Mars plans for AI agents that do the weekly grocery run

The plan rests on research Mars published with the data firms Kantar and Synthesis, built from more than 180,000 data points across four continents. Under that forecast, staples are bought in the background: an agent proposes the basket from what you like and what is running low, and you only approve it, while stores and apps shift towards entertainment and discovery, Consumer Goods Technology reports.

Mars is working on both halves. Step one is data: through deals with young tech firms, Mars wants to see how chatbots describe its brands today and where those answers get their facts. The next aim is to turn an agent’s discovery of a brand into a finished purchase. Second, it is planning immersive store sections with retailers. Reynolds also wants Mars’s data team talking directly to Walmart’s, and its connected-commerce work has sat with Publicis since last summer.

Paid placement inside chatbots is already on offer: ChatGPT’s Go plan card says that plan may include ads, as our ChatGPT page recorded on 31 October.

In the Mars research, the agent’s list needs only the shopper’s confirmation, so a snack that never makes that shortlist never reaches the moment of choice; that is why Mars’s first step is measuring how its brands show up in AI answers, ahead of any change to the shelf.