Starbucks discontinued its AI-powered “Automated Counting” inventory system across North America less than a year after deployment, reverting back to manual inventory processes. The tool, developed with NomadGo, used computer vision and spatial intelligence to automate stock counting for products like milk and syrups.
- The system reportedly struggled with miscounts and product mislabeling, including confusion between similar inventory items.
- Starbucks had initially claimed the tool enabled inventory counts to occur 8x more frequently across stores.
- The rollout had reached more than 11,000 North American stores before being retired.
- Employees reportedly had to manually verify many AI outputs, eliminating expected efficiency gains.
- The case is becoming a visible example of the operational gap between controlled AI demos and real-world enterprise deployment.
Source: Restaurant Dive, Reuters