Hi all—Blake here. Buckle up, because this month’s edition is going to be juicy.

A few weeks ago, OpenAI quietly killed Instant Checkout. Six months from launch to eulogy. And honestly? The postmortem tells you more about where grocery e-commerce is heading than anything that shipped this year.

Let's get into it.

As always, stick around for the Featured Insights at the end! We’re forwarding you to an incredible read from FMI that’s worth your time.

🤖 Own the Agent, Rent the Surface
✈️ Your Best Digital Customers Are Quietly Leaving
💬 Featured Insight: FMI and NIQ unveil the latest e-commerce report

001 // OpenAI Tried to Own the Commerce Layer. It Couldn't.

When OpenAI launched Instant Checkout last fall, the premise made sense on paper—hundreds of millions of people already use ChatGPT to research products, so why not let them buy right there?

The problem was execution. OpenAI tried to be both the discovery layer and the transaction layer without building any of the infrastructure that commerce actually requires. Inventory sync wasn't there. Tax collection wasn't there. Fraud prevention wasn't there. The product data was scraped, not structured, which meant availability and pricing were often wrong.

Retailers scrambled to integrate. Walmart made roughly 200,000 products available. Shopify aimed for over a million merchants. What actually shipped? Somewhere between a dozen and 30 Shopify stores, depending on who's counting—and a Forrester analyst described the inventory management as "disastrously absent from the plan."

Conversion told the rest of the story. Walmart reported that checkout through ChatGPT converted at roughly a third the rate of its own website. People were happy to browse and research inside the chatbot, but when it came time to hand over a credit card, they wanted to be somewhere they recognized.

By March 4th, Daniel Danker—Walmart's EVP of AI Acceleration, Product and Design—called the whole arrangement "a very temporary moment in time." Not soon after, OpenAI confirmed the shutdown.

002 // Walmart Rewrote the Architecture and the Power Dynamic

Here's where it gets interesting: Walmart didn't walk away from ChatGPT—it restructured the relationship entirely.

Instead of OpenAI controlling the checkout, Walmart embedded Sparky (its proprietary AI shopping assistant) directly into ChatGPT and Google Gemini as a dedicated app. The technical architecture is the key detail. When a shopper asks ChatGPT to help them find something, the request routes to Sparky, which searches Walmart's inventory, presents options, and handles the purchase—all within the ChatGPT interface, but entirely on Walmart's rails. OpenAI gets the exposure. Walmart keeps the customer data, the transaction, and the post-purchase relationship.

Early data from the pilot suggests it's working. According to one report citing internal Walmart data, users who access Sparky through ChatGPT complete purchases at roughly 70% the rate of those on Walmart.com directly—a massive improvement over what Instant Checkout delivered. And the purchases look different: the top-selling categories through ChatGPT aren't household staples but vitamin and protein supplements, originating from prompts that don't start with shopping intent at all. Danker noted at a recent investor conference that these purchases are coming from conversations that don't begin with commerce—meaning the agent is capturing demand that wouldn't have reached Walmart through its own app.

Walmart is also in conversations with Anthropic about bringing Sparky to Claude, which tells you how the retailer views these platforms: as distribution channels, not technology partners. Own the agent, rent the surface.

003 // The Template Is Set. The Question Is Who Follows It.

What Walmart did here isn't just a partnership adjustment—it's a structural proof point for a principle that, frankly, should have been obvious: the retailer who owns the agent logic, the product data, and the customer relationship controls the experience, regardless of which AI platform the shopper is using. The moment you hand that to a platform, you become a catalog inside someone else's system.

For national retailers with engineering teams and billion-dollar tech budgets, this is challenging but achievable. For regional grocers, the question gets harder—but it doesn't get less urgent.

Because here's what's forming underneath all of this: a world where AI agents, not shoppers, are the ones querying your store. An agent needs structured access to your inventory, your promotions, your loyalty data, your fulfillment logic—it needs an interface layer (call it MCP, call it an agentic API, call it whatever you want) that exposes your store's operating logic in a way any model can interact with.

Without that layer, the agent doesn't know your store exists. It can't recommend your products, build a cart, or complete a transaction. You're not competing—you're invisible.

Walmart built that layer and is now deploying it across every major AI surface. The question for the rest of the industry isn't whether this pattern will spread—it's how quickly, and whether you'll have something ready when it does.

