Loss prevention has always operated under pressure—but today that pressure is structural, not cyclical.
Shrinkage, organized retail crime, shoplifting, and self‑checkout (SCO) abuse are escalating at the same time even as retailers face labor shortages and margin compression. What has changed is not just the scale of the problem, but the tools available to address it. Artificial intelligence has moved from experimentation to execution across retail—and loss prevention can no longer afford to be an exception.
According to the NVIDIA State of AI in Retail and Consumer survey, 91% of retail and CPG organizations are now engaged with AI, and 58% are actively deploying AI solutions, up from 42% just a year earlier. This shift is driven by results: 89% of respondents report AI has increased revenue, and 95% say AI has reduced annual costs.
The message for loss‑prevention leaders is unmistakable: AI is no longer optional infrastructure. And visual AI is its most underutilized application.
Shrinkage is no longer a visibility problem—it’s about speed
Retailers already have extensive camera coverage. The failure point is not visibility, but response timing.
Traditional surveillance systems are largely forensic. Video is reviewed after an incident, long after loss has occurred. In an environment where organized theft and SCO abuse happen in seconds, this model is fundamentally misaligned with reality.
The NVIDIA survey shows why this matters. Across retail functions, AI is delivering value by enabling real‑time decision‑making, not retrospective analysis. 54% of respondents report AI has increased employee productivity, and 52% report improved operational efficiency. Loss prevention operates under the same constraints—and stands to gain the same benefits—by shifting from manual monitoring to AI‑driven detection.
Visual AI transforms cameras into live intelligence systems, surfacing high‑risk behaviors as they happen rather than documenting them after the fact.
Self‑Checkout changed the risk equation—and the math
SCO lanes are now a permanent fixture in US and European retail. They increase throughput and reduce staffing requirements, but they also introduce new loss vectors that traditional LP methods struggle to manage.
The challenge is scale. SCO transactions move too quickly, and volumes are too high, for human monitoring to be consistent or cost‑effective. Adding staff undermines the economics of SCO; removing oversight accelerates shrink.
This is precisely the kind of tradeoff AI is already helping retailers resolve elsewhere. The NVIDIA report shows that 66% of respondents prioritize AI technologies such as data analytics and generative AI to optimize operations, while 47% are already using or assessing AI agents—systems capable of reasoning and acting autonomously.
Applied to SCO, visual AI enables targeted intervention. Rather than blanket oversight, AI identifies anomalous behaviors—missed scans, item substitution, barcode switching—in real time, allowing staff to intervene selectively. Fewer alerts, higher confidence, better outcomes.
Labor constraints make automation inevitable
Loss prevention has always relied on people: store detectives, associates, regional LP teams. But, the workforce reality has shifted.
The NVIDIA survey identifies lack of AI talent as the top barrier to AI implementation, rising to 46% year over year. This mirrors broader labor constraints in physical retail, where staffing shortages and wage inflation limit the ability to scale human oversight.
AI’s value, however, is not replacement—it is augmentation. Retailers report that AI improves productivity precisely because it allows employees to focus on judgment and action rather than constant monitoring.
Visual AI extends LP teams in the same way. It reduces the need for continuous observation, prioritizes exceptions, and enables faster, safer interventions—without increasing headcount.
Scale, latency, and inference define whether LP AI works
One of the most consequential findings in the NVIDIA survey is the growing focus on AI inference. As AI moves into production, 41% of respondents cite cost efficiency and total cost of ownership as top concerns, while 35% prioritize latency, accuracy, and throughput.
For loss prevention, these are not abstract metrics.
Visual AI must run continuously across thousands of cameras, deliver alerts fast enough to enable intervention, and operate economically at enterprise scale. Systems that generate excessive false positives, introduce latency, or drive unpredictable inference costs will fail—regardless of model accuracy.
This is why retailers are increasingly favoring open, flexible AI stacks. 79% of respondents say integrating open‑source models and tools is moderately to extremely important, reflecting the need to tune AI systems to proprietary data, store layouts, and operational realities.
Loss‑prevention AI that cannot scale cleanly will remain stuck in pilot mode. And pilots do not stop retail shrinkage.
Why SAI Group’s visual AI aligns with loss prevention reality
Against this backdrop, SAI Group’s visual AI approach aligns with how modern loss prevention actually operates.
Rather than treating surveillance as a passive security function, SAI Group positions visual AI as real‑time operational intelligence for physical environments.
This approach resonates with LP leaders because it focuses on:
- Detection in the moment, and not on post‑incident review
- Leveraging existing camera infrastructure to control costs
- Scaling consistently across large store networks
- Reducing false positives to protect staff time and credibility
The NVIDIA survey makes it clear that retailers are doubling down on AI where it proves business impact. 92% of executives plan to increase AI budgets, with optimization of production workflows cited as a top investment priority. Visual AI that performs reliably at scale fits squarely within this execution‑focused mindset.
Loss prevention is becoming a strategic function again
As AI delivers validated results across retail, loss prevention has an opportunity to move from defensive cost control to strategic margin protection.
Retailers that integrate visual AI into LP operations can:
- Reduce shrinkage without increasing labor
- Improve associate safety and confidence
- Protect customer experience while enforcing controls
- Respond faster to organized and adaptive theft tactics
Those that do not will continue absorbing losses that AI‑enabled competitors are actively preventing.
The bottom line
The NVIDIA survey concludes that retail is moving beyond debating AI adoption to optimizing deployment. Loss prevention must be part of that shift.
With 95% of retailers already seeing cost reductions from AI and 90% planning increased investment, the question is no longer whether AI belongs in loss prevention—but how quickly it can be operationalized.
Visual AI is no longer an emerging technology for LP teams. It is becoming core infrastructure.
And in an environment where shrinkage pressures continue to rise, the cost of waiting is higher than the cost of acting.
About SAI
As a leader in computer vision technology, SAI Group delivers cutting-edge, multi-modal AI solutions into retail environments. Using a unique platform approach, its technology uses existing camera systems to target losses, increase store safety, and underpin operational efficiencies.
All solutions are built from the ground up to ensure the highest levels of security and data protection, respecting the privacy expectations of the public and operating to stringent ethical standards while delivering substantial value to our clients. Globally, SAI monitors millions of transactions per day, protecting the revenues from tens of millions of product sales and hundreds of millions of customer interactions. Its models also accurately identify anti-social behaviour, aggression and violence, helping to de-escalate situations with real-time interfaces to security officers and operations centres.
