- Starbucks retires AI inventory tool after nine months due to real-world challenges
- The AI could not recognize or distinguish between inventory items, forcing manual intervention
- Other AI and technical improvements continue to roll out during the ‘Back to Starbucks’ transformation
Starbucks has officially ended its much talked about ‘Automated Counting’ AI warehouse program across all of its North American stores just nine months after it was launched in September 2025.
Developed with Seattle-based computer vision company NomadGo, the app was designed to use on-device 3D spatial intelligence, computer vision, augmented reality and LiDAR sensors to give stores real-time visibility into inventory shortages.
CEO Brian Niccol had hoped the tool would free baristas from unproductive administrative work to deliver more to customers, but reports suggest the technology ultimately failed to revolutionize the store manager.
Failed Starbucks AI system being pulled just months after launch
“Our goal is simple — if it’s on the menu, customers should be able to order it,” Starbucks said (via Pakinomist), which justifies the place of technology in its stores.
In early 2026, the global coffee chain announced its ‘Back to Starbucks’ transformation plan, targeting revenue growth, more comparable store sales and over 2,000 net new stores globally (including about 400 in the US).
The problems came when the technology was implemented and tested on a large scale, because stores found that the computer vision model struggled with basic spatial awareness and object recognition, often overcounting items, overlooking inventory or mislabeling products.
Most notably, the tool was unable to distinguish between similar items, such as whole, oat, and near milk cartons. Starbucks also inadvertently showed the app completely missing a syrup bottle in a promotional video.
Ultimately, workers complained about having to force the AI to read shelves by waving and angling tablets in certain ways to trigger sensors, which made it slower than entering the details manually in the first place.
Starbucks has since deleted the associated blog posts praising the tool for its effectiveness, marking a complete 180 from early positive narratives.
“Thank you for stopping automatic counting,” one worker wrote. “The idea behind it was great, but the execution proved difficult.”
The company has now reverted to manual stock control, but remains committed to a revamped, high-frequency store replenishment model to prevent customers from being met with out-of-stock drinks.
Although Starbucks is in the limelight for this particular example, the issues primarily revolve around computer vision and artificial intelligence outside the realms of text. In unpredictable, densely packed retail warehouses with hidden labels, lighting variations and other variables, automated systems still face major challenges.
Starbucks is still committed to change
Although CEO Brian Niccol had previously introduced this efficiency measure to address chronic product shortages and long wait times, Starbucks announced a 9% increase in Q2 quarterly revenue to $9.5 billion, as well as a 7.1% increase in comparable store sales across North America.
“We have more work to do, but we are pleased to see the combination of our growth and cost discipline starting to show up in margins,” added CFO Cathy Smith.
Another tech improvement that doesn’t seem to have gone away with the automation of stock checks is Starbucks’ new Smart Queue system, which was introduced in the company’s previous ‘Back to Starbucks’ plan. It is designed to balance and prioritize incoming tickets across in-store, mobile, drive-through and delivery orders to ensure customers are seen immediately.
Looking ahead, even though the inventory management tool failed, the company hasn’t removed mention of its AI plans, such as “leveraging artificial intelligence to support partners, including supply chain and planning tools.”
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds.
