Starbucks scraps NomadGo AI inventory tool after 9 months of miscounts
AI promised 99% accuracy but couldn't count milk bottles correctly.
Starbucks has abruptly retired its AI-powered inventory management tool from NomadGo after only nine months, according to an internal newsletter obtained by Reuters. The system, which leveraged on-device 3D spatial intelligence, computer vision, and augmented reality to scan shelves in under 30 seconds, was frequently miscounting and mislabeling items — the very problem it was supposed to solve. CEO Brian Niccol, who joined from Chipotle in 2024, had championed the tool to address inventory shortages that were hurting sales.
The failure echoes a broader pattern of AI tools causing costly errors in the food and beverage industry. A Pizza Hut franchisee recently sued the chain’s parent company after an AI efficiency system allegedly caused $100 million in lost revenue. Starbucks will now revert to manual counting for all inventory categories. Meanwhile, the company continues testing other AI tools like Green Dot Assist for baristas and a ChatGPT-powered drink recommendation app, but the inventory debacle raises questions about the reliability of computer vision solutions in complex retail environments.
- Starbucks retired NomadGo's AI inventory tool after 9 months due to frequent miscounts and mislabeling of milk and beverage components.
- CEO Brian Niccol spearheaded the tool to solve inventory shortages; the tool claimed 99% accuracy but failed in practice.
- Similar AI failure cited: a Pizza Hut franchisee lawsuit over a $100M loss from an AI efficiency system.
Why It Matters
Highlights the gap between AI vendor promises and real-world reliability in complex retail operations.