For years, the dominant technological narrative was that AI was an inevitable, top-down replacement for human systems. Leaders assumed that mandating AI use would lead to widespread efficiency. Eager governments poured hundreds of millions of dollars into software to replace legitimate human workflows and automate decisions. That strategy has its limits, not just ethically, but practically. Today, front-line workers are pushing back, revealing that these AI mandates have their limits.

For a generation, South Korea was at the forefront of technological innovation. When the government announced the introduction of AI digital textbooks across elementary and secondary schools, the move was framed as the future of education by some and a dehumanization of teaching by others. Over $385 million was spent in 2024 alone on the project in an attempt to build more intelligent software. Algorithms would track student progress, change lessons, and save teachers time by grading assignments. But it didn’t work.

Soon, more than 50,000 parents signed a petition urging the National Assembly to eliminate the textbooks. Growing reports of screen fatigue, privacy risks, and software glitches were paired with teachers having issues, including glitchy systems that led to more work. Lawmakers passed legislation stripping AI software of its legal status, making it optional instead. Since then, AI textbooks are rarely used, causing huge software projects to be in limbo.

This failure in South Korea has been part of a global pattern of AI meeting human rejection. In the United States, federal and state courts attempted to streamline court case backlogs using AI filing systems. Instead of efficiency, the software led to bias and hallucinations, forcing federal judges to issue more regulations.

Similarly, in the UK, corporate HR departments rushed to use AI to screen thousands of job applicants. The tools showed bias, punishing neurodivergent candidates and non-English speakers. The social backlash was severe, and protests led to major firms scrapping the hiring.

These events show that AI still can’t replace human judgement in many places. But more importantly, it shows that the limits of AI are not just in its hallucinations, but also in our overconfidence in it. When policymakers try to bypass human expertise, social pushback is often preceded by technological failure. Flawed tech policy has thus created a trail of expensive and abandoned software projects, showing that AI is not a silver bullet but a potential tool.