Starbucks damaged its reputation in South Korea by blindly following AI-generated advice. Meta alienated thousands of employees with an AI-driven reorganisation that caught everyone off guard. Indifference comes at a steep cost, writes columnist Simone van Neerven.
South Korea in rage
On 18 May this year, Starbucks launched its “Tank Day” campaign in South Korea to promote a new tumbler (coffee mug). Within hours, the country was in uproar. Customers smashed their Starbucks cups and mugs, many deleted the company’s loyalty app, government agencies cut ties with the coffee chain, and even the president weighed in, calling Starbucks “low-grade peddlers”. Revenue fell by 26% in just one week, and the CEO of Starbucks Korea was forced to resign.
The problem was that Starbucks had launched the campaign on South Korea’s national day of remembrance, known simply as 5/18. On 18 May 1980, a popular uprising against the military regime of Chun Doo-hwan began in the city of Gwangju. The army crushed the protests with brutal force, leading to a ten-day massacre. Official figures say around 170 people were killed, but independent estimates put the death toll at between 500 and 2,000.
The launch date was not the campaign’s only mistake. Its slogan, “tak on the table!” in Korean, also caused outrage. “Tak” was also the word used in a controversial statement given by police in 1987 about the death of student activist Park Jong-chul in custody. Police had said the activist collapsed and died after an interrogator slapped the table forcefully, when in fact the activist died after he was tortured. The phrase has since become a symbol of the authorities’ attempt to cover up what really happened.
Indifference and gullibility
It later emerged that the marketing team had used an AI tool to develop both the campaign and its slogan. The suggestions were adopted without considering their historical or cultural implications. To make matters worse, senior management approved the campaign without even opening and reviewing the attachment containing the marketing materials.
This highlights a common mistake when using AI: accepting its suggestions without properly checking or fact-checking the output. The Starbucks Korea case shows just how costly that lack of scrutiny can be.
More AI-blunders by Starbucks
In September 2025, Starbucks introduced an AI-powered inventory management tool across more than 11,000 stores in the United States. Using a tablet, employees could scan shelves, after which the software automatically counted items such as milk cartons, syrups, and other stock. Starbucks claimed the system was 99% accurate and could complete stock counts up to eight times faster than staff.
But just a few months later, in May 2026, Starbucks pulled the plug on the initiative. The problem was that the system had been trained and tested under ideal conditions: neatly organised shelves, good lighting, and orderly storage areas. In real-world stores, however, conditions are far less consistent. The software mislabelled products, confused different types of milk, skipped items, and recorded stock incorrectly. This led to disrupted inventory management, empty shelves, and growing frustration among baristas, who were left to deal with the errors.
In the rush to embrace the AI trend, many organisations are keen to position themselves as early adopters. This example shows that deploying AI tools too quickly, without sufficient real-world testing or input from employees, can easily lead to serious operational problems.
Chaos and discontent at Meta
Meta has not been immune to similar issues. In mid-June 2026, Meta CTO Andrew Bosworth circulated an internal memo in which he openly acknowledged that the leadership team had made a big mistake with the company’s so-called “AI reorganisation”.
In May, Meta cut around 8,000 jobs, roughly 10% of its workforce. The company said the restructuring was necessary to fund its multi-billion-dollar investment in AI. At almost the same time, about 7,000 other employees received a very different message: they were informed by email that they had been selected for a new internal AI initiative. However, this reassignment had been made without their prior knowledge or consent.
The reorganisation followed another controversial decision a month earlier, when Meta introduced software to record employees’ keystrokes and mouse movements to collect data for training AI models. The move sparked widespread dissatisfaction among staff. On internal channels, many voiced their concerns, with one of the most upvoted comments reading: “This makes me super uncomfortable. How do we opt out?” Bosworth responded bluntly: “There is no opt-out on company laptops.”
Employees reacted strongly, and a petition against the tool quickly gained widespread support. Meta backed down and decided in late June to temporarily pause the programme.
Lowest morale in years
Employees have also made their feelings about the restructuring very clear. On Reddit, one employee wrote that the selection process for the new AI unit appeared arbitrary. Others were even more outspoken, describing the work as “a soul-crushing gulag”. Cynical memes circulated widely on internal Slack channels, and several employees spoke of a “deep and growing sense of fear” within the company.
During one of the weekly ‘Tuesdays with Boz’ sessions, Bosworth acknowledged that employee morale had fallen to a near-record low. “Maybe not the absolute lowest in the last 20 years, I think that was during the Cambridge Analytica scandal in 2016, but it comes pretty close”, he said.
Human touch
Meta aimed to scale up its AI efforts rapidly to stay ahead of the competition. However, the combination of a rushed reorganisation and a lack of transparency led to confusion, uncertainty, and growing resistance among employees. As a result, trust in leadership was significantly undermined.
Meta’s AI strategy, therefore, appears to stumble not so much on the technology itself, but on the human side of change. This case illustrates how poor communication, limited transparency, and top-down decision-making can quickly lead to chaos and widespread dissatisfaction among staff.
Joining the AI craze
The examples of Starbucks and Meta show that the biggest failures are rarely caused by the technology itself, but by human decisions. AI recommendations are accepted without question, systems are rolled out before they have been properly tested in practice, and employees are often left out of the process. The technology is rarely the main problem; it is usually how organisations choose to use it.
Yet there is a clear sense of overexcitement. Wim T. Schippers would probably have called it an “AI craze” [author’s note: AI-gekte]. The fear of falling behind pushes organisations to move faster than is sensible, with predictable consequences: rushed decisions, costly mistakes, and, above all, exhausted employees.
In a world that is speeding up, slowing down is sometimes the most forward-thinking choice. But who dares to hit the brakes when everyone else is accelerating?
This article was originally published in Dutch on MT/Sprout, the most popular business and management platform in the Netherlands.
don’t miss out!
get my columns straight into your mailbox: