Retailers spent more on in-store technology last year than the year before. Losses from empty shelves, mispriced goods, and poor inventory management went up anyway.
A Coresight Research study published in late May 2026, based on 200 senior U.S. retail executives, puts the total cost of in-store operational failures at $196.4 billion across hardware, grocery, and mass merchandise this year. That is up from the equivalent of 5.5% of gross sales in 2025 and 4.5% in 2024. Sector sales are growing at roughly 3% annually. The losses are outpacing revenue growth by a wide margin.
The reason, the research argues, is not that AI tools do not work. It is that retailers are deploying them in the wrong sequence. Most are prioritizing pricing automation software, with 43% directing capital there, and supplier collaboration platforms at 36%. Only 33% are investing in the shelf-scanning hardware that those same software tools depend on for their data. An automated pricing engine that is fed inaccurate stock counts will produce inaccurate prices. That is exactly what is happening: mispricing rates reached 13% in 2026, a four-point rise in two years.
The correct sequence starts with getting a real-time, accurate picture of what is physically on the shelf. That means cameras, sensors, or scanning robots moving through the store continuously. Only once that data foundation is in place do pricing, forecasting, and supplier tools have something reliable to work with.
BJ's Wholesale Club has done this at scale. The retailer deployed Simbe's autonomous scanning robot, called Tally, across all 244 of its warehouse clubs. The robot roams aisles multiple times per day, capturing images of shelves and checking stock levels, pricing accuracy, and product placement. BJ's used that data to build real-time digital models of each club, which the company then applied to route planning for online order picking. The result was a 40% year-on-year improvement in picking efficiency.
Albertsons is pursuing a broader version of the same logic. The grocery chain has committed to $1.5 billion in productivity gains across three fiscal years, backed by $2 billion in capital spending for 2026. AI-powered demand forecasting and computer vision are already lowering inventory and fulfillment costs, according to its Q4 earnings call in April.
Lowe's has taken a different path to a similar destination. Rather than robots, it uses AI-directed task lists that tell store associates exactly which shelves need restocking, based on real-time sales data. The company's productivity programme saved 80 non-productive labour hours per store per week and issued performance bonuses of up to $5,000 to store managers where results were documented.
The adoption gap between large and small retailers is stark. Among retailers with more than $5 billion in annual revenue, 73% have fully deployed store intelligence tools. Among those below $1 billion in revenue, that figure is 42%. Larger retailers benefit from the economics of scale: the hardware cost per store falls as it is spread across hundreds of locations, and the data generated becomes more useful as the store network grows.
For mid-sized retail operators outside the U.S., the findings carry a straightforward implication. The technology itself is not the obstacle. The sequence is. Buying a pricing tool before you have working shelf data is like buying a GPS for a car with no fuel gauge. The software will generate outputs, but those outputs will be wrong. The first investment worth making is the one that shows you what is actually on your shelves right now.
Ninety-seven percent of retail decision-makers surveyed said they have deployed or plan to deploy store intelligence technology within the next year. The question is no longer whether to invest. It is whether to get the order right before spending the money.