Starbucks ends AI inventory programme after nine months amid accuracy concerns

Starbucks has discontinued its AI-powered inventory management system across North America less than a year after its launch, following reports of persistent inaccuracies and operational challenges.

The Automated Counting programme was introduced in September 2025 as part of Starbucks’ wider effort to improve store efficiency and reduce administrative tasks for staff. Developed in partnership with Seattle-based technology company NomadGo, the system combined computer vision, augmented reality, LiDAR sensors and 3D spatial intelligence to monitor stock levels in real time.

The initiative was intended to help stores identify shortages more quickly and ensure that popular menu items remained available to customers. However, despite initial optimism, the technology struggled to perform reliably in busy retail environments, leading Starbucks to abandon the programme after only nine months.

Inventory technology struggled in real-world conditions

Starbucks originally promoted the system as a way to improve product availability and streamline store operations. The company said the goal was to ensure customers could order any item listed on the menu without encountering stock shortages.

“Our goal is simple – if it’s on the menu, customers should be able to order it,” Starbucks said.

The technology relied on store employees using tablets to scan shelves and storage areas. The software then analysed stock levels using computer vision and sensor data. However, reports from stores suggested the system frequently produced inaccurate results.

Employees found that the software often overcounted inventory, failed to detect products and incorrectly identified items. One commonly cited issue was its inability to distinguish between similar products, including cartons of whole milk, oat milk and almond milk.

The challenges became more apparent as the programme expanded across the company’s North American network. In one promotional demonstration, the application reportedly failed to recognise a syrup bottle, highlighting concerns about its reliability.

Store workers also reported that the scanning process itself was cumbersome. Staff often had to move tablets at specific angles or wave them around shelves to properly activate sensors. As a result, some employees found that manually recording inventory was faster and more accurate than relying on the automated system.

Following the programme’s cancellation, Starbucks removed online content that had previously highlighted the technology’s benefits.

One employee welcomed the decision, writing: “Thanks for discontinuing Automatic Counting. The thought behind it was great, but the execution was proving difficult.”

Company shifts back to manual stock checks

With the AI system no longer in use, Starbucks has returned to manual inventory counting while continuing efforts to improve stock management across its stores.

The inventory programme had formed part of the company’s broader “Back to Starbucks” transformation strategy announced in early 2026. The initiative aims to increase revenue, improve comparable store sales and support the opening of more than 2,000 net new stores worldwide, including approximately 400 locations in the United States.

Although the automated counting system failed to meet expectations, Starbucks has remained committed to improving operational efficiency and reducing product shortages. The company continues to focus on more frequent store replenishment practices to help ensure ingredients and products remain available when customers place orders.

The difficulties Starbucks has experienced also highlight broader challenges facing computer vision technologies in retail settings. Unlike text-based AI systems, visual recognition tools must operate in environments where products are tightly packed, labels can be obscured, and lighting conditions vary significantly.

These factors can make accurate product identification difficult, even for advanced systems. As a result, many retailers continue to face obstacles when attempting to automate inventory management using artificial intelligence.

Starbucks continues broader AI and technology investments

Despite abandoning Automated Counting, Starbucks has not stepped away from technology-driven improvements. The company continues to invest in digital tools designed to improve customer service and store operations.

One example is the Smart Queue system introduced under the Back to Starbucks strategy. The technology helps manage incoming orders from multiple channels, including in-store purchases, mobile orders, drive-through customers and delivery services. By prioritising orders more effectively, Starbucks aims to reduce waiting times and improve the customer experience.

The company has also reported encouraging financial results during its ongoing transformation efforts. In the second quarter, Starbucks recorded a 9% increase in revenue, reaching US$9.5 billion. Comparable store sales across North America also rose by 7.1%.

“We have more work to do, but we’re pleased to see the combination of our comp growth and cost discipline starting to show up in margins,” said CFO Cathy Smith.

Starbucks leadership continues to view artificial intelligence as an important part of future operations, despite the setback. The company has retained references to plans for using AI in areas such as supply chain management and employee scheduling.

While the inventory-counting project did not achieve its objectives, Starbucks appears committed to exploring other applications of AI that may deliver more consistent results at scale. The company will now focus on technologies that can support staff and improve efficiency without disrupting day-to-day store operations.