Starbucks recently quietly stopped a set of artificial intelligence tools for store beverage inventory management in North America because the system frequently made "mistakes" in actual applications. Not only did it fail to reduce the burden on front-line employees, but it continued to make "hallucinatory" errors in the inventory of ingredients such as syrups.
One of the core goals of current Starbucks CEO Brian Nicol after taking office is to drive sales growth while increasing the company's profit margins. He tried to improve operational efficiency through a series of technological upgrades, but at least in the inventory management link, the AI solution did not bring the expected results.
According to reports, Starbucks recently informed employees in an internal communication that the company has decided to stop using an AI inventory tool called Automated Counting (AC), and that milk and other ingredients for drinks will return to the same manual inventory method as other products. The AC app was developed by NomadGo, a Seattle-based technology company, and had been in testing for years before Nicole took over Starbucks.
Nicole decided to promote and deploy AC in all Starbucks stores in North America in September last year, hoping to use this system to make inventory management faster and more efficient. The tool was originally designed to automatically count the inventory of syrups, milk and other beverage ingredients. Stores are equipped with tablet terminals with cameras and lidar sensors to identify products that need replenishment.
However, in actual operation, AC does not show its "intelligent" side. Internal sources revealed that the app frequently made mistakes when identifying specific items on the shelves. For example, it could not identify a bottle of mint syrup, but would "count out of thin air" the number of other bottles on the same shelf that did not exist. Therefore, employees described it as "hallucinatory" inventory. Starbucks subsequently confirmed to the media that the project had been terminated. The explanation given was that the company decided to "unify the inventory counting method" among stores in order to continue to strengthen the consistency and execution of large-scale operations.
In order to fill the gap left by the abolition of AI tools, the company is considering more traditional and reliable solutions, including increasing the frequency of beverage ingredient distribution and improving its own supply chain. Starbucks also shared screenshots of employee feedback internally. Many front-line employees welcomed the company's abandonment of this automated system and believed that returning to a simple and stable approach would be beneficial to work. NomadGo stated that it will continue to polish and improve its application products based on Starbucks’ experience and feedback.
In fact, AC is just one of Starbucks' series of technological attempts in recent years. In recent years, the coffee chain giant has been involved in high-tech concepts such as NFT digital collections and 3D printing stores. At the same time, it has also caused public controversy due to the alleged "dark mode" design in the application that helps the company retain a large amount of prepaid funds. However, these attempts have mixed performance in terms of actual business value and user reputation.
Starbucks' AI experiment at the store level ended with "phantom" inventory, which to some extent reflects the current practical difficulties faced by the retail industry when introducing cutting-edge technologies: in a complex and ever-changing front-line environment, over-reliance on immature intelligent systems is likely to neither achieve cost savings nor truly improve efficiency. Currently, Starbucks is returning its focus to standardized management and supply chain optimization, and the industry will continue to pay attention to how it will re-find a balance between technological innovation and store operations in the future.