Starbucks has officially pulled the plug on its AI-powered inventory management system after just nine months in the wild. The coffee giant confirmed to Fortune it made an operational decision to revert to a single model of counting inventory following widespread barista complaints about inaccuracies and workflow disruptions. Reuters first reported the discontinuation earlier this year, with sources saying the NomadGo-provided app often miscounted or mislabeled items—failing to even identify bottles sitting on shelves.

The Counting Catastrophe

The automated counting tool was designed to track beverage components like milk and syrups, helping stores manage shortages by monitoring inventory levels. According to Carl Addison, a shift supervisor of nine years based in Shoreline, Washington, the system created more problems than it solved. Stores were required to rearrange back-of-house storage—a time-intensive process that ate into staff hours. When the app overcounted products, it wouldn't trigger enough restocks for items running low. When it undercounted, stores received shipments of things they already had plenty of. "It started off not particularly accurate and got less accurate over time," Addison told Fortune.

Barista Feedback Was Brutal

Starbucks did share some positive barista responses about the automated counting tool, noting improvements in inventory processes and interface usability. But the public-facing comments paint a different picture. One barista's response cut through the corporate speak: "Thanks for discontinuing Automatic Counting! The thought behind it was great, but the execution was proving difficult." Addison was more blunt about where retail AI stands today. "I would love AI if I felt like it worked, but have to say... I just don't feel like it's a solid fit for a retail environment, where accuracy and speed are both really important," he said. "And it just doesn't feel like it can really deliver on those fronts for us."

The Bigger AI Bet at Starbucks

This stumble comes as CEO Brian Niccol pushes forward with his broader "back to Starbucks" turnaround strategy, which relies heavily on AI tools to reverse slumping sales. Other deployments include Green Dot Assist—an iPad app providing recipe cards, ingredient substitutions, and machinery troubleshooting guidance—and Smart Queue, a tool that sequences orders to improve speed and efficiency. Former CEO Laxman Narasimhan noted in early 2024 that customers were abandoning mobile orders due to long wait times and product availability issues. So far, the strategy appears to be working: Starbucks reported a 7.1% increase in quarterly comparable U.S. sales last quarter, beating analysts' expectations of 4.5%, with revenue climbing 9% to $9.5 billion.

The Retail AI Reckoning

Starbucks isn't alone in wrestling with AI deployment challenges. Earlier this month, a major Pizza Hut franchisee sued the chain over its Dragontail Artificial Intelligence system, claiming it gave gig workers visibility into internal systems that they exploited for personal benefit—selecting higher-tip orders and bunching deliveries, resulting in cascading operational breakdowns. As global restaurant automation is projected to balloon into a $28 billion market this year, the pressure to deliver results is intense. Santiago Gallino, a Wharton professor of operations, information, and decisions, offered a stark assessment: "Right now, there is more hype than actual benefit." He criticized retailers for rushing AI deployments before they can provide concrete returns on investment.

Lessons From Zara's RFID Journey

Gallino pointed to Zara as an example of how to do automation right—though not without patience. The fast-fashion retailer spent over a decade refining its Radio-Frequency Identification tagging system, ultimately improving inventory accuracy across its entire supply chain. "The Zara case study is less an argument about technology generating universal benefits to retailers, but rather an example of a company doing the research and iterating a technology's use to fit its specific needs," Gallino explained. The professor's bottom line on ROI expectations should sound alarm bells across the industry: "One general theme that to me is still a little bit perplexing... how return on investment seems to be not a main consideration—the promise that down the road all this is going to make sense. That is something that can be out of focus in the middle of the hype."

Key Takeaways

  • Starbucks' NomadGo inventory AI lasted only nine months before being retired due to chronic accuracy failures
  • The system miscounted bottles, over-shipped unneeded items, and under-ordered essentials—getting worse over time
  • Baristas report the tech required extensive workflow changes without delivering reliable results
  • Brian Niccol's broader "back to Starbucks" AI strategy is still moving forward with other tools like Green Dot Assist and Smart Queue
  • Wharton professor Santiago Gallino warns that retail AI deployments are plagued by hype exceeding actual benefit

The Bottom Line

Starbucks' inventory AI whiffed hard—but that's the game when you rush half-baked models into production environments where accuracy actually matters. The coffee giant's broader AI bet might still pay off, but this nine-month experiment is a textbook case of why retail automation needs more iteration and less press release-driven deployment.