004 // Retail Media Just Found Its Next Battleground

One more dimension worth watching: Walmart has already started running ads inside Sparky, using a format it calls "Sponsored Prompts"—brand recommendations that surface within conversational shopping flows. A shopper asks Sparky for a protein powder recommendation, and a sponsored suggestion appears naturally in the response, connected directly to add-to-cart.

This is what retail media looks like when it's woven into the intelligence layer. The ad appears at the exact moment of intent, inside the conversation, measured end-to-end—a fundamentally different model than banner ads or keyword bidding, and one that's only possible because Walmart owns the agent.

If your AI experience belongs to someone else, your retail media opportunity does too.

001 // The Retention Problem Nobody's Measuring

There's a pattern I keep seeing in conversations with grocery operators: acquisition metrics look healthy, new users are signing up, first orders are coming through, digital is "growing." But when you ask about the second order—or the fifth—the picture gets murkier.

The data backs up what the conversations suggest. Industry-wide, ecommerce sees 70-77% annual churn, meaning more than three-quarters of customers never come back after their first purchase. Grocery does better than most categories because people need to eat every week, but that built-in advantage masks real vulnerability—subscription retention rates in grocery often hover in the 30s past the first year beyond the first year, which means even among your most committed digital customers, you're losing more than half.

And here's the uncomfortable finding from PwC's 2025 customer experience survey: true brand loyalty dropped to 29% last year, down five points from 2024. Nine out of ten executives think loyalty is growing. Only four out of ten consumers agree.

002 // The Second Trip Is Where You Win or Lose

The first order is expensive to acquire and operationally fragile—the customer doesn't know the interface, they're not sure the substitutions will make sense, they don't trust the produce. Everything is a test.

Earlier Bain/Google research found only 42% of first-time online grocery shoppers say the experience actually saves them time, which is the entire value proposition of the channel. That number climbs to 63% by the third order, as customers learn the system, save their favorites, and build routines.

The gap between those two numbers is where retention lives or dies, and it's shaped by decisions most grocers treat as afterthoughts—substitution quality, delivery window reliability, reorder friction, loyalty integration, and whether the experience feels personalized or generic.

003 // Retention Is a Systems Problem, Not a Marketing Problem

The instinct is to throw marketing at retention: send a coupon, trigger a re-engagement email, offer free delivery on the next order. Those tactics have a place, but they're patching symptoms. And those tactics don't move the 70-77% churn number we opened with. Most grocers lack the infrastructure to even measure retention properly—they see top-line order growth, but not which customers came back, which dropped off, or where the second trip slipped away.

Real retention in digital grocery is a systems problem. Does your substitution engine have enough structured data to make intelligent replacements? Does your loyalty program actually flow into the digital experience, or does it live in a separate silo? Can your fulfillment system reliably hit the delivery window the customer selected? Does the reorder experience surface what they actually bought last time, or make them start from scratch?

Every one of those questions is an architecture question, and they connect directly to the same infrastructure principles we've been talking about all year—clean data, integrated systems, intelligence wired into operational workflows. The retailers pulling ahead on retention aren't running better campaigns. They're making the second trip feel like the twentieth.

For most regional grocers, building this layer from scratch is impossible, but having it isn't an option anymore either.

ARLINGTON, VA — FMI–The Food Industry Association and NielsenIQ (NYSE: NIQ), a leading consumer intelligence company, find that grocery shopping is no longer a choice between store and screen, but instead a seamless blend of both in the latest Digital Engagement Transforms Grocery Shopping 2026 report. With omnichannel behaviors as the new normal, total U.S. online grocery sales will reach $452 billion by 2028, making ecommerce a primary engine of grocery progress.

The report finds that ecommerce now drives nearly three-quarters of total grocery dollar growth. In 2025 alone, online sales contributed close to 75% of total grocery dollar growth, while in-store sales remained relatively stable.

"Ecommerce is poised for extensive growth,” said Jack O’Leary, director of ecommerce strategic insights for NIQ. “With online grocery sales increasing at a projected 11.6% annual rate through 2028 and already accounting for about one-fifth of total grocery spending, success now hinges on how well retailers connect digital and physical experiences into one cohesive journey.”

Read the full article →