Starbucks Kills AI Inventory Tool After Nine Months of Milk Mix-Ups

A High-Profile Retreat

Starbucks has pulled the plug on its AI-powered "Automated Counting" inventory system across North American stores, ending a nine-month experiment that promised to modernise stock management but left employees battling persistent miscounts and mislabelled products. An internal company newsletter dated this week confirmed the move, telling staff that "Automated Counting will be retired" and that beverage components and milk would now be counted the same way as every other inventory category.

How the System Worked

Developed in partnership with technology firm NomadGo, the tool used tablets fitted with cameras and LiDAR sensors to scan shelves and automatically tally items such as syrups and milk. It was deployed across more than 11,000 company-operated locations. Starbucks launched the technology in September 2025 as part of CEO Brian Niccol's broader "Back to Starbucks" turnaround strategy, which aimed to address the product shortages he argued were hurting sales. At launch, the company said the system enabled inventory counts eight times more frequently than before and cut the task to under 15 minutes per session.

Milk Mix-Ups and Missed Items

The promise faltered in practice. Reports earlier in the year revealed that the tool frequently miscounted or mislabelled products, with similar-looking goods like different types of milk proving especially troublesome. In some cases, items were missed entirely. For a business where running out of an ingredient means turning away paying customers, those errors quickly eroded confidence in a system that had been pitched as a major operational upgrade.

Back to Manual Counting

In a statement, Starbucks said the decision was made to "standardize how inventory is counted across coffeehouses as we continue to focus on consistency and execution at scale." The company added that it is working towards more frequent daily replenishments and continued supply chain improvements, summing up its goal simply: "if it's on the menu, customers should be able to order it." NomadGo, which also supplies inventory solutions to restaurant chains including Taco Bell and KFC franchises, said it is "continuously learning from customer and user feedback" to improve its products. The retreat is a notable example of a high-profile AI deployment being rolled back after real-world results failed to match the marketing.