When Starbucks scrapped an AI-powered inventory counting tool back in May, just nine months after revealing the new system, it landed as a surprise to those tracking the coffee giant’s high-tech ambitions. A new report from Fast Company tells the inside story of how the national rollout disintegrated — and why the Redmond, Wash.-based startup behind it was left “blindsided.”
Known as “Automated Counting,” the tool was built in partnership with NomadGo to scan backroom storage shelves using iPad Pros equipped with computer vision, spatial computing, and augmented reality. It was designed to automatically tally coffee bags, milk, syrups, and other key supplies.
The idea was to turn an hour-long manual chore into a 10-to-12-minute job so baristas could focus on making drinks and connecting with customers.
The technology was deployed rapidly across all 11,300 company-operated Starbucks locations in North America. But almost immediately, real-world store environments triggered rampant glitches, according to Fast Company.
Baristas reported camera errors — such as shiny refrigerator reflections doubling milk counts or the app misidentifying syrups and trash cans — while stores with spotty Wi-Fi frequently had their counting progress wiped out entirely mid-scan.
According to Fast Company, the technical breakdowns stemmed from both software limitations and outdated infrastructure. While NomadGo’s computer vision achieved 99% accuracy in controlled tests, CEO David Greschler noted that computer vision inherently struggles when inventory changes — requiring up to six weeks of retraining for seasonal holiday cups or limited-time packaging that NomadGo developers sometimes only learned about once items hit store shelves.
Compounding the problem, people involved in building the tool pointed to Starbucks’ backend network, which relies on a legacy IBM AS/400 system dating back to the 1990s, making it difficult for cutting-edge AI to process real-time store data reliably.
When Starbucks notified NomadGo on April 3 that it was pulling the plug, the startup was reportedly blindsided. Greschler called the decision “a complete surprise,” telling Fast Company that “there’s nothing you can do when leadership and strategy change.”
Within days of losing its centerpiece enterprise client, NomadGo was forced to lay off a large chunk of its 30-person workforce, according to the report, including the technical team that managed the Starbucks integration. Six weeks later, on May 18, Starbucks formally notified baristas that Automated Counting was retired, instructing them to rip the QR tracking codes off backroom shelves and return to manual tallies.
A Starbucks spokesperson provided GeekWire with this statement on Monday:
“Human connection is at the core of our business, which is why we have invested $500 million to put more partners (employees) in our coffeehouses. We use technology to support human connection, not to replace it. This tool was designed to simplify a routine task and give partners more time with their customers. When it fell short, we listened to feedback and changed course. That is what innovation looks like at Starbucks: listening, learning, and adapting.”
GeekWire also contacted NomadGo, and we’ll update this story when we hear back.
Despite retiring Automated Counting, Starbucks has pushed forward with other AI initiatives across its business. The coffee giant is building an AI-powered ordering companion inside its mobile app to translate cravings into custom recipes, while 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 generative AI virtual assistant built to help baristas quickly look up recipes, standards, and store operating procedures.
Read the full article here