Ask most people what AI in grocery looks like, and they'll picture a chatbot helping you build a shopping list, or a recommendation engine nudging you toward a recipe. That's the version that gets headlines. It's not the version that's actually changing how grocers and consumer packaged goods (CPG) companies operate.

‍

The more consequential shift is happening in the back end — in forecasting, replenishment, quality control, and inventory. Recent reporting from Grocery Dive has made this split increasingly clear: while consumer-facing AI still struggles to earn shopper trust, operational AI is tackling problems that have plagued grocery for decades, from shrink and stockouts to markdowns and mis-shipments. The question for grocers and CPG suppliers isn't whether to invest in supply chain AI anymore. It's whether that investment is built on a foundation solid enough to work.

‍

‍

What "Supply Chain AI" Means in Grocery and CPG

‍

Supply chain AI, broadly, is the use of machine learning and real-time data to improve how goods move — the decisions, handoffs, and exceptions that happen between raw materials and the customer. In grocery and CPG, that definition gets a harder edge. Margins average around 1.7%, according to FMI — The Food Industry Association, which leaves little room for waste or error. Product is perishable, often measured in hours rather than weeks. And digital grocery has pushed fulfillment down to the individual item, exposing gaps that bulk-movement supply chains were never built to catch.

‍

That combination — thin margins, perishable product, item-level expectations — is why grocery has become one of the highest-stakes proving grounds for supply chain AI. When it works, the savings are measurable. When it's built on bad data, the results are, too, just in the wrong direction.

‍

‍

Where Grocers Are Actually Putting AI to Work

‍

The investment is real, and it's concentrated in operations rather than the storefront. Hy-Vee partnered with Relex to sharpen fresh product forecasting, ordering, and replenishment. Heritage Grocers Group is using AI to manage the complexity of pricing promotions at scale. Grocery Outlet recently adopted a system from Afresh to modernize ordering across departments. Kroger built Sage, an AI assistant that helps frontline associates manage schedules and store tasks rather than helping shoppers pick a recipe. Albertsons is using AI to inspect incoming produce for quality before it ever reaches a shelf.

‍

These aren't experiments chasing buzz. They're targeted fixes for operational pain points that have direct P&L impact. That's a big part of why they're succeeding where consumer-facing tools have moved more slowly — adoption is still uneven across the industry. A 2025 FMI survey found just 47% of grocers reported using AI in their operations, compared with 93% of suppliers, though large grocers are moving much faster, with 77% already using AI in some form.

‍

‍

The Catch: AI Is a Capability, Not a Strategy

‍

Here's where the industry conversation gets more honest. As one grocery product leader recently told Grocery Dive, AI is a capability, not a strategy — and a personalized promotion or a smart reorder recommendation is only as good as the data feeding it. If a system recommends a product that isn't actually on the shelf, or forecasts demand based on inventory counts that were accurate three days ago but aren't now, the AI hasn't failed. The data underneath it has.

‍

This is the same problem grocery has wrestled with long before anyone called it supply chain AI: the physical data gap between what a system records and what's actually happening on the floor, in transit, or in the backroom. Most grocery and CPG operations still run on ERP systems, periodic manual counts, and barcode scans at fixed checkpoints. Those tools capture a single moment — a pallet scanned at receiving, a count completed on a Tuesday — but between those moments, nothing is known. A product can sit in a backroom, ride an extra day in transit, or drift out of temperature range, and the digital record won't reflect any of it until someone notices.

‍

That gap is exactly why experts advising grocers on AI keep circling back to data quality before technology selection. You don't necessarily have to wait until your data is perfect to start, but layering AI on top of data that's wrong just produces confident, wrong answers faster.

‍

‍

Closing the Gap With Continuous, Physical Data

‍

This is where the sensing layer becomes as important as the AI model sitting on top of it. Supply chain AI needs a continuous, real-time stream of what's actually happening to product, not a snapshot from the last scan. This is what Physical AI was built to address.

‍

The Wiliot Physical AI platform processes signals from battery-free IoT Pixels using purpose-built AI and ML models, turning passive product movement into real-time, actionable intelligence. Instead of relying on manual scans or fixed checkpoints, IoT Pixels attach to cases, pallets, and reusable assets and continuously transmit location, temperature, humidity, and movement data as products flow from supplier to shelf.

‍

Freshness and cold chain. The transition to digital grocery has already exposed how thin the margin for error is when a system says a product is available and a picker can't find it, or finds it compromised. Case-level temperature monitoring catches excursions that zone-level sensors miss entirely, delivering up to a 90% reduction in temperature-related spoilage for grocery and CPG operators.

‍

Inventory accuracy. Continuous, zone-level visibility across backrooms, coolers, freezers, and the sales floor replaces periodic manual counts with inventory intelligence that's always current — turning FIFO into FEFO, first expired, first out, automatically, and pushing inventory and receipt accuracy above 99%.

‍

Shipment accuracy. Automated shipment verification and automated receiving catch incorrect or incomplete loads at the dock door, before trucks leave, cutting mis-shipments by up to 90%.

‍

It also turns FSMA 204 traceability from a manual documentation burden into an automatic byproduct of continuous monitoring — case-level temperature and location history, generated as goods move rather than reconstructed after the fact. Grocery and CPG operators typically see ROI on this kind of continuous visibility within a year.

‍

‍

The Next Frontier: Agentic Retailing Needs Real Data, Not Just Real Models

‍

The industry's next conversation is already underway: agentic AI that doesn't just recommend an action but takes it — reordering, rerouting, flagging quality risk, adjusting a promotion — without a human reviewing every step. Category managers, supply chain analysts, and demand planners are being told to expect decisions that move at machine speed rather than meeting-cadence speed.

‍

But agentic systems can only act on what they know, and they can only know what the sensing layer tells them. Industry voices are increasingly framing this as a "data fabric" that has to run across the entire enterprise before agentic AI can be trusted with real decisions. That's not a software problem alone, it's a physical visibility problem. An agent can't reroute inventory it can't see, or flag a spoilage risk it has no signal for.

‍

That's precisely why the investment in continuous sensing and the investment in AI are converging into the same investment. Grocers and CPG companies that build the physical data foundation now, with case-level visibility into location, condition, and movement, are the ones positioned to move from automated tasks today to genuinely agentic operations tomorrow.

‍

‍

The Bottom Line

‍

The grocers making real progress with supply chain AI aren't necessarily the ones with the flashiest tools. They're the ones treating AI as one part of a larger system, one that starts with trustworthy, continuous, item-level data about what's actually happening in stores, distribution centers, and trucks. Whether you have AI in place already or are just starting to build toward it, that foundation is what turns a promising pilot into measurable, lasting ROI.

‍

If your grocery or CPG supply chain still relies on periodic scans and manual counts to feed your AI investment, there's a visibility gap worth closing. Talk to Wiliot about what continuous, scan-free intelligence looks like for your operation.

‍