At a Lawson store in Tokyo, a lemon tart topped with pickles is drawing curious glances. It is one of the latest experiments from Japan's convenience store giants, who are turning to artificial intelligence to dream up desserts that shoppers never knew they wanted.
FamilyMart and Lawson, two of the country's largest chains, have begun using AI to generate product ideas, hoping to break out of the familiar flavour combinations that dominate the market. The move comes as AI's role in the workplace sparks anxiety across industries, but here the technology is being used as a brainstorming tool rather than a replacement for human creativity.
“AI supplied the ideas, but humans still had to turn them into something worth eating,” said a spokesperson for FamilyMart, which fed years of sales data into its system to develop a French-style canelé made with sweet potato purée and topped with caramel sauce. The chain said it wanted to create something novel yet commercially viable.
In July, FamilyMart test-launched its first AI-assisted product: a biscuit sandwich filled with thick milk cream. The response was positive enough to encourage further experiments.
Lawson took a different route. Instead of relying on historical data, the company asked AI to propose combinations “unconstrained by existing ways of thinking or market trends,” according to a statement. The result included a lemon tart paired with pickles, a bread roll filled with red bean paste, margarine and yoghurt cream, and a Japanese-style salad presented like a parfait.
Human taste testers weigh in
For Shingo Shiina, a 64-year-old architect who sampled the pickle-topped tart at a Lawson outlet, the combination was initially baffling. “I was like ‘huh’ at first, but it actually goes pretty well, doesn’t it? The pickles. I’m surprised,” he said after tasting it. “It’s a strange flavour, but it’s tasty. It really is delicious.”
Shiina was equally impressed by the AI's originality. “You’d probably never come up with this, right? Pickles and this. I don’t think a human would ever think of that. That’s AI for you,” he added.
The experiments reflect a broader trend in Japan, where companies are exploring AI not just for efficiency but for creative leaps. The technology has already been used to design everything from train schedules to marketing campaigns, and now it is entering the culinary world.
Yet the role of human developers remains crucial. Both chains emphasise that AI only proposes ideas; it takes skilled food scientists and chefs to refine them into products that meet quality and taste standards. “The AI can suggest a combination, but it can't tell you if it will actually taste good,” noted a Lawson representative.
For European observers, the Japanese approach offers a glimpse into how AI might augment rather than replace human expertise in the food industry. As European retailers and manufacturers grapple with similar questions about automation and innovation, Japan's convenience store experiments could serve as a case study in balancing algorithmic suggestions with human judgment.
Whether these AI-inspired desserts become permanent fixtures or fade as gimmicks remains to be seen. But for now, they are drawing attention—and customers—eager to taste what the machines have cooked up.
