When the company asks you to teach AI the most valuable "judgment," are you willing?

The company is requiring employees to use AI models to "extract" this implicit knowledge, but employees clearly know they are training systems that could replace them; Anthropic's chief economist points out that this is not a technical issue, but an organizational one.
(Background: The more you work diligently, the faster AI eats your job? "Colleague.skill" reveals the brutal truth of knowledge distillation)
(Additional context: For every 4 white-collar workers, 1 middle-aged worker hits a career bottleneck: AI accelerates rewriting career rules)

Table of Contents

Toggle

• First acknowledge the "natural human resistance"
• Enhancement or replacement: companies say the former, economists say the latter
• The market is already splitting: AI skill positions have grown nearly 8 times

Corporate servers store financial reports, contracts, customer data. But the most valuable part: "What does this customer fear hearing most," "Why does this process need to bypass the third step," "Why was that decision wrong back then," no one writes into any system. It resides in the minds of veteran employees, passed down through word of mouth and trial-and-error.

The business world calls this "tacit knowledge." The biggest data collection task in the AI era is not crawling more web pages, but extracting this knowledge from the human brain. But the problem is, employees are very aware of what they are doing.

First acknowledge the "natural human resistance"

Thomas Andersen, VP of AI and Machine Learning at chip design company Synopsys, bluntly stated at a Bloomberg forum last week: "A large amount of genuine expert knowledge, of course, resides in someone's brain, and I have to extract it first."

Extraction, download, mining—these terms used by corporate leaders clearly reveal the core issue: they view employees' brains as a database, and AI as a reading tool.

This logic is not new. Throughout history, each industrialization has involved similar processes: Taylorism decomposed workers' motions into standard procedures, filming them for newcomers to imitate. Ford's assembly line broke down craftsmanship into machines. The difference this time is, what is being broken down is not craftsmanship, but judgment.

Andersen admits that employees do have "natural human resistance" to this. Even those pushing this system have to acknowledge that employees' concerns are "natural."

Enhancement or replacement: companies say the former, economists say the latter

The standard corporate rhetoric is "augmentation," not "replacement." Walmart's AI rollout documents explicitly frame it as "investment in employees." Amazon founder Jeff Bezos and Microsoft CEO Satya Nadella have publicly stated that AI will "empower" workers.

But economists observe that: historically, software developers and the companies that hire them tend to choose "automation" over "augmentation," because automation reduces labor costs.

Matthew Call, management professor at Texas A&M University Mays Business School, straightforwardly suggests: employees should use personal AI tools rather than company models; issues related to how their knowledge is collected and used should be subject to collective bargaining.

U.S. Senator Bernie Sanders takes an even more radical stance: establishing a sovereign wealth fund to share AI's profits, which he describes as "built on human collective knowledge," benefiting the public rather than just flowing to tech company shareholders.

The market is already splitting: AI skill positions have grown nearly 8 times

PwC analyzed data from 27 countries, covering over 1 billion job openings, and concluded that: by 2025, positions requiring specific AI skills will grow at a rate nearly 8 times faster than the overall job market. Salary growth is also higher.

But a closer look at the structure reveals key differences. The fastest-growing, highest-paying roles are not "positions that operate AI," but "positions that amplify human judgment with AI." Pure automation roles—where AI directly replaces human output—are growing more slowly and offer lower salary advantages.

PwC's data also shows that companies using AI to enhance human expertise achieve higher productivity and growth returns than those solely using AI to cut labor costs.

This conclusion has practical significance for employees: what’s most valuable in your brain is not "knowing how to do," but "knowing why you do it." The former can be extracted, trained, and standardized. The latter, at least so far, has not been truly learned by any AI system.