Lawson taps AI for pickle-flavored lemon tart

Japan's convenience store chains compete on a constant churn of limited-time items, and both Lawson and FamilyMart are now testing AI as an idea generator rather than a sales forecaster. Lawson's approach is deliberately detached from past sales figures or market data, with the stated goal of generating ideas unconstrained by existing thinking, which is why a pickle-flavored lemon tart exists at all. FamilyMart's path is more incremental: it fed sales data into AI, and after a July test of a biscuit sandwich filled with thick milk cream proved a success, it decided to properly pursue AI in product development and plans more such items in the coming months.

Notably, neither company has handed product development entirely to machines. Human specialists remained involved at FamilyMart, and AI was used primarily for brainstorming. That division of labor is the practical takeaway: AI is being asked to widen the range of ideas, while people still judge what can actually be made and sold.

The stakes hinge on whether shoppers respond to novelty for its own sake. FamilyMart's upcoming nationwide canele launch, following its successful July test, will be one signal, while Lawson's Tokyo-area rollout tests whether an AI-generated theme like 'What on earth is this? But it's delicious.' translates into repeat purchases rather than one-time curiosity.