is that most of it goes undetected. Of the thieves who slip out unnoticed, about eighty percent will come back to steal again. This is the problem that Veesion, a Paris-based startup founded in 2018 by two French engineers, set out to solve.
The company's answer is a system that watches security camera feeds in real time, using artificial intelligence to recognize the physical gestures that signal theft. When a customer pockets an item, or slips something into a stroller, or consumes a product while still in the store without paying, the software flags it. The technology doesn't rely on facial recognition—which is restricted in Europe—and doesn't attempt to identify who the person is. It simply watches for the movements themselves: the reach, the conceal, the walk.
Veesion spent roughly three years developing the product before launching it in France, where it found early success with major retailers like Carrefour and Intermarché. From there, the company expanded into Germany, Belgium, the Netherlands, and Portugal. Spain came next, and the startup arrived at the beginning of 2022. Within the first few months, they had installed their software in more than three hundred fifty stores across the country—supermarkets mostly, but also cosmetics shops, pharmacies, and jewelry stores.
The installation is straightforward. Veesion connects a software package to the store's internet router, which then links to the existing security cameras. When the algorithm detects suspicious movements, it sends an alert. Pablo Blanco, who runs Veesion's operations in Spain and Latin America, explains that the system catches classic theft patterns: items hidden in pockets, goods placed in baby carriages, merchandise tucked into fanny packs, or customers who make a legitimate hundred-euro purchase while also stealing something else.
In the first weeks after installation, stores typically see between ten and fifteen theft alerts per week. Not all of them are real—Blanco notes that between sixty-five and seventy percent of the software's alerts actually represent genuine theft attempts. But even with that margin of error, the impact is significant. Stores using the technology have reduced their unaccounted losses by more than sixty percent. For retail operations where only about five percent of theft is normally caught, and where roughly half of that is committed by repeat customers, this represents a meaningful shift in loss prevention.
The company generated more than one million euros in revenue during 2021. With a first funding round of ten million euros already secured, Veesion used that capital to push into new markets. Now they're preparing a second round of financing, with their sights set on the United States. The ambition for 2022 is to triple their customer base, turning a French engineering project into a global retail security standard.
Notable Quotes
The software doesn't identify people—it only analyzes their gestures. Facial and biometric recognition is prohibited in Europe.— Pablo Blanco, Veesion country manager for Spain and Latin America
In the first weeks after installation, stores typically catch between 10 and 15 thefts per week.— Pablo Blanco, Veesion
The Hearth Conversation Another angle on the story
Why does a store need AI to catch shoplifting? Don't they already have security guards and cameras?
They do, but the problem is scale and attention. A guard can only watch one section at a time. A camera records everything, but nobody sits and watches hours of footage looking for subtle hand movements. The AI does that automatically, in real time, across every camera feed at once.
So it's watching for specific movements—like reaching into a pocket?
Exactly. The algorithm learned what theft looks like by studying thousands of hours of actual shoplifting. It recognizes the pattern of the reach, the pause, the concealment. It's not trying to identify the person; it's identifying the gesture.
What about false alarms? If it flags someone adjusting their jacket, wouldn't that create chaos?
That's the real engineering challenge. They report about sixty-five to seventy percent accuracy, which means roughly one in three alerts is a false positive. But even with that noise, stores are cutting their losses by more than half. The alternative—catching almost nothing—is worse.
Does it matter if the person is a regular customer or a first-time thief?
The software doesn't know the difference. It just sees the gesture. But interestingly, the data shows that repeat customers account for about half of all detected theft. So the stores are catching people they see all the time.
Why expand to America if Europe is working?
Europe is a proof of concept. The U.S. market is vastly larger, and American retailers lose billions to shrinkage every year. If Veesion can solve this problem at scale in the States, that's where the real revenue is.