'AI slop' covers workplace as young people place uncritical trust

Nguyen Hai, a project manager at a Hanoi software company that builds enterprise AI transformation solutions, recalls how developers used to manually write every single line of code before the rise of "vibe coding"—the practice of programming via natural language AI prompts.

While manual coding took longer, it guaranteed total control over a project from concept to implementation, he says.

"Now we routinely run into junior staff using AI without checking the work, completely unaware of what their code actually contains.

"The application might run temporarily, but the developers don't understand the system design and can't explain or troubleshoot it".

To Manh Hoang, CEO of Datapot, a firm specializing in data training, solutions, and AI transformation, says shoddy AI output dressed in polished packaging is invading companies in every industry.

He points to an ongoing survey launched in September 2025 by HR tech firm BetterUp and Stanford University, which reveals that 40% of employees have received "AI garbage", work products or reports that appear complete but add zero value.

Flipping the script, half of all respondents admitted to sending unvetted, AI-generated content to their own colleagues.

These show up most frequently in professional services like law, finance, human resources, marketing, and technology

Claude app interface on a mobile device. Photo by Luu Quy

Drawing from Datapot's experience in training and supporting AI adoption at more than 150 Vietnamese enterprises, Hoang recommends two fundamental rules to stem the tide: avoid complete reliance and verify everything.

He says the over-reliance is particularly acute among junior employees, who often lack core domain knowledge and issue AI prompts before developing the expertise needed to evaluate the results.

"Over-reliance degrades core skills, effectively outsourcing critical thinking and reasoning to AI models.

"Early-career professionals need to master their fields independently, separate from AI, so they build genuine capability before leaning on digital tools".

But industry experts also warn against treating all AI output as inherently flawed.

In Stack Overflow’s 2025 Annual Developer Survey, based on over 49,000 responses across 177 countries, 66% of developers cited AI solutions that are 'almost right, but not quite' as their top frustration with AI tools.

Left uncorrected, subtle errors accumulate into costly architectural flaws that require expensive rework down the line, diminishing product quality and enterprise service standards.

From a managerial standpoint, businesses must establish clear AI policies even if they haven't launched full-scale digital transformations.

"At minimum, leadership must outline which tasks permit AI usage and spell out clear protocols for how to use it," Hoang says.

He adds that corporate guardrails lag far behind adoption rates: while 66% of Vietnamese workers currently use AI on the job according to the Vietnam National University, Hanoi's, December 2025 Annual AI Report, only a fraction of organizations have formal rules governing its use.

Pham Dieu Linh, HR director at Synnex FPT, echoes those concerns, emphasizing that organizations must broaden their operational management beyond human teams to encompass the AI tools being deployed alongside them.

Companies must first pinpoint which tasks and workflows are suited for AI integration, and from there, leadership can design clear processes distinguishing human-AI collaboration from human-only tasks, complete with corresponding operational policies and evaluation metrics, she says.

"When an enterprise sets out to boost productivity through AI, it needs to define where to start, what humans remain responsible for, and how success is measured."

"Only then can AI usage be controlled and scaled sustainably across the organization, rather than splintering into fragmented, ad-hoc individual habits".

A December 2025 survey by recruitment platform Top found that over 59% of companies planned to invest in employee AI skill training in 2026.

But only 38% intended to map out AI-integrated workflows, and just 7% planned to implement security and compliance guidelines.

"Most companies are focusing their investments on individual-level software skills rather than building institutional systems," Linh observes.

Hai says his firm had to move quickly to implement new rules: engineers are permitted to leverage AI but must review and validate the output down to every single line of code.

"If you wait until technical debt snowballs out of hand, AI won't be able to handle new tasks, and engineers won't understand the underlying system well enough to intervene."