Starbucks Scraps Inventory Management AI 9 Months After Introduction

Starbucks introduced an AI tool called Automated Counting to automate inventory management tasks. The tool used an iPad camera to identify items on shelves and count quantities, aiming to reduce a task that used to take 1 hour down to around 10-12 minutes.

However, the following serious technical flaws occurred in actual field use:

• Visual errors: Reflections of light off refrigerators were mistaken for actual products, causing duplicate counting
• Identification errors: Misclassifying types of milk or confusing items such as syrups
• Network dependency: In environments with unstable Wi-Fi connections, data was lost and work entries were deleted

This project reportedly consumed more than $10 million (about 13 billion KRW) in development and deployment over several years, but was ultimately scrapped entirely after 9 months. Starbucks has now returned to its previous manual inventory management method.