Starbucks is “disciplined about where automation adds value.” |Photo: Shutterstock
When Starbucks announced its AI inventory calculator in early September, it seemed like a great example of how restaurants can benefit from AI.
It was practical, concrete, and clearly impactful. According to Starbucks, by using iPads with computer vision, employees can now review inventory in minutes instead of an hour, and eight times more frequently than before.
And importantly, it wasn’t just a pilot in a few stores. As of early September, thousands of Starbucks stores had implemented AI for inventory management, and the rest of the chain was expected to do so by the end of the month.
Starbucks hasn’t been shy about this technology, which it developed in partnership with a company called NomadGo. A video was also created in which employees talk about how AI makes their jobs easier and allows them to spend more time in the store.
So it came as a surprise when Starbucks announced last week that it was ramping up its AI inventory system. The company said it wanted to standardize the inventory process for all products, and it appears that AI was only used for milk and other beverage items.
However, Reuters also reported that the AI is not working well. There were problems with counting and labeling, which resulted in inaccurate numbers and extra time for staff to correct mistakes.
Social media posts by employees That seems to have been borne out in recent months. One person complained on Reddit that counting using AI takes twice as long as doing it manually.
And Starbucks more or less accepted that the old-fashioned pen-and-paper method of inventory management was more effective. “We continue to use technology across our operations,” the company said in a statement to sister publication Nation Restaurant News. “This change reflects discipline around where automation adds value.”
This isn’t just a Starbucks problem. Across industries, there are examples of AI failing to deliver on its promise of increasing efficiency and reducing costs. In a survey of limited-time restaurant chains, In a March survey by POS company Qu, only 9% said AI was having a significant impact on their business, and 43% said they saw only limited value.
And in some cases, AI may be actively making things worse. For example, earlier this month, a major Pizza Hut franchisee sued the company over its requirement to use AI-based software called Dragontail, which is designed to help restaurants more effectively manage delivery orders.
The system was counterproductive, leading to longer delivery times and a significant drop in sales, according to the complaint. Franchisees argued that manually entering third-party delivery orders into the POS, as they had previously done, was better for their business than an automated system.
When operators say they prefer “tablet hell” to AI, something is wrong. (Pizza Hut declined to comment, citing pending litigation.)
These AI challenges aren’t just for restaurants. AI will impact productivity across industries. There were various situations. Some studies also show that while AI can help save time for some workers, it can also create more work for others. Harvard Business Review calls this “AI Worklop” and defines it as “AI-generated work that pretends to be good work, but lacks the substance to meaningfully advance a given task.”
An HBR survey of 1,150 U.S. workers found that 40% had dealt with a backlog of work in the past month. (Maybe some of them were Starbucks baristas?)
Of course, we’ve heard about the success stories of AI in restaurants, some of which we’ve covered in this column. AI can predict how many donuts a restaurant will sell or Analyze business data Doing this manually would have taken a long time. That’s great!
Then again, just a week ago, Starbucks would have been counted among the success stories. So we no longer know what to believe.
This gap between what AI is supposed to do and what it actually does is likely the result of misaligned priorities. There is a lot going on right now with the success of AI. Investors are pouring billions of dollars into AI companies. Technology companies like Meta and Amazon are restructuring their businesses around this. Boards are asking, “What is our AI strategy?”
All of this puts pressure on AI companies to sell more AI and on AI buyers to do more with AI.
As a result, a square peg is pushed into a round hole. So even when this technology reaches restaurant operators in the field, it doesn’t work as intended.
This was a hot topic at last week’s National Restaurant Association show. Many operators said they were skeptical of all the AI on display.. They felt much of it was marketing bullshit and questioned whether it worked as advertised, so some resorted to building their own AI tools. The Starbucks news proves exactly why they feel that way.
This doesn’t mean AI doesn’t work or never will work in restaurants. There are pockets that you probably already have. But this business is fundamentally physical and human in nature, and AI still has much to learn in these two areas.
Before the AI boom, restaurants were doing well (relatively). Like any technology, it’s a helpful tool. But there are forces at work in restaurants beyond their control that it’s starting to feel like an obligation.
The situation at Starbucks is in desperate need of a reality check. As the chain suggested, restaurants need to think about where AI will actually bring value and where it will just create more work.
