FamilyMart will launch its first AI-generated dessert later this month, marking a significant shift toward using generative AI to expedite product development and respond to changing consumer trends.
The company plans to release approximately 10 AI-driven products by the end of the fiscal year following successful preliminary trials.
According to FamilyMart, this initiative builds on a limited launch in July, where the strong sales of an AI-recommended cream sandwich encouraged corporate leadership to expand the technology across more product categories.
The first commercial result of this process is a sweet potato canele with caramel sauce, priced at 285 yen ($2.58).
FamilyMart’s development system analyses point-of-sale data along with product attributes to evaluate characteristics shared by high-performing items and to identify flaws in underperforming lines.
The company’s data analysis revealed that consumer demand in the dessert category favours portable, one-handed snacks that feature contrasting internal and external textures, such as a crispy caramel-coated exterior and a moist sweet potato filling.
The AI system also determined the retail price point after identifying a decline in trial rates for new dessert items priced above $2.72.
While competitor networks, including 7-Eleven and Lawson, use AI for inventory management and supply chain routing, FamilyMart is among the first to apply the technology directly to commercial recipe generation.
The company estimates that this technology will reduce workloads related to trend research, supplier meetings, and prototype planning by approximately 20 per cent.
However, internal development teams will retain final approval over all AI proposals to assess the feasibility of manufacturing and the sourcing of raw materials.
- Further reading: 7-Eleven taken to task over franchisee store takeovers.
