Kim, 35, is a white-collar worker in charge of planning at a small-to-mid-sized company in Seoul. He used to have normal disagreements with his boss about strategy and execution. Now, those discussions have been replaced by something far more frustrating — a third party in the room that never gets tired and never admits it's wrong: Google Gemini.
Every time Kim submits a market analysis report, his boss inputs it into Gemini and instructs it to "find errors." Even when Gemini produces irrelevant critiques that completely ignore the company's internal circumstances or business background, the boss still demands revisions, claiming the report is "wrong."
"I have to find original materials and counterarguments to prove that 'your AI is wrong,'" Kim says. "The stress is extreme because responding to AI's critiques has been added to my existing workload."
He's not alone. A 31-year-old named Park working at a Seoul startup describes a similar dynamic — but from a different angle. After his boss began trusting AI answers more than his own explanations, Park started packaging his own judgments as AI output. He writes his logic and evidence, then inputs it into the AI with the prompt "Explain this in your own words," captures the AI's response, and sends it to his boss.
Content that was dismissed as "subjective" when he explained it directly is accepted if it appears to come from AI.
"It's more disheartening that my years of experience at the company are less trusted than 'my opinion disguised as AI,'" Park says. "I feel like a mere 'AI operator.'"
A survey of 1,250 knowledge workers across 10 countries found that 30% said they cannot perform tasks without AI, and 28% reported trusting AI more than their own judgment. This "reversal phenomenon" — where AI's generalizations are accepted without verification while human judgment based on field experience is ignored — is creating a new kind of workplace stress that didn't exist before AI became so capable.
For Kim, the most exhausting part isn't the extra work. It's the feeling that his expertise and institutional knowledge — things he built over years — now count for less than a chatbot's confident but context-free analysis.