California-headquartered Appriss Retail announced Wednesday it had released a new, generative AI-powered tool that can help identify organized retail crime (ORC) rings and repeat offenders.
Generative AI is often thought of as a way to power chatbots, like ChatGPT, or image generation tools, like Dall-E. But the technology also has the capability to ingest and process natural language patterns—how a person would describe something—to recognize patterns.
That feature of the technology is how Appriss has built out its latest model, which it calls Incident + ORC Intelligence. The model uses data and narratives written by associates to link incidents that are potentially related to one another.
Pedro Ramos, chief revenue officer at Appriss, said the ability to include narratives in the system’s analysis of incidents improves the likelihood of a link, because narratives don’t have to be as precise as a system may require data points to be.
“We understand that the skills it takes to be a good store manager or a good store associate don’t necessarily translate to being able to capture every minute detail of a shoplifting incident, but people can write a very basic narrative. That’s why we took the approach of using generative AI, not just to look at the data elements, but also to look at the narratives and try to find common themes,” Ramos told Sourcing Journal.
For instance, a store associate may write a short report on an incident that states, “Two men entered the store in yellow ski masks around 1:00 p.m. on a Tuesday. They specifically sought out athletic apparel and shoes and left without paying for over $700 worth of merchandise.”
If the system identifies a potential link between that incident and other instances of retail crime, the system will flag it to store administrators. Employees can vet the purported pattern, identifying whether the connection the AI model made is valid and makes sense. In turn, the model learns from the user’s input.
Photo courtesy of Appriss Retail.
“The end user has a mechanism in the backend to tell the model if they made a mistake on one or two transactions; the model learns from them, so there are mechanisms [controlled by people] that help the feedback loop in order to train the models,” Ramos explained.
He noted that, despite the models having the ability to learn from users, companies own their data and remain in control of it as needed.
If an employee agrees with the link the system has drawn, they can share that information with local law enforcement.
“The back end of the system has various mechanisms that are designed to fit the most commonly used ways that law enforcement wants to adjust this data. If a retailer wants to provide access to their data to a third-party, we have the mechanism to do that, too,” Ramos said.
Ramos said the most likely customers for this technology are retailers with less robust loss prevention teams and smaller store footprints. Typically, big-box retailers, like Target and Home Depot , employ loss prevention teams to help stave off issues in their stores.
Appriss declined to disclose the companies using the technology already, but Ramos noted that “several dozen” retailers have either implemented the technology or are in the process of negotiating on using it.
It’s too early to tell what the long-term outcomes of the technology will be, but a nearer-term impact could be understanding the geographies that are most affected by retail crime; the types of products that crimesters have a specific affinity for and more. Ultimately, that could influence store-level decisions.
“If a retailer has better information and makes better decisions on product allocation , how much product they put on the shelf, it hardens the target. By that, I mean, it makes it less profitable for an organized retail criminal to come into the store and target that store, because [if] there’s less availability [of product], without impacting sales, then there’ll be several byproducts: [fewer] ORC criminals coming in; [fewer] better sales and lower shrink,” Ramos said.
Get updates delivered to you daily. Free and customizable.
It’s essential to note our commitment to transparency:
Our Terms of Use acknowledge that our services may not always be error-free, and our Community Standards emphasize our discretion in enforcing policies. As a platform hosting over 100,000 pieces of content published daily, we cannot pre-vet content, but we strive to foster a dynamic environment for free expression and robust discourse through safety guardrails of human and AI moderation.
Local News
Comments / 0