Why Starbucks is shutting down its AI inventory system just months after launch

For a company serving millions of coffees every day, running out of oat milk or caramel syrup is more than a small inconvenience. It can slow service, frustrate customers and hurt sales. That is why Starbucks decided to bring artificial intelligence into its stockrooms.

The idea sounded futuristic enough to impress any tech enthusiast: cameras, LIDAR sensors and AI-powered software scanning shelves and automatically tracking inventory in real time. No more manual counting. No more guessing whether a store had enough cold foam for the morning rush.

But after just nine months, Starbucks is abandoning the system and returning to old-fashioned manual inventory checks.

So what exactly went wrong?

What were Starbucks’ AI inventories?

The company rolled out its “Automated Counting” system across North American stores in September 2025. The tool was developed with tech firm NomadGo as part of Starbucks’ broader push to modernise store operations using AI.

More from Tech

Employees used handheld devices equipped with cameras and LIDAR technology to scan shelves and storage areas. The software would then identify products and calculate stock levels automatically.

The goal was straightforward: reduce the time staff spent counting inventory and improve supply chain efficiency. Starbucks also hoped the system would help tackle persistent shortages of popular ingredients, an issue chief executive Brian Niccol had linked to weaker sales performance.

At the time, the company heavily promoted the technology as a way to make stores run more smoothly. Starbucks executives argued it would free employees from repetitive backroom tasks and allow them to focus more on customer service.

The failure

In practice, however, the AI system struggled with something humans usually manage without much trouble: recognising different cartons of milk.

According to Reuters report, the software frequently confused similar-looking products, miscounted items or failed to identify them altogether. Variants such as oat milk, almond milk and dairy milk reportedly caused repeated problems.

For baristas already juggling mobile orders, long queues and customised drinks, the added frustration of correcting AI mistakes quickly became exhausting.

Employees appear relieved to see the system disappear. Some workers reportedly said the idea behind the technology was promising, but the actual implementation created more complications than convenience.

The shutdown is particularly notable because Starbucks had only recently claimed the software was improving inventory management. As recently as February, the company continued praising the technology publicly.

Still, the AI experiment reflects a larger trend across the retail and restaurant industries, where companies are rushing to automate routine work using artificial intelligence tools.

What happens now?

Starbucks stores will now return to manually counting inventory, the same method employees use for other stock categories.

An internal company communication reportedly informed workers that beverage ingredients and milk products would once again be counted by hand rather than through automated scanning, reports Reuters.

The reversal comes during a broader turnaround effort led by Brian Niccol, who joined Starbucks in 2024 after successful stints at Chipotle and Taco Bell.

Niccol has pushed aggressively to improve efficiency and revive growth through operational changes and AI-driven tools. Investors have largely responded positively, with Starbucks shares rising this year despite ongoing profitability pressures.

But the failed inventory system is also a reminder that even in the AI era, replacing simple human judgement is not always as easy as it sounds.