Skip to content

News & Insights

In Grocery, AI Pays Off When It Runs on Good Data and Experienced Judgment

Grocery operators are using AI for everything from predictive store ordering, automated inventory tracking and personalized shopper engagement. With so many options on the table, the question is not whether to use AI, but how to use it in a way that will actually result in accurate, revenue-driving decisions.

The answer starts with the data AI has to work with. For example, when food companies bought general-purpose AI to handle recipe development, it failed. The technology knew nothing about their specific product formulations, equipment, costs, or regulatory requirements, and most of their product knowledge sat in old emails and lab notebooks where no system could use it. The tools that did work handled the routine checks on cost, nutrition, and compliance so product development teams could focus on the product.

Grocery merchandising works the same way. AI improves margins only when it starts with good data and is backed by experienced people who make the decisions. At Radian, we recently enhanced our Merchandising Planning Suite to include AI-enabled capabilities across assortment, base price, promotion, shopper marketing, and space management on a single platform.
As merchandising teams are tasked to do more with leaner teams, respond to market volatility in weeks rather than quarters, and defend every decision with data, most AI tools promise that speed, but often deliver the most obvious (at best) and/or dubious (at worst) recommendations merchants can’t explain to senior leadership, suppliers, or themselves. Based on that experience, we have seen what separates AI projects that deliver from the ones that stall.

Where AI Adds Value

The questions merchants ask have not changed all that much, but AI can help get to an answer far more quickly. In a promotion review across a full category, pulling, formatting, and calculating the data once took up most of the effort. Now teams complete their initial analysis almost immediately.

AI is also built for needle-in-a-haystack work that no merchant has time for. In our work with grocery partners, we use AI to review thousands of items, shelf placements, and planogram blocking decisions at once, flagging anomalies that are easy to miss by hand. That review is important but rarely urgent, so before AI, it was often pushed aside.

What AI cannot do is replace judgment. We have seen AI describe commodity market conditions with complete confidence, but it was clearly wrong when reviewed by anyone who has bought produce or protein. It will suggest promotion plans that overlook seasonality, competitive dynamics, recent history or key market conditions and does not consider how real trade relationships actually work. Catching that takes someone who has both experience and expertise in the industry. Radian’s belief is that decisions about price image, assortment direction, and space allocation should stay with experienced merchants.

Where AI Projects Stall, and What to Do Instead

Retailers want AI to make the work easier and more automatic. It can, but only with accurate data and guardrails built around the business. Grocers differ in small ways that affect the bottom line, such as how trade areas behave, UPC versus random weight items are managed or how product hierarchies are built. Off-the-shelf tools sold as a ready-made answer miss those differences.
That’s why it’s important to choose partners and solutions that start by understanding how a retailer’s stores, categories, items and data relate to one another. It is always specific to the business, and takes hands-on expertise to get right. Without it, no AI tool will produce recommendations a merchant can trust.

AI That Adds Up

AI options for grocers will only multiply, but the ones that make a positive impact will share two things. They will start with information and guardrails (i.e., business rules) that are accurate, complete, and specific to the business, and they will leave the most important decisions to merchants who know the categories. When both are in place, AI stops being one more tool to manage and starts paying for itself.