This is the sixth post of a 10-part series: AI Reality Check for Grocers. Best to begin here if you haven’t already.
Grocery operators spend enormous resources fighting shoplifting. Cameras. Security tags. Loss prevention teams. AI-powered surveillance systems. The industry treats shrink as an existential threat, and for good reason. At margins of 1-3%, every stolen item matters.
But there’s a more expensive theft happening that nobody’s talking about. It doesn’t show up on a security camera. It doesn’t trigger an alert. And the grocers experiencing it are actually paying the thieves to do it.
Every order that runs through a marketplace is a data point walking out the door. And that data isn’t just being collected. It’s being fed into AI models that are getting smarter about grocers’ operations every single day. The platforms running their ecommerce are building AI to run their stores. Grocers are unknowingly writing the tuition checks.
This isn’t a customer experience problem. It’s an operational AI problem. And unlike shoplifting, it compounds quickly and silently.
🔍What’s actually being stolen?
Earlier in this series, I argued that grocery AI is being misallocated toward customer-facing digital experiences when the real value is in operations: shrink, out-of-stocks, inventory accuracy, and demand forecasting. Those are the problems that determine whether the business makes money.
Here’s what makes the platform AI problem so damaging: the data leaving through their partners is precisely the data that trains powerful operational AI models. And it's being used to build one. Most grocers don't know it's happening.
Take substitution patterns. When a shopper accepts a substitute on a delivery order, that’s not a minor inconvenience. That’s a signal that a primary item was unavailable. A stockout. An inventory failure. Captured at scale, across thousands of orders, substitution data trains an AI model to understand where out-of-stocks are chronic, which categories have the worst inventory accuracy, and which supply chain relationships are consistently failing.
That’s not marketing intelligence. That’s operational intelligence. The kind that drives replenishment decisions, informs forecasting models, and directly affects margin.
Instacart launched its Consumer Insights Portal in September 20251. It’s built on daily activity from nearly 100,000 stores and more than 1,800 retail banners. It surfaces SKU-level performance, search behavior, substitution patterns, and promotion impact, based on actual transactions. And its data is sold to CPG brands.
Think about it. The grocers’ operational failure data is processed by Instacart’s AI, packaged, and resold to the brands with which grocers negotiate every year.
It gets worse. In January 2026, at CES, Instacart unveiled Data Hub2. This clean room solution gives CPG brands and agencies access to Instacart’s first-party grocery purchase signals. The positioning is retail media. The reality is an AI demand intelligence engine trained on transaction data flowing through grocers’ operations.
DoorDash is building the same thing. The company ended 2025 as the leading third-party marketplace3 by order volume across grocery and retail in the United States, adding 33 new U.S. grocery partnerships over the past year. Their ChatGPT integration moves users from recipe browsing directly into a DoorDash grocery cart.
It’s framed as a discovery feature. What it actually is is AI learning what shoppers want before they walk through the grocer’s door.
⚠️Why This Gets Worse Over Time.
Shoplifting is a static loss. Someone takes an item. It’s gone. The damage is discrete and contained.
AI-powered data theft doesn’t work like that. It gets worse over time.
AI models improve based on data. Every order flowing through Instacart or DoorDash makes their AI marginally better at predicting demand, identifying inventory failures, and understanding how grocery categories actually behave.
That intelligence accumulates. The gap between what the platform’s AI knows about your operation and what your internal systems know widens with every order.
Most grocers have no idea.
Rodrigo Dellacqua, co-founder of grocery commerce VTEX, put it plainly: “If grocers can’t access their insights just because it’s locked up in a third-party app, they’re always going to be a step behind.4”
A step behind on replenishment. On forecasting. On knowing which substitutions are masking chronic inventory problems. On understanding where demand is going before it gets there.
These are margin decisions. And the AI making them doesn’t belong to the grocers themselves.
Kroger's Mistake Should Scare Every Grocer.
Kroger has been closing automated fulfillment centers and leaning more heavily on Instacart, DoorDash, and Uber for ecommerce fulfillment.
On the surface, that’s a rational cost move. Automated fulfillment is expensive, and the unit economics haven’t worked. But every order routed through those platforms trains their AI. Not Kroger’s.
Every incremental order through DoorDash is another training data point in DoorDash’s AI model. Kroger is retreating from the infrastructure that would have given it operational control and, in doing so, accelerating the flow of data into the platforms it depends on. The platforms get smarter. Kroger becomes dependent.
This pattern is playing out across the industry. Grocers are optimizing for short-term unit economics, quietly outsourcing the data that trains the AI that will define competitive position for the next decade.
The Wynshop Acquisition Says Everything.
When Instacart acquired Wynshop5, it eliminated a platform that grocers had specifically adopted to reduce their dependence on the platform.
That’s not a delivery company move. That’s an AI company move. More surface area. More data. A richer training set. The transaction data flowing through partner stores isn’t a byproduct of the delivery business. It is the business.
💡The question grocers should be asking.
Grocers have sophisticated systems for catching shoplifters. They track shrink by category, by store, by shift. They know exactly how much merchandise they’re losing and where.
They have almost no visibility into what their partners are stealing and what AI is being built from it.
The raw material for operational AI is flowing out through their partners every day.
And for many grocers right now, the most sophisticated AI being trained on that raw material lives inside a platform system, not an internal one. The platform’s AI knows more about how grocers’ stores perform than their own systems do. And it’s being sold to CPGs to weaken their negotiating power.
🛑Grocers need to stop funding their thieves.
That’s not a technology problem. It’s a strategy problem. And unlike shoplifting, there’s no security camera that’s going to catch it.
The fix isn’t complicated, even if it’s not easy. Build first-party data infrastructure. Own the channels where your customers transact. Invest in internal AI capability before the gap between what your systems know and what your partners’ systems know becomes impossible to close.
The grocers who do this now will own their operational future. The ones who wait are funding someone else’s.
The merchandise walking out of grocery stores shows up in shrink reports. The data walking out through the grocers’ partners, and the AI being built from it, doesn’t show up anywhere.
At least with shoplifting, grocers know what they're losing.
AI Reality Check for Grocers is a 10-part series that cuts through the hype and focuses on where AI actually creates value, from the shelf to the supply chain.
“Instacart Introduces Consumer Insights Portal.” Progressive Grocer: https://progressivegrocer.com/instacart-introduces-consumer-insights-portal
“Instacart Debuts Data Hub, a Clean Room Offering for Enhanced Media Performance.” Instacart: https://investors.instacart.com/news-releases/news-release-details/instacart-debuts-data-hub-clean-room-offering-enhanced-media
“DoorDash Claims Top Spot in 3rd-Party Grocery and Retail Delivery.” Progressive Grocer: https://progressivegrocer.com/doordash-claims-top-spot-3rd-party-grocery-and-retail-delivery
“Whose cart is it anyway? Wynshop, Instacart and the battle for grocery data.” The Packer: https://www.thepacker.com/news/retail/whose-cart-it-anyway-wynshop-instacart-and-battle-grocery-data
“Whose cart is it anyway? Wynshop, Instacart and the battle for grocery data.” The Packer: https://www.thepacker.com/news/retail/whose-cart-it-anyway-wynshop-instacart-and-battle-grocery-data