As retailers explore where and how to use agentic artificial intelligence inside their operations, deployment in the supply chain, and specifically returns management, is emerging as a logical starting point. 

Why? For one, returns and shrink today are costing retailers $796 billion in total retail loss. Of that number, returns represent $706 billion in total return volume, with $100 billion lost to preventable fraud and abuse. Connected data and AI can address these losses. 

Retailers today utilize machine learning models that are trained on retailer data and a consortium of shopper data to evaluate returns transactions in real time. Agentic AI goes further to support the internal workflow, including when analysts investigate fraud patterns, identifying incidents across locations, and helping retailers act faster.  

But it’s not about agentic AI alone. Retailers need a multi-layered approach to AI to turn losses into profits.  

Unlike earlier AI technologies that primarily analyze data and deliver insights, multi-layered AI, with AI agents embedded inside internal processes, can accelerate workflows by taking direct, guided action. The layers work like this: 

Data lays the foundation. First, retailers and supply chain teams must have unified, omnichannel data, specifically customer identities and linking across cards, addresses, emails and more, flowing into one single, trusted view to support accurate decisions. This is the foundation for how AI helps support returns.  

Machine learning screens returns. The data intelligence allows retailers to then implement machine learning models to screen return transactions. AI reviews the streamlined data, including a customer’s shopping history, behavior patterns, loyalty information and mor. The ML decision intelligence then recommends if a return should be approved, denied, or flagged as a warning to the consumer. 

Agentic AI accelerates workflows. The data intelligence also makes it possible for AI agents to accelerate processes by allowing analysts to look up transactions, pull case details for law enforcement, assign work items, identify and execute next steps, or retrieve results for specific time periods across stores or distribution centers, all in natural language, saving hours of time.

Why Data Matters Most

Of course, for AI to truly improve operations in returns management, and to power decisions that trickle down throughout the supply chain, clean and enriched returns data must be at the core. Retailers need data flowing to one centralized place for the AI layers to access, and that data must be founded on decades of retail transaction history, unique omnichannel consumer identifiers, and behaviors that are stitched across credit cards, emails and addresses, and a pool of cross-retailer consortium data. 

When cases are linked automatically across store locations, whether online or not, returns and incidents are always visible and connected. This is an algorithmic AI at work; an unglamorous but critical aspect of returns. 

Success with AI at any level depends on data intelligence, and retailers must first streamline data into one trusted view to empower AI agents.

Factoring in Consumer Perception of AI

The goal of decision intelligence and automating workflows is to reduce unnecessary losses within returns, but they also must improve the returns experience for consumers.

 With this in mind, retailers need to be transparent about the use of AI within returns, as consumers predominantly trust humans more than AI to process returns. AI can dramatically improve returns processes, but privacy concerns mean that automation should operate within clear guardrails. AI agents don’t replace human judgment, but act on clearly defined rules and structures put in place by loss prevention and supply chain teams.

AI is not handing returns over to machines. Instead, dashboards shift from giving human teams AI-powered insights about what has happened inside an operation, to a real-time window on what AI agents are doing, and how teams can alter and guide decisions as needed.

Retailers are already looking at how they should restructure around agentic AI systems. In fact, consulting firm BCG’s data shows more than a third of enterprise businesses are using agentic AI inside their operations. But retailers need to further explore agentic AI inside returns and logistics.

Today, dashboards flag costly return rates, suspicious refund activity, or locations with abnormal behaviors. The insights are helpful, but typically arrive after the fact.

The agentic AI promise, however, is that companies can move from analysis to real-time decision-making. Dashboards flag a suspicious return as it happens, and update inventory and logistics data in the process. Loss prevention, finance and supply chain teams all have access to the decisions being made, and can adjust as needed.

AI has already proven to be an asset in limiting returns abuse and fraud, and improving the logistics of returns. Agentic is the next phase. Done openly and with cross-channel data at its core, retailers can keep all parts of the organization aligned and working together toward reducing loss and maintaining a positive consumer experience. 

Vishal Patel is chief product officer at Appriss Retail.