> ## Content Index > Fetch the complete content index at: https://giovanni-gallucci.ghost.io/llms.txt > Use this file to discover other available public pages before exploring further. # Reorder my usuals - URL: https://giovanni-gallucci.ghost.io/reorder-my-usuals/ - Published: 2026-09-30T15:12:31.000Z - Updated: 2026-09-30T15:12:30.000Z - Description: Every roundup from Groceryshop ran DoorDash's 50 percent bigger carts. Modern Retail printed the part they skipped: the chat window is the feature people ignore. What they use is the button that reorders what they already bought. That is a retention machine, not a discovery one. - Author: giovanni gallucci - Tags: Organic Social, Retail Distribution, Consumer Trends, Brand Positioning, Food and Beverage Weekly *Listen to this episode of the Food and Beverage Trend Report, then use the piece below.* The briefing I recorded this morning told you that 75 percent of people trust an AI agent more than they trust their friends. I went and read the report it came from. That is not what it says, and the actual sentence is worth more to you than the one we said out loud. The number comes from [Accenture's Consumer Pulse Research 2026](https://www.accenture.com/content/dam/accenture/final/accenture-com/document-fy26/q4/Accenture-Consumer-Pulse-2026.pdf?ref=giovanni-gallucci.ghost.io), a survey of 25,590 consumers across 16 countries including the United States. Worth saying plainly: Accenture sells consulting into exactly the transformation this report says is coming, so read it the way you would read any vendor's research. The finding is that 74 percent would trust a personal AI agent more than their best friend *to make a purchase on their behalf*. Not trust in general. Trust with an errand. And two sentences later the same report gives you the number nobody quoted. Seventy-four percent would let an agent handle routine tasks. Thirty-two percent are ready to hand over the purchase decision. Nine percent would go fully autonomous. Seventy-four will hand you the list. Thirty-two will hand you the choice. That gap is the only real estate a small food brand has left, and I did not hear one person at the conference last week talk about it. ## What actually happened on that stage Groceryshop ran at Mandalay Bay last week and AI was the whole show. [Grocery Dive's five takeaways](https://www.grocerydive.com/news/takeaways-groceryshop-conference-2026-las-vegas/831421/?ref=giovanni-gallucci.ghost.io) carried the line every roundup picked up: [DoorDash](https://www.linkedin.com/company/doordash/?ref=giovanni-gallucci.ghost.io) co-founder Andy Fang said early results from Ask DoorDash include a 50 percent increase in cart size and checkout five times faster than normal. That is the quote that traveled. [Modern Retail covered the same panel](https://www.modernretail.co/technology/how-kroger-giant-eagle-and-doordash-are-building-their-agentic-ai-assistants/?ref=giovanni-gallucci.ghost.io) and printed the part the roundups left out. Fang said the problem they found is the chat window. "Oftentimes when people think of using AI to order, they think of a chat window popping up," he said. When it pops up, people are not sure what to talk to the bot about, so they ignore it. What users actually appreciate, per Fang, is being able to prompt "reorder my usuals" during browsing instead of digging through their order history. The other thing DoorDash is building is easier swapping for out-of-stock items. So the celebrated conversational assistant's best-performing features are a reorder button and a substitution engine. Then I went back to the launch. [Grocery Dive's June 12 brief on Ask DoorDash](https://www.grocerydive.com/news/doordash-grocery-tools-conversational-shopping-assistant-agentic-ai/822735/?ref=giovanni-gallucci.ghost.io) lists what shipped on day one: shoppers can prompt it to reorder their last cart, shop frequently purchased products, and get suggestions based on past purchases. The backward-facing features were in the box from the start. Three months later the co-founder's reported lesson is that those are the ones people use. ## It is a retention surface wearing a discovery surface's clothes This morning's show said the app kills the browse. I said it, and it is close, and close is the problem. Killing the browse would at least be neutral. Everybody loses the aisle together, big brand and small, and we all start over somewhere else. That is not what this does. The app freezes the basket. It pays out to whoever was already in the cart, every week, automatically, and it has no opinion about anyone else. Nobody is discriminating against you. You are simply not in the history, and the history is the product. [Kroger](https://www.linkedin.com/company/kroger/?ref=giovanni-gallucci.ghost.io) says so out loud. Greg Foran, on the company's first quarter earnings call in June, said 95 percent of Kroger's transactions are tied to a loyalty card. His words: "That means we can measure actual purchase behavior, not just intent, and that's increasingly valuable to brands and advertisers." Intent is what you can earn with a good label and a good story. Behavior is a receipt. He is telling you which one the machine is built on. Kroger's chief digital officer Yael Cosset framed the assistant as closing the gap between inspiration and the basket. [Giant Eagle](https://www.linkedin.com/company/giant-eagle/?ref=giovanni-gallucci.ghost.io) was blunter about the math. Shweta Prabhu, who runs digital, e-commerce, martech, loyalty, data analytics and AI there, said the chain's omnichannel shoppers carry a 30 percent higher lifetime value than single-channel shoppers, so "every dollar that we invest in making that experience seamless for our guests is one that will keep our most valuable customers coming back to us." Kroger has announced its intention to acquire Giant Eagle, with the deal expected to close next year. Coming back. Keep. Every executive on that stage described a retention program and the trade press filed it under discovery. ## The ad slot was there on day one Here is the part that should decide your quarter. Kroger launched its assistant across most of its banners this summer, built with a grocery AI platform called Cooklist, and it had [product listing ads in it at