Report: Starbucks ditches AI innovation tool and leaves first Seattle location ‘blind’.
When Starbucks scrapped an AI-powered inventory accounting tool back in May, just nine months after unveiling the new system, it surprised those tracking the coffee giant’s high-tech ambitions. A new report from Fast Company tells the inside story of how the national rollout fell apart — and why the Redmond, Wash.-based startup behind it was left “blindsided.”
Known as “Automatic Counting,” the tool was developed in partnership with NomadGo to scan indoor storage shelves using iPad Pros equipped with computer vision, spatial computing, and augmented reality. It is designed to automatically count bags of coffee, milk, syrups, and other essentials.
The idea was to turn an hour-long manual process into 10 to 12 minutes so baristas could focus on making drinks and interacting with customers.
The technology was quickly implemented in all 11,300 company-operated Starbucks locations in North America. But almost immediately, real-world shopping centers caused widespread problems, according to Fast Company.
Baristas have reported camera errors – such as a shiny fridge display that doubles the amount of milk or an app that doesn’t correctly identify syrup and trash cans – while shops with spotty Wi-Fi often count their progress completely removed during scanning.
According to Fast Company, the decline in technology stems from both software limitations and outdated infrastructure. Although NomadGo’s computer vision achieved 99% accuracy in controlled tests, CEO David Greschler noted that computer vision is inherently problematic when changing inventory — requiring six weeks of retraining for seasonal holiday mugs or limited-time packaging that NomadGo engineers sometimes learn just as items hit store shelves.
Adding to the problem, people involved in the tool’s development point to Starbucks’ backend network, which relies on a legacy IBM AS/400 system from the 1990s, making it difficult for advanced AI to reliably process real-time store data.
When Starbucks notified NomadGo on April 3 that it was pulling the plug, the startup reportedly shut down. Greschler called the decision “a total surprise,” telling Fast Company that “there’s nothing you can do when leadership and strategy change.”
A few days after losing a key business customer, NomadGo was forced to lay off most of its 30 employees, according to the report, including the technology team in charge of the Starbucks integration. Six weeks later, on May 18, Starbucks officially notified baristas that Automated Counting had been retired, ordering them to rip the QR tracking codes off the back shelves and return to manual counting.
A Starbucks spokesperson provided GeekWire with this statement on Monday:
“Human interaction is the core of our business, that’s why we invested 500 million dollars to put more associates (employees) in our coffee places. We use technology to support human interaction, not to replace it. This tool is designed to simplify the routine work and give our associates more time with their customers. When it fell short, we listened to the feedback and changed course. That’s what Starbus looks like learning and adapting. That’s what Starbu looks like, learning and adaptability.
GeekWire has also reached out to NomadGo, and we’ll update this story when contacted.
Aside from abandoning Automated Counting, Starbucks has moved forward with other AI initiatives throughout its business. The coffee giant is building an AI-powered ordering companion within its mobile app to translate wishes into custom recipes, while it’s testing a ChatGPT integration that suggests drinks based on a customer’s mood or outfit.
For store staff, the company continues to rely on Green Dot Assist, a virtual AI productivity assistant designed to help baristas quickly check recipes, standards, and store operating procedures.
