Starbucks has discontinued its AI inventory tool nine months after its launch due to persistent issues with milk identification.

      The chain is reverting to manual inventory counts throughout North America, concluding one of CEO Brian Niccol’s more prominent technology investments and adding another entry to the record of “enterprise AI pilots that failed to withstand real-world application.” As reported by Reuters and confirmed by the company, Starbucks has discontinued the AI-driven inventory tool it introduced in its North American locations last September.

      "Effective immediately, Automated Counting will be discontinued," a memo issued on Monday stated. "Beverage components and milk will now be counted like other inventory categories in your coffeehouse," meaning they will be counted manually.

      Developed by Seattle-based NomadGo, the tool utilized tablet-mounted cameras and LiDAR technology to scan shelves of syrups, milk, and other beverage ingredients, providing automatic inventory counts in place of manual counts for selected categories. It had been in development for years and was rolled out nationwide after Brian Niccol became CEO in September 2024 as part of his "Back to Starbucks" initiative.

      According to reports from Reuters and internal documents, the issue arose from the tool's difficulty in distinguishing between similar white liquids. The app frequently made errors in counting or labeling, especially with products that looked alike, such as oat milk and dairy milk. A promotional video released by Starbucks during the launch depicted the system failing to identify a bottle of peppermint syrup on a shelf while counting other nearby bottles, a mishap that seems more critical in retrospect.

      In a statement to Reuters, Starbucks portrayed the decision as a move towards standardization rather than a retreat. The company stated that it stemmed from “the desire to standardize how inventory is counted across coffeehouses as we strive for consistency and effective execution at scale,” adding that it is transitioning to more frequent daily replenishments and enhancing supply chain efficiency.

      An internal note from the company included a comment from an employee expressing gratitude for terminating the program: “The idea behind it was excellent, but the execution was challenging.” This decision is significant because inventory management was meant to be the straightforward aspect. Over five years, four CEOs of Starbucks have attributed lost sales to the company’s difficulties in reliably stocking stores. By early 2024, the company acknowledged that less than a third of deliveries to its distribution centers were timely and complete.

      Automated Counting was intended to provide the chain with the real-time store-level visibility it had been lacking and was a key operational fix for Niccol. Additionally, this decision comes at a time when the overall record of enterprise AI is beginning to appear less favorable than initial presentations. Last year, MIT’s NANDA initiative discovered that 95% of enterprise generative AI pilots yielded no measurable impact on profits and losses, despite an estimated expenditure of $30 to $40 billion, with only 5% advancing to production.

      While the Starbucks tool was not generative AI, the pattern of its failure is reminiscent: automating a deeply integrated, store-level workflow proved to be more challenging than the demonstrations implied. The financial landscape is sufficiently mixed that the decision could be interpreted in two ways. Starbucks recently reported its highest quarterly sales growth in two and a half years, and its stock has increased by 24% so far in 2026, yet operating margins in its main North American market have dropped to 9.9%, down from 18% two years ago.

      Niccol has continued to invest in other technological initiatives, including AI tools designed to prioritize orders and support baristas during busy periods. NomadGo stated to Reuters that it is “continuously learning from customer and user feedback” to enhance its offerings. The upcoming challenge will be whether manual counts and daily replenishments can achieve what the algorithm failed to do — keeping peppermint syrup available on the shelves.