On the last day of March, Shanghai is overcast, hovering just above 15 degrees Celsius. Today is World Backup Day, a reminder to back up your data. But perhaps what you should really be backing up is your job security.
An Uncomfortable Statistic
Yesterday, TechCrunch published a piece citing a Quinnipiac University poll conducted between March 19 and 23: among 1,397 US adults, only 15% would accept a job where “their direct supervisor is an AI program that assigns tasks and schedules shifts.” 70% worry that AI will reduce employment opportunities. Among the employed, 30% are “very concerned” or “somewhat concerned” about AI eliminating their jobs.
To be honest, 15% is higher than I expected. Think about it, it means one in seven people is saying, “Sure, let AI manage me.” Considering the survey asks about a rather extreme scenario—not “would you use AI tools to assist your work,” but “would you let AI be your direct boss”—this ratio is quite intriguing. It shows some people no longer care whether their manager is carbon-based or silicon-based. Maybe they’ve been deeply hurt by human bosses, or maybe they genuinely believe AI distributes tasks more fairly than humans.
But the truly interesting part of this poll isn’t the 15%. It’s the tension that arises when viewed alongside another set of data: Gartner predicted in an October 2024 report that by the end of 2026, 20% of companies will use AI to “flatten” their organizational structures, cutting over half of existing middle management positions. Korn Ferry’s 2025 survey also corroborates this trend—41% of employees say their companies are already reducing management layers.
On one hand, the vast majority says, “I don’t want to be managed by AI.” On the other hand, companies are saying, “But we are going to use AI to lay off the people managing you.” Nobody is asking if you’re willing; the gears are already turning.
What Exactly Are Those Being Laid Off Doing?
To say something that might offend people: the daily work of many middle managers can be summarized in one phrase—information couriers.
Meetings are held above, and you pass the spirit down. Work gets done below, and you summarize the progress up. In between, you coordinate cross-departmental wrangling, push delayed projects, and approve a few expense reports. Fortune magazine published an article last year directly titled Surviving the Great Flattening: The coming extinction of the middle manager. The author, George Pesansky—a veteran transformation consultant for Fortune 500 companies—put it bluntly: if your competitive edge is “I know more than others,” prepare to be flattened. Because AI will soon make everyone know just as much.
Someone named Keith Anderson on LinkedIn compiled a list of middle management functions AI is taking over: information carrying, routine processing, and operational coordination. His exact words were: “If your primary function is compiling weekly reports, scheduling meetings, and passing data between superiors and subordinates, you have become an expensive piece of infrastructure.” It sounds harsh, but think about it—an AI agent can indeed do these things in seconds. It doesn’t need coffee, doesn’t take time off, and doesn’t zone out after 3 PM on a Friday.
Bloomberg and Live Data Technologies conducted an analysis: in 2023, middle managers accounted for a third of all layoffs. It’s not that a third of middle managers were fired, but rather one in every three people on the layoff lists was a middle manager. This ratio, how should I put it, is no longer just “being affected.”
But the Lesson of Klarna is Right There
Talking up to this point, it’s easy to fall into the narrative trap of “AI is omnipotent,” so we need to pour some cold water on it.
The most frequently cited success story for AI replacing human labor is Klarna. In 2024, the Swedish payment company partnered with OpenAI to deploy an AI customer service assistant. In its first month, it handled 2.3 million conversations, equivalent to the workload of 700 full-time agents. Resolution time dropped from 11 minutes to 2 minutes. Klarna excitedly announced this would bring a $40 million annual profit boost.
And then what? Customers started complaining. When faced with complex problems, the AI provided scripts, not solutions. In scenarios requiring human judgment like fraud, payment disputes, and financial hardship, describing the AI’s performance as “crashing” wouldn’t be an exaggeration. A Forrester survey showed that 55% of employers already regret laying off workers because of AI, predicting that by 2027, half of the “AI-attributed layoffs” will be reversed. Klarna itself also quietly started rehiring human customer service agents, pivoting to a human-machine hybrid model.
