Starbucks quietly pulled its AI inventory tool months after launch due to inaccurate counts and workflow disruptions.

(SeaPRwire) –   Starbucks has discreetly discontinued its AI-driven inventory management system just nine months after its launch.

The coffee chain confirmed to that it has opted operationally to return to a single inventory counting method, reversing a September announcement about deploying the automated counting tool.

Supplied by NomadGo, the application tracked stock levels of drink ingredients such as milk and syrups to monitor shortages. In February, Reuters—which first broke the news of the tool’s termination this week—quoted Starbucks sources stating the app frequently made counting or labeling errors, “hallucinating” stock by not detecting bottles on shelves.

“We experiment with concepts in our stores, pay close attention to partner input, and adjust to provide a better, more reliable experience,” a company representative said in a statement to .

NomadGo did not promptly reply to ‘s request for comment.

Carl Addison, a nine-year Starbucks shift supervisor in Shoreline, Wash., informed that the automated counting app forced stores to reorganize backroom storage, a lengthy undertaking. He noted the app’s mistakes complicated staff workflows. An overcount meant insufficient shipments of low-stock items, while an undercount resulted in inadequate supplies of needed products.

“It began with limited accuracy and deteriorated further over time,” Addison remarked.

Starbucks provided with several barista comments on the automated tool, some indicating it enhanced inventory procedures and the interface for checking stock levels.

“Thank you for ending Automatic Counting! The concept was excellent, but the implementation became problematic,” one response stated.

Brian Niccol’s ‘back to Starbucks’ plan

Starbucks has rolled out numerous AI tools under CEO Brian Niccol’s “back to Starbucks” initiative, aimed at boosting declining sales and optimizing operations. Recent AI introductions encompass Green Dot Assist, a store iPad app that offers recipe guides, suitable ingredient swaps, and equipment troubleshooting. Another is a Smart Queue feature that organizes orders to enhance speed and efficiency. Former CEO Laxman Narasimhan noted in early 2024 that customers were canceling mobile orders due to extended waits and item unavailability.

To date, Starbucks’ revival plan—which also involves installing more comfortable seating and simplifying the menu—seems effective. The firm recently posted a 7.1% rise in quarterly comparable U.S. sales, surpassing the 4.5% growth analysts predicted. Quarterly revenue grew 9% to $9.5 billion.

Retail’s automation challenges

Starbucks’ choice to reinstate its old inventory system highlights wider implementation struggles as the retail sector adopts AI. This month, a significant Pizza Hut franchisee filed a lawsuit against the chain regarding its AI system. The franchisee alleged that Pizza Hut’s Dragontail AI platform provided gig workers greater access to internal systems, allowing them to use the AI for personal gain—such as picking high-tip orders and grouping deliveries—leading to late orders and “cascading operational failures.”

With the global restaurant automation market projected to surge to $28 billion this year, these technologies face mounting pressure to perform.

Given the current stage of AI evolution, the difficulties retailers encounter in scaling the technology have prompted Santiago Gallino, a Wharton professor of operations, information, and decisions, to assert: “Currently, the hype outweighs the tangible benefits.”

“Many retailers feel compelled to claim they are engaged in AI projects and innovations, launching them before they can yield solid, actual returns,” he told .

Gallino praised Starbucks’ move to withdraw the automated counting tool. He mentioned inventory management remains a persistent retail problem, and while tech progress helps address significant inventory issues, optimization tools are not a cure-all.

Other firms, like Zara, have invested years honing their use of algorithmic tech. The fast-fashion chain introduced a microprocessor-based tagging system over ten years ago, using Radio-Frequency Identification (RFID) tags on inventory, which boosted count accuracy and simplified item tracking throughout its network.

Gallino suggests the Zara example is not primarily about tech delivering universal retailer advantages, but rather illustrates a company researching and adapting a technology to meet its unique requirements. Although retailers bear the responsibility to harness emerging tech, AI will only prove sustainable for these businesses if it provides a return on investment.

“One overarching theme I still find somewhat puzzling is how, on multiple fronts, [return on investment] often doesn’t appear to be a primary concern—relying on the promise that it will eventually pay off,” Gallino commented. “That focus can be lost amidst the hype.”

Addison, the shift supervisor, said that at this point, the technology does not optimally support barista workflows.

“I would embrace AI if I believed it functioned well, but I must admit…I just don’t think it’s a good match for a retail setting, where precision and speed are crucial,” he said. “And it simply doesn’t seem capable of meeting those needs for us.”