Top News

Last month, after spending three hours tweaking a draft to satisfy nearly 20 rounds of automated feedback, Hanoi communications specialist Minh Phuong, 26, stared at another cold notification: “Article unsatisfactory. Further edits required.”

A year earlier, she had been thrilled by AI’s potential to streamline editing and spark new ideas. She had spent weeks creating custom prompts to train the software in her personal writing style, running every document through the tool before submitting it to management.

But what began as a productivity boost soon degenerated into a grueling ritual of algorithm-driven demands.

“I spent three hours fixing a piece based on AI suggestions, longer than it would have taken me to write from scratch,” she says about the “article unsatisfactory” episode. “It made me question my own competence.”

In Ho Chi Minh City, 29-year-old event manager Hong Anh is confronted by the same problem.

One time AI generated her project slide deck in 10 minutes, but it took her four hours to smooth out the disjointed phrasing.

Her supervisor then returned the draft along with an eight-page report generated by an AI evaluation tool that completely contradicted the logic established by her tool.

“I wanted feedback grounded in my boss’s real-world experience, but all I got were machine-generated metrics,” she says. “Both of us got sucked into a revision loop just to appease invisible algorithms.”

Desperate for a solution, she added several other AI tools into her work, only to worsen the noise. Instead of leaving the office on time, she found herself working two to three hours overtime just to train her digital tools. Managements are equally caught in the trap.

Dang Tien, 35, a department head in Hanoi, routinely uses AI to review staff drafts to cut down on his workload. He paid subscriptions to several premium platforms to cross-examine submissions, hoping to achieve flawless results.

“Every tool produces different feedback,” he says. “In the end, I spent twice as much time reconciling the original draft with conflicting computer recommendations.”

This cycle of employees relying on AI to execute tasks and managers also using AI to grade them is creating a passive, gridlocked work environment.

A study by the Upwork Research Institute found that 77 percent of employees feel AI has actually increased their workload, while 47 percent have no idea how to meet expected productivity targets using the technology.

Researchers at Harvard Business Review have dubbed this “AI brain fry”, a state of severe mental fatigue triggered by over-reliance on artificial intelligence.

Symptoms include chronic distraction, elevated stress, and an unsettling feeling of perpetual incompleteness.

A joint study published in early 2026 by Boston Consulting Group and Harvard Business Review found 14 percent of the US workforce suffering from AI brain fry.

In fast-paced, information-heavy fields like marketing and communications, that number jumps to 26 percent.

Chu Tuan Anh, director of the Aptech International Software Developer Training System (Ho Chi Minh city), says unchecked tech dependence is introducing unprecedented workplace strain. “AI was designed to liberate human labor, but mismanaged, it becomes a burden that drains human intellect.”

Experts blame misaligned expectations and inadequate technical skills for this. Managers overestimate software capabilities and assign unrealistic workloads, while staff lack the engineering and fact-checking skills needed to catch erroneous outputs.

Le Dinh Duy, deputy director of the GenAI Product Center at tech firm FPT Smart Cloud, says the core problem is humans using AI to evade critical thinking. “Employees outsource thinking to AI, while bosses outsource evaluation to another AI. When nobody takes ownership of the content, work grinds to a halt.”

He emphasizes that generative models produce text based on statistical language patterns, not subject-matter expertise.

Chasing machine-defined perfection means surrendering decision-making authority to a system that takes zero professional responsibility, he points out.

To break the cycle, he advises workers to establish clear evaluation criteria before generating AI content, independently verify dates and figures, and cap AI-guided revisions at two iterations.

Workers must be able to justify the reasoning behind their final submissions rather than rely blindly on automated suggestions.

Anh recommends that managers adopt a “1.5 time multiplier” rule, meaning if they complete a task using AI in one hour, employees should be given 1.5 hours so that they have adequate breathing room to verify data and refine outputs.

For individual workers, he proposes a three-step framework: think independently for five minutes before writing a prompt, treat AI as a reference tutor rather than an absolute authority, and ensure you can explain the logic of any AI-generated solution before adopting it.

After months of stress, Phuong revamped her approach. She now gathers research and outlines projects manually before using AI purely to polish sentence structures, strictly blocking the software from generating unverified data.

In Da Nang, 30-year-old marketing specialist Hoang Nam took a similar stand.

When his director used AI to critique a campaign strategy, Nam relied on ground-level survey data to defend his strategy and highlight flaws in the software’s counterarguments. “AI only boosts productivity when you control the input and keep your independent judgment intact,” Nam says.