General Mills Uses an AI Stand-In for Dieters on Weight-Loss Drugs
Lisa is built from interviews, social posts, and sales data, and shows the average taste of her group rather than one real person.
The Journal. · How Big Food Is Navigating the GLP-1 EraGeneral Mills says it has created AI personas to advise its product teams, including one named Lisa who helped develop a brand aimed at GLP-1 users that will debut soon.
General Mills says the personas provide answers in a fraction of the time required for traditional testing.
Some foods carry GLP-1 friendly badges or callouts for protein or fiber, the attributes companies are highlighting to catch the attention of these shoppers.
Lisa stands in for women in their thirties and forties with a long history of dieting who are now on GLP-1s. She was built from interviews, social media posts, and sales data.
General Mills says the benefit is breadth as well as speed. Drawn from immense data, she reflects the average of her group rather than one specific person. That is also the tradeoff, an average can guide a recipe but cannot show how any real shopper will react at the table.
Food companies are increasingly using customer data to decide what to make. Conagra, for example, "sifts through credit card data and social media posts," according to reporter Amira McKee. The data includes late-night delivery orders, fast food menus, and Reddit conversations, which the company uses to see what consumers want.
The next thing to watch is that new brand, when it reaches shelves and shows whether an average drawn from data picked what real GLP-1 users want to eat.
Questions people ask
Why do GLP-1 users need different foods?
GLP-1 drugs, originally for diabetes and now widely used for weight loss, curb appetite and slow digestion. Users eat much less at a time, need protein to preserve muscle, and often report nausea. That is why food makers are emphasizing protein and fiber, and smaller but satisfying portions.
Do AI personas replace taste tests and focus groups?
Traditional focus groups bring a small number of people into a room for a few hours, so results can hinge on who happened to show up. An AI persona aggregates thousands of interviews, posts, and purchases, so it can answer questions in minutes. The tradeoff is that it shows the average preference, not how any real person will react in a taste test.