FamilyMart uses AI to accelerate confectionery product development

FamilyMart is implementing artificial intelligence to accelerate its confectionery product development process, reducing the idea generation phase from approximately two weeks to a single day.

The company utilizes AI by feeding it extensive datasets, including two years of ID-POS (customer identification) data, sales figures, customer demographics, purchase timing, and repeat customer analysis. To ensure market relevance, the AI also incorporates industry trends from sweets market reports and internal group policy documents.

In one instance, the AI generated approximately 1,000 product ideas based on specific constraints and successful past product attributes. Human developers then reviewed these suggestions to select a few viable options. For the first AI-assisted product, ‘Oimono Canelé ~with Caramel Sauce~’, the AI provided feedback on flavor combinations and pricing; data suggested that keeping the price below 300 yen would improve trial rates, leading to a retail price of 285 yen.

While AI handles the initial brainstorming and data analysis, human staff remain central to the process. Developers intervene to ensure technical feasibility, such as modifying an AI-suggested internal sauce structure to a top-drizzled format to accommodate existing factory equipment and technical capabilities.