launch](https://www.modernretail.co/technology/krogers-ai-shopping-assistant-has-ads-at-launch/?ref=giovanni-gallucci.ghost.io). Brian Spencer, marketing director for Kroger Precision Marketing, told Modern Retail that showing up in the assistant requires no additional setup by advertisers. If a brand already has a product listing campaign running with Kroger Precision Marketing, it may appear in the AI assistant's recommendations. Read that again. The one slot inside the assistant that is not somebody's purchase history populates itself from campaigns that were already running. If you were buying Kroger retail media in July, you were inside the agent in August and nobody had to call you about it. If you were not, there is no button for you to press. And the timeline is the tell. Walmart put ads in its Sparky assistant in January, a year after the tool launched. OpenAI started testing ads in ChatGPT in May, years in. Kroger shipped them on day one. The lag between a new discovery surface opening and that surface being rented out has gone to zero. Every feed you have ever lost went through this, and it used to take a few years. ## Store hours The most useful thing anybody said last week came from a regional grocer in St. Louis, and it is almost too dumb to print. [Schnucks cleaned up its data before launching its assistant](https://www.grocerydive.com/news/schnucks-ai-shopping-assistant-grocery-technology/831101/?ref=giovanni-gallucci.ghost.io) this summer. Caleb Carr, their senior director of data science and engineering, said that on top of unifying their data streams they had to add product information that was missing, go gather operational facts nobody had ever written down, and break internal jargon into terms shoppers would actually use. And when it went live, they realized they had never given the thing their stores' operating hours. A chain with real engineers and a real budget shipped an AI shopping assistant that did not know when its own stores were open. Multiple speakers at the show said the same four words about all of this: garbage in, garbage out. That is what the gap actually is. Not sophistication. Completeness. And completeness is boring, which is the good news, because boring is the only category of work a founder can finish on a Tuesday night without hiring a vendor. The jargon line is the one I would tattoo on the wall. Your internal vocabulary is invisible to the machine. If you call it your house blend and the shopper says medium roast, the field the assistant reads is blank. ## What to do this week, and all of it is free I am not going to pretend this is good news for a small brand. The mechanic pays incumbency and it pays it automatically. So these are all mitigate-and-respond, and none of them cost anything. **1\. Ask your existing buyers to put you in a history. Paste this.** *If you buy \[FULL PRODUCT NAME\] more than once, go set it as a reorder in whatever app you shop.* *Not a sales thing, and there's no code. Grocery and delivery apps are building assistants that fill the cart for you now, and the feature people actually use is the one that reorders what they already bought. So whatever's already in your history keeps getting bought, and everything else quietly stops getting seen.* *We're not in many histories yet. You can put us in one in about eleven seconds.* *If your app listens, the words are \[FULL PRODUCT NAME\], \[CATEGORY NOUN\]. That's what it matches on.* Three rules. Use the words a human would say out loud, not your SKU language. Aim it at people who already bought, because this is a retention mechanic and you should use it as one. Attach no discount, because you want this product bought on autopilot and a code teaches people to wait for the next one. **2\. The still.** Your own hand, your own phone, showing what a grocery or delivery app actually returns when you search your own category. Shot at the bench where you pack orders, with the tape gun and the boxes still in frame. Not a flat lay, not a cleared table, not a screenshot dressed up as a graphic. If the app got you wrong, that wrong result is the picture. The proof is that you went and looked. **3\. The video, five shots, 15 to 30 seconds. The jargon test.** Shot one, zero to five seconds, you talking to the camera: "I make \[FULL PRODUCT NAME\]. It's a \[CATEGORY\]." Say both out loud. Shot two, over your shoulder at the phone, asking a grocery app for that category. Shot three, the result, unedited, including the part where you are not in it. Shot four, your hand writing on whatever paper is on the bench: the words it matched, the words it missed. Shot five, you saying the full product name again, because search is the one route nobody can outbid you on. No burned-in captions. Do not reshoot the search to get a better result. **The free internal move.** Write down the five phrases a shopper would say out loud to find something like what you make. Then open your own product titles, your website listings and every retailer data sheet you have filed, and search for those exact phrases. Every phrase that is not in there is a field the assistant reads as missing, and missing reads as nothing. Count them. That number is your backlog, and it costs you an afternoon and no money. Two ways to get this wrong. Do not pick a fight with the app, because nobody has ever rooted for a brand yelling at software. And do not stage a screen you did not get, because anybody with the same app checks it in twenty seconds. Seventy-four percent will hand over the errand. Thirty-two percent will hand over the choice. Go find out which words the machine already knows you by, and spend the rest of the quarter on the people who would say your name to it without being asked. This piece is the extension of this week's episode. The show is the briefing. You can subscribe to The Food and Beverage Trend Report on [Apple Podcasts](https://podcasts.apple.com/us/podcast/id1896775833?ref=giovanni-gallucci.ghost.io) or [Spotify](https://open.spotify.com/show/033jPRykJcvZDhtgSwZBhf?ref=giovanni-gallucci.ghost.io), and the back catalog lives on the [show page](https://giovanni-gallucci.ghost.io/food-and-beverage-weekly/). If your product is better than your feed, [come find me on LinkedIn](https://www.linkedin.com/in/galluccinet?ref=giovanni-gallucci.ghost.io).