Starbucks Drops AI Inventory Counting System After Widespread Errors

Starbucks has terminated the use of its AI‑powered inventory tool, Automated Counting, which was developed with Seattle‑based startup NomadGo. The system employed iPads with vision‑computer technology and LiDAR sensors to scan shelves in the chain’s roughly 11,300 North American stores, aiming to cut a one‑hour manual count to about ten minutes.

The rollout encountered persistent operational problems: camera reflections caused duplicate counts, unstable Wi‑Fi halted progress, and the software struggled with product changes, requiring up to six weeks of retraining. Staff reported frequent miscounts of milk, syrups and other items, leading the company to revert to manual records. Starbucks said it will invest $500 million to boost staffing in stores and continue improving its supply‑chain processes. The project, part of the “Back to Starbucks” plan overseen by CEO Brian Niccol, is estimated to have cost more than $10 million.