This article was written by Daniel Vitiello, CEO and co-founder of Cooklist, an agentic AI platform that combines meal planning, personalized search and AI-driven product bundling to simplify shoppers’ path to checkout and increase basket size.
Dominant technologies drive sea changes in the way that consumers obtain information and make purchase decisions. The World Wide Web created e-commerce. Google’s rise drove search engine optimization and pay-per-click keyword advertising. And smartphone adoption led to mobile apps and anytime/anywhere shopping.
Now, large language models, or LLMs, are doing it again, but far more quickly and consequentially than their predecessors. And a big part of the change has to do with how much more LLM users are writing in prompts. A 2023 study conducted by researchers at universities in the U.S. and Hong Kong found that ChatGPT queries contained more than three times as many words as Google searches.
What are shoppers saying with all of those extra words? More than half of respondents to a survey conducted by Semrush said they use AI when they begin looking for products. Another 52% specify constraints upfront such as budget, features or specific use cases. And 57% use AI to narrow down their choices.
It isn't hard to understand why this is happening. A search bar trains consumers to compress intent into succinct keywords that synopsize the essence of what they are looking for: “gluten-free pasta,” “fish patties,” or “Coca-Cola.” A conversational AI shopping interface, on the other hand, invites shoppers to explain their whole situation: what they want, why they want it, the conditions that matter most and the kind of results that would actually help. Here’s an example:
“I am planning a birthday party for my five-year-old daughter Ariel this Saturday. I need all the ingredients for an easy-to-prepare lunch for 20 kindergarten children that avoids nuts and other allergens. Maybe mac and cheese, but then an alternative for the ones who don’t like that, plus juice boxes, grapes and strawberries, and do you have ready-made cupcakes with sprinkles? Vanilla and chocolate please! Oh, and don’t forget disposable plates, tablecloth, birthday candles and hats. The Little Mermaid theme if you have it!”
This kind of “conversational shopping” has vital implications for grocers. For decades, the industry inferred customer intent through loyalty cards and basket analysis after the sale, and then reasoned backwards. AI inverts that. Shoppers are now declaring their intent before the purchase, without even being asked. It makes simple keyword search look positively primitive by comparison. Here are three ways that grocers can adapt:
Master your metadata
Shoppers’ declared intent is worthless if your AI can't act on it. To meet the higher accuracy expectations of conversational shopping, your product metadata may need an upgrade. Beyond standards like brand, price and category, conversational commerce requires machine-readable characteristics and details about ingredients, allergens, prep time, nutritional category, dietary fit and recipe use cases. Consider this shopper request, for instance:
“My family eats too much junk food! We are gaining weight and spending too much on snacks. Help me plan five healthy, high-protein dinners that I can cook in less than one hour each and that cost under $25/dinner. We are two adults and two hungry teenage boys. The guys love beef, chicken and pork, but I just became a vegetarian, so please offer non-meat alternatives for some of the meals! And include some greens.”
In a case like this, your results will be much better if your product metadata includes standardized details about food, ingredients, nutritional content, serving yields, prep time, package counts and recipe use cases.
Be better with bundles
Digital shoppers will no longer be content to simply write up their own shopping list and then go out and find everything they need item by item. Instead, they will expect to merely describe what they are trying to achieve in conversational language and have their grocer translate their broad-stroke requests into a checkout-ready basket. Begin thinking in terms of complete product “bundles” that are personalized according to each shopper’s preferences, quantity needed, and the recipe, family meal mission or other intent expressed in their search descriptions.
In the case of the metadata example above, for example, instead of serving up a list of meats, vegetables and grains to search through, a shopper might see the following meal bundle: pasta shells, tomato sauce, one pound ground beef, one quarter pound plant-based meat alternative for mom, a head of broccoli, three white onions, a stick of butter, two garlic bulbs and a bag of shredded cheese. This solution offers convenience without removing autonomy: shoppers can review, select or modify the bundles that they like.
Use shoppers’ “why” to buy (and to sell)
It’s no longer enough to make merchandising decisions based on what you think shoppers want. Now you finally can understand why they want those products in the first place. And you must, but not only in order to sell it to them at the moment. There is also a powerful opportunity to discover unmet needs and subtle shifts in consumer preferences in real-time, instead of waiting for quarterly or annual sales reports. Analyze shopper prompts for core motivations and emerging preferences, and cross-reference them with actual purchases. Then use this information to improve merchandise purchasing decisions.
The context of “why” that shoppers offer in their longer prompts is also helpful when you seek to personalize experiences. The conversation about Ariel’s birthday party makes it clear that cupcakes were only for a special occasion. But the conversation about high-protein dinners and vegetarian meal preference demonstrates a new shopping pattern, budget and dietary restrictions just in time, not after six months.
LLMs are driving a tectonic shift in how shoppers obtain information and make purchase decisions. Soon, e-commerce experiences that offer only a static search bar will seem as outdated and inflexible as “exact match” search or “desktop browser only” stores seem today. However, conversational commerce is not just a smoother search interface for shoppers. It also offers grocers the clearest and quickest read on shoppers’ “why.” This declared intent offers enormous benefits to grocers — but only if they adapt their digital experiences to listen.
