The Real Threat AI Poses to Employment: Not Job Loss, But Wage Decline

Redefining the AI Employment Threat

Public discourse around artificial intelligence's impact on the job market has long fixated on one anxiety-inducing question: How many jobs will AI replace? However, a new study raises a more alarming—and more realistic—perspective: AI's true threat to employment may not be mass unemployment, but the sustained decline of wage levels.

This conclusion upends our conventional understanding of AI's impact. In previous narratives, automation and AI were cast as "job killers," as if vast numbers of positions would suddenly vanish one day. But reality tends to evolve in more covert and gradual ways: jobs may still exist, but their value, bargaining power, and corresponding compensation are being quietly eroded.

Why AI Brings Wage Cuts Rather Than Unemployment

AI Changes the Value Composition of Jobs

When AI can handle a portion of tasks that previously required human completion, jobs don't disappear immediately—they get "redefined." A position that once demanded senior expertise may, with the assistance of AI tools, lower its dependency threshold on human experience. This means employers can hire workers with "AI + basic skills" at lower salaries to replace formerly high-paid senior experts.

In other words, AI hasn't eliminated the job—it has diluted the job's "skill premium." Skill Premium is a core concept in labor economics, referring to the additional wage returns that workers with specific skills or higher education levels receive relative to low-skilled workers. Over the past few decades, technological progress has generally been considered "Skill-Biased Technical Change" (SBTC)—meaning new technologies tend to boost the productivity and wages of high-skilled workers. However, AI's uniqueness lies in its erosion of the premium for mid-to-high-skill positions—jobs whose core value lies in knowledge processing, analytical judgment, and professional experience, which is precisely what large language models and generative AI excel at assisting or even partially replacing. When the unique human capabilities needed to complete a task are partially substituted by machines, employees' leverage in salary negotiations diminishes accordingly.

Chain Reactions on the Labor Supply Side

If AI enables more people to perform work that previously had high entry barriers, the supply of qualified labor increases correspondingly. Under basic economic logic, when supply increases while demand remains constant (or even decreases due to efficiency gains), the price—that is, wages—naturally faces downward pressure.

This phenomenon echoes the "Task-Based Framework" proposed by economist David Autor. Under this framework, jobs are not indivisible wholes but are composed of a series of specific tasks. AI may only replace some of these tasks, reducing the human requirements for the entire position from "comprehensive expertise" to "basic competency + AI collaboration," thereby altering the equilibrium wage level in the labor market. When the effective labor supply pool expands dramatically due to technological assistance, skills that were once scarce become widely accessible, and market pricing mechanisms naturally adjust downward.

This effect is particularly likely to hit white-collar positions with moderate skill levels and relatively standardized tasks. These positions will neither be fully automated nor able to maintain their original scarcity premium, ultimately falling into the awkward situation of "the job still exists, but income is declining."

Why Stealthy Wage Decline Is More Alarming Than Unemployment

Compared to mass unemployment—an overt, statistically measurable shock—the slow erosion of wages is a threat that's harder to detect and harder for policy to address in a timely manner.

The unemployment rate is a core indicator closely watched by governments and media. When mass layoffs occur, they typically trigger immediate social attention and policy intervention. However, wage stagnation or gradual decline—especially the decrease in real purchasing power against an inflationary backdrop—often gets buried in macroeconomic data, making it difficult to generate clear policy signals.

Notably, wage stagnation is not unique to the AI era. Data from the U.S. Bureau of Labor Statistics shows that since the 1970s, growth in median real wages in the United States has significantly slowed, falling far behind productivity growth. This "Productivity-Pay Gap" phenomenon has been attributed to multiple factors: globalization, de-unionization, winner-take-all market structures, and more. The introduction of AI may accelerate this trend, as it simultaneously exerts pressure from both the supply side (increasing effective labor) and the demand side (reducing corporate dependence on high-paid talent). The key difference is that previous technological changes primarily impacted blue-collar manufacturing jobs, while AI is extending similar pressure to white-collar knowledge workers.

For ordinary workers, the destructive power of this threat is equally significant. Job retention may create a "false sense of security," leading people to mistakenly believe they haven't been affected by AI, while their economic standing is being progressively undermined. In the long run, wage stagnation affects consumer spending power, weakens the middle-class foundation, and has far-reaching effects on the overall economy.

Personal and Policy Strategies for the AI Era

Individual Response: From Being Replaced to Leveraging AI Tools

For individual workers, this research sends a clear signal: merely holding onto a job is not enough—the key is how to maintain or even enhance your bargaining power in the AI era.

This means proactively moving up the value chain—developing capabilities that AI struggles to replace, such as complex decision-making, creative thinking, cross-domain integration, and interpersonal collaboration. From a cognitive science perspective, these capabilities remain difficult for AI to replicate due to the fundamental technical limitations of current artificial intelligence. Existing AI systems (including large language models) are essentially pattern recognition and statistical prediction systems, adept at handling tasks with clear patterns but still fundamentally limited in areas requiring common-sense reasoning, contextual awareness, ethical judgment, and genuinely innovative thinking. What psychologists call "Tacit Knowledge"—judgment capabilities that are difficult to explicitly articulate and rely on accumulated experience—remains a unique human advantage. Furthermore, work involving trust-building, emotional resonance, and complex social interactions derives its value from human relationships themselves, rather than pure information-processing efficiency.

At the same time, treating AI as a tool to amplify your own productivity—rather than passively waiting to be replaced by tools—is an important path to maintaining wage competitiveness. Those who can skillfully use AI tools to dramatically improve output quality and efficiency may actually gain greater bargaining power in the new environment.

Policy Level: Focus on Wage Indicators, Not Just Unemployment Rates

From a policymaking perspective, this research reminds decision-makers that unemployment rates alone cannot serve as the sole yardstick for measuring AI's impact. Wage levels, real purchasing power, and the labor share of national income may better reflect AI's deeper effects on the labor market.

Labor Share of Income is a key macroeconomic indicator for measuring the distribution of economic outcomes. Research from both the International Monetary Fund (IMF) and the Organisation for Economic Co-operation and Development (OECD) shows that over the past forty years, the labor income share in most advanced economies has been declining persistently, with the capital income share rising correspondingly. If AI further suppresses wage levels, this trend could intensify, leading to widening income and wealth inequality. From a macroeconomic perspective, workers are the primary drivers of consumption, and wage stagnation suppresses aggregate demand, potentially creating a vicious cycle of slowing economic growth. This is why some economists have begun discussing the necessity of redistribution mechanisms such as "Universal Basic Income" (UBI) and "robot taxes."

How to ensure workers share in the dividends of technological progress while productivity improves—avoiding excessive concentration of gains on the capital side—will become a critical public policy issue going forward.

Conclusion

AI's impact on employment is evolving from the binary question of "will we lose our jobs" into the complex proposition of "how is the value of work being redistributed." The value of this research lies in redirecting our attention from dramatized unemployment panic toward the more real and pervasive phenomenon of wage erosion.

The real challenge may not be machines stealing our jobs, but rather imperceptibly lowering the price of our labor. Understanding this point is what will enable individuals, businesses, and society to face the AI era with greater clarity.