Is 2025 the turning point for loss prevention AI?
20th January 2025

Jason Souloglou, CEO at SeeChange, says retail theft is at crisis point but a loss prevention strategy powered by Computer Vision and AI will reduce losses, enhance the shopper experience and improve operations. Hear from Jason.
Why is addressing theft now a top priority for grocery executives?
We have reached a crisis point with regard to losses and theft in grocery retail and it’s been fuelled by a number of factors – COVID, lifestyle changes and pressures on income.
One of the factors is the way the press has, for instance, reported against self-checkouts and demonized the tech. It’s given those, who were maybe on the edge of things, an excuse to behave the way they have in supermarkets by taking stuff and justifying it to themselves.
I came across a thread on a Reddit forum, where the conversation was about theft from supermarkets but from both sides of the equation – shoppers who were stealing and attendants who had to suffer it. What was really interesting was that most of the people that ended up stealing from checkouts ended up in that place accidentally: they were interacting with the self-checkout and something happened, which allowed them to get away with something they didn’t intend. And when they got away with it, it opened the door to look for future opportunities.
I think that’s given fuel to the thinking, ‘it’s not my fault if the technology is hard to use and annoying’.
Different retailers have distinct priorities in respect of retail crime too, but whether it’s tackling anti-social behavior, shrink at checkout or in aisle, theft has been promoted to be their top priority and they are rolling out aggressive programs to combat shrink and curb losses using artificial intelligence across checkouts but also CCTV feeds.
To what extent is 2025 the turning point for loss prevention AI?
Up until now, early-adopters have ‘dabbled’ with computer vision and AI for loss prevention, but now solutions are officially out of the lab and ready to roll out at scale. The reason for this is two fold: there’s increased trust and acceptance of the latest tech solutions and the go to market model has changed.
If you take self-checkouts, for example. Self-checkouts were invented so that shoppers who chose to go down that route could do so and the attendants could do other things. Up until now, that technology has been incomplete because it’s not really been intelligent enough to remove all the friction. But we have finally got to the point whereby, adding this extra level of intelligence with AI, you have a more intelligent process, results of which are highlighted below. This takes away the friction and provides a level of intelligence and automation to implement the original thinking around self-checkouts in the first place.
While there has been a backlash against self-checkouts with some retailers removing them from their stores, there is an increasing number of units out there and ECR reports anecdotal evidence that suggests that in some forms of retailing, as much as 80% of customer transactions in supermarkets might be processed through variants of SCO technologies.
Secondly, it’s no longer the case of new start-ups needing to convince tier one grocery retailers of the role of new tech in loss prevention. Rather, established trusted global vendors, such as Diebold Nixdorf, Toshiba and NCR, as well as security players like Mitie, understand how AI enables a complete solution for their customers.
On the checkout side, SeeChange partners with Diebold Nixdorf and on the security side, we partner with Mitie. Diebold Nixdorf has integrated Vision AI into its checkout solutions to deliver a frictionless experience. Mitie is partnering with SeeChange Vision AI to bring prioritized insight into physical security providing a safer environment for customers and employees.
The combination of Diebold Nixdorf and Mitie at our leading customers provides a perfect opportunity to bring these two aspects together in what we are calling a connected store. That means connecting different types of devices in a network, which are working together to prioritize real-time events to reduce losses, make the journey more frictionless for the customer and improve the overall operations for the store itself. This integrated approach is providing a single view of an entire store and realizing our original vision.
Is there a natural relationship in retail AI deployments with employees and customers?
How can AI help retailers measure and control their losses?
Retailers know they have losses but not necessarily where they are in-store. We know that the cost of stock theft is as high as €100m for individual retailers. Connecting feeds from sensors and devices into a computer vision or loss prevention AI platform enables retailers to monitor in the aisles, at the checkout and exit for a safer and more efficient store environment. But significantly, it enables retailers to answer that critical question of how much shrink is in the back of the store, in the aisle, at checkout and how much inventory is being walked out of the store.
We already know that AI at the checkout cuts losses by up to 50%, imagine the impact when extrapolated across the store?
There are numerous use cases to show the impact of solutions like SeeChange on shrink. Intermarche at their Toulon store, for example, upgraded their self-checkouts to be able to detect unscanned items and differentiate between intentional and unintentional fraud. For the retailer, often items are not scanned due to handling errors. One common error concerns avocados – as customers don’t know if they are a fruit or a vegetable. These errors cause losses for the store of up to 3% of all transactions but since self-checkouts were upgraded to leverage computer vision and AI, that figure is around 1% and significantly interventions are down.
Another example is fresh produce recognition technology which not only assists in correct item identification but speeds transaction time. Aldi Süd piloted the technology at self-checkouts in Cologne. Items were recognised four times faster, which means transactions per hour can be increased by up to 40%. Further, the tech means more shoppers with fruit and veg will be willing to use the self-checkouts and staff interventions could be reduced by up to 45%.
We’ve talked a lot about loss, but does computer vision and AI have other benefits?
In addition to curbing losses, we often talk about operational efficiency, as well as the positive impact these solutions have on employees.
For example, self-service can help compensate for labor shortages and with line busting. Queues can hamper sales with research showing 75% of physical shoppers sometimes leave a line before it’s their turn. Addressing labor shortages, self-service checkouts are ‘always on’, unlike staffed checkouts so they improve overall efficiency and free up time to focus on more value-added tasks.
One of the positive effects we’ve also seen is for systems to reduce staff stress and anxiety levels. We heard this directly with staff at Intermarche. They told us they felt more confident because there was less theft, and video evidence meant they knew why SCOs were being blocked. That meant they had to intervene less often, and interventions were more likely to be focused on helping customers.
It makes sense, doesn’t it? If the technology is nudging shoppers to correct a miss-scan or accidental mistake, you don’t have the same level of interventions. And that same concept of ‘nudging’ can be applied across the store.
How will loss prevention AI help retailers better secure their stores in the future?
We’ve talked about the real time element of the technology, which allows you to spot an anomaly in the moment but there’s also an offline element, which is all of this insightful data that you’re collecting over a period of time. That gives you insight into macro versus micro trends.
For example, we learned from a trial in Belgium, in a city centre store, that the most stolen item was sushi. It was not an instantaneous finding but an interesting insight learned over a period of time. We also learnt that the self-checkout with the highest level of fraud, was the one closest to the attendant station – sounds illogical, until you realize it’s because the attendant is often not present at the station.
Those types of insights allow you to understand which stores in a network are the most targeted, the times of day that show higher fraud levels, or if the layout of a self-service area is a contributing factor. It also starts to level the playing-field with online retailers who can tap into a wealth of shopper insights based on pages visited, time on page, basket analysis, left items, etc.
Armed with these insights, you can move to the next level, which is predictive action. That predictive element is the next evolution of this technology.