Thought Leadership
After just nine months, Starbucks is retiring the AI system it had deployed to count stock automatically. Technical failures had become a burden on day to day operations across its coffeehouses.
The American coffee chain is permanently ending the use of its artificially intelligent inventory measurement system across all of its North American stores. The operational decision comes only nine months after the technology was rolled out company wide as part of a broad turnaround campaign. The end of the program was made official through an internal company memo obtained by the news agency Reuters, the contents of which were independently confirmed by two company employees.
According to the instructions in the document, the automated counting process is being discontinued with immediate effect. Going forward, drink components and dairy products will once again be tracked using the conventional manual inventory methods that already apply by default to every other product category in the stores. The end of the project marks a strategic setback for the company’s modernization efforts, since the technology had been announced as a central element for boosting efficiency.
NomadGo Was Meant to Fully Replace Manual Stock Counts
The failed system was developed by NomadGo, a technology company based in Seattle. Its technical architecture relied on LiDAR sensors combined with the built in cameras of the tablet computers that serve as work tools in the stores. Employees were supposed to use the tablets to scan the storage shelves. The software was programmed to automatically generate precise counts of syrup bottles, milk cartons, and similar liquid ingredients.
Although development of the technology had begun before current chief executive Brian Niccol took office, he pushed the rollout across the North American coffeehouses shortly after assuming leadership of the company in September 2024. The primary goal of the automation was to fully replace the previously time consuming manual stock counts carried out by baristas. Management expected the system to deliver significantly faster recording speeds along with a clear reduction in human error during daily demand assessments.
The AI Failed at Basic Product Identification
In everyday store operations, however, significant functional shortcomings emerged. As investigations by industry analysts revealed, the artificial intelligence struggled with even the basic task of identifying products. The camera system showed a high error rate and regularly confused visually similar packaging, such as different milk varieties and fat content levels. On top of that, stock already sitting on the shelves was systematically overlooked by the algorithms.
The system’s lack of maturity was, curiously enough, even on display in an official promotional video that Starbucks released during the launch phase. In one sequence of the marketing footage, the system scans a row of bottles on a shelf but skips a clearly visible bottle of peppermint syrup entirely, while the containers directly to its left and right are recorded correctly. Unreliable data of this kind led to miscalculations in follow up orders at the stores.
Starbucks Has Long Struggled with Product Gaps
In an official statement to the media, Starbucks management tried to frame the end of the program as a proactive move toward process standardization rather than admitting a technological failure. The company justified the step by pointing to its decision to standardize the way inventory is counted across all coffeehouses, while continuing to focus on consistency and execution at scale. In parallel, the company is now pursuing a shift to daily restocking cycles and aiming to implement continuous improvements throughout its supply chain.
In this context, the leadership stressed its overarching objective with a clear directive: its goal is simple, in that if an item is on the menu, customers should be able to order it. A lack of availability of core products has weighed on the company’s balance sheet for years. Several successive chief executives blamed unexpected gaps on the shelves for noticeable revenue losses, which is why inventory automation had been positioned as a cornerstone of the Back to Starbucks turnaround strategy.
AI Systems Remain in Use for Drink Orders
The internal feedback from staff that was shared as the decision came down was considerably less diplomatic than the official press statements. On internal communication channels, numerous store employees expressed relief at the end of the automatic counting. One worker summed up the general mood in a message to management, thanking them for ending the system and noting that while the idea behind it had been great, putting it into practice had simply proven too difficult.
Despite the failure of the inventory counting effort, Starbucks is holding on to other digital infrastructure. The company’s broader technology push still includes AI driven systems for optimally sequencing complex drink orders, along with digital assistance tools meant to ease the burden on staff at the espresso machines during busy peak hours. NomadGo, the developer, said in a statement that it continuously learns from customer and user feedback in order to keep refining its own software offerings.
