Taichung buffet restaurant uses AI checkout to stop being "ripped off", owner says customer complaints dropped by 90%

A buffet restaurant in Taichung has introduced an AI image recognition pricing system, with the owner reporting a 90% drop in customer complaints after a three-month trial. The global AI-driven retail visual checkout market is expected to reach $3.99 billion in 2025 and exceed $5.05 billion by 2026. (Context: University of California study on "AI brain fog": 14% of office workers are driven crazy by Agents and automation, with a 40% higher intention to quit.) (Background: Y Combinator startup guide interpretation: What are the future development trends for AI Agents?) A buffet plate weighing 600 grams and worth 250 NTD might have previously sparked a pricing dispute (or led you to silently complain that it was too expensive!). But with the help of AI, such disputes may be significantly reduced in the future. According to a CTS report, this buffet restaurant in the Nantun District of Taichung City integrated an AI image recognition pricing system three months ago: customers place their trays on a scale, a camera captures the items simultaneously, the system automatically identifies the dishes and calculates the cost based on weight, and the screen instantly lists the unit price and total amount for each dish. Mr. Hsu, the owner, stated that since the system went online, customer complaints have decreased by 90%. In the traditional model, "manual visual estimation" is the basis for billing. The owner has one standard in mind, while the customer has another; the gray area between the two is a breeding ground for complaints. Mr. Hsu admitted that no matter how well employees are trained, manual pricing inevitably leads to errors. The error itself does not necessarily constitute a loss, but it creates a sense of uncertainty, making consumers wary every time they check out. One customer who tested the system said: "Knowing exactly how much I've taken and how much it costs makes me feel more at ease." The logic behind the AI image recognition pricing system is not complex, but the difficulty in implementation lies in the irregular shapes of food, the wide variety of items, and the different ways they are stacked. Some netizens speculated that the pricing logic is related to "area ratio + weight estimation," leading to discussions about whether "higher piles of food result in different prices" and even heated debates about "meat-hiding techniques." The AI retail visual checkout market is already heating up globally Analysts estimate the AI-driven retail visual checkout market will reach $3.99 billion in 2025 and $5.05 billion by 2026, with a compound annual growth rate of 26.6%. Faced with rising labor costs and severe labor shortages, business owners are increasingly interested in any technology that can reduce reliance on manual labor. The British company Winnow is an early player in this space. Its camera system can identify over 1,000 types of food and is primarily used in five-star hotel kitchens. Its focus is not on pricing but on reducing food waste; the company claims it has helped clients reduce food waste by approximately 50%. While the two application scenarios seem different, the underlying logic is the same: using machine vision to replace human judgment, turning "what the food is and how much there is" into quantifiable data. Some people online joked that with AI replacing manual visual pricing, "the aunties might lose their jobs." However, what these systems currently replace is mainly the "pricing judgment" step, rather than the overall labor involved in serving food, preparing meals, or maintaining the dining environment. Pricing is the easiest part of the buffet service process to standardize and the most prone to disputes. The machine's advantage here is clear: it does not give different prices based on mood, fatigue, or subjective impressions. In the short term, the jobs of the aunties cannot be replaced, but in the future, fewer people will certainly be needed for the pricing or ordering process.