At Groceryshop, grocery and delivery executives made one thing clear: agentic AI is moving past the chatbot demo. Kroger, Giant Eagle and DoorDash are building assistants that fit into existing shopping routines, close the gap between inspiration and basket, and avoid the interface fatigue that comes with popping up a chat window.
Practical prompts over intrusive chat
DoorDash co-founder Andy Fang said users often don’t know what to ask a bot when a chat window appears, so they ignore it. The delivery platform is instead building intelligence into the browsing journey, such as a one-tap prompt to “reorder my usuals” or tools that suggest alternatives when an item is out of stock.
That reframe changes how AI should be designed: not as a separate destination, but as support inside the flow a shopper is already using.
Turning inspiration into a basket
Kroger’s chief digital officer Yael Cosset framed the assistant’s job as reducing the distance between meal inspiration and basket creation. Shoppers can add personal context like a photo of a handwritten list or a family recipe, and the assistant builds a cart that fits those constraints.
The retailer says it is layering its personalization capability into the tool, and it creates a new surface for CPG brands through retail media and recipe content.
Omnichannel is the bigger prize
Giant Eagle is using agentic AI to convert single-channel shoppers into omnichannel shoppers. The reason is simple: its omnichannel shoppers have a 30% higher lifetime value than customers who buy only in-store or only online.
The retailer has also reorganized to support that goal, combining ecommerce operations with marketing and merchandising teams last year. Giant Eagle’s Heather Feather said the website and app now receive as much weekly traffic as its physical stores, which changes where investment and priorities should sit.
The strategic lesson is that agentic AI is decision-making infrastructure, not a conversational gimmick. When designed natively, it reduces the thinking burden for busy families, creates higher-frequency baskets, and gives brands a new placement in the purchase path. For marketers, the metric to watch is not how many people talk to the assistant, but how often shoppers move from a prompt to a completed order with less manual effort.
What this means for marketing teams
- Agentic AI works best when it solves a specific, repetitive task rather than asking shoppers to learn a new interface.
- Personal context like lists, recipes and dietary needs can turn a generic recommendation into a ready-to-checkout basket.
- Backend team integration matters as much as the model: digital, marketing and merchandising need shared goals.
Kroger’s Cosset offered a useful guardrail for leaders under pressure to deploy quickly: speed with precision and direction is good, because speed alone brings risk.
Source: Digiday
