Societies that prepare institutions and education will win the AI revolution

Ryu Keun-kwan

The author is an emeritus professor of economics at Seoul National University and a former commissioner of Statistics Korea.

The AI revolution is no longer a future possibility. It has already begun. Yet, as history repeatedly demonstrates, technological progress does not automatically produce social progress. Technology merely expands what is possible. Whether those possibilities ultimately generate growth and prosperity or instead deepen inequality depends on how well a society prepares for change.

One useful way to understand the AI era is through two J-curves. The first is the "time-axis J-curve." Whenever a transformative technology emerges, confusion and adjustment costs often appear before productivity gains become visible. Existing industries, institutions and regulations require time to adapt, creating a temporary period of stagnation. As the technology gradually spreads throughout the economy, however, productivity rises and economic growth accelerates.

The first challenge of the AI era, therefore, is institutional preparedness for this time-axis J-curve. To accelerate technological diffusion and fully capture its benefits, governments must modernize education, research, industrial policy and data governance before disruption arrives. The longer these reforms are delayed, the longer the period of uncertainty and slow growth will persist before the economy reaches the upward side of the curve. This is fundamentally a question of institutional readiness rather than technological capability.

An even more significant transformation, however, will emerge through a second pattern: the "skill-axis J-curve." Labor demand in the AI era is likely to take a J-shaped form based on workers' skill levels. Highly skilled workers will increasingly collaborate with AI, significantly improving both their productivity and market demand. Workers with medium or moderately low skill levels, by contrast, are more likely to be displaced through automation, causing demand for their labor to fall sharply into the bottom of the J-curve. Demand for lower-skilled workers may also stagnate or decline. Occupations requiring physical labor, including cleaning, caregiving, logistics and construction, may maintain limited demand only until physical AI technologies become widely deployed.

The more serious concern is that the bottom portion of the skill-axis J-curve may become unusually broad. As automation affects a large share of middle-skill occupations, the center of the labor market itself could weaken. Since skill levels are closely linked to income, this transformation presents a substantial threat to middle- and lower-income households.

The central policy challenge of the AI era is therefore society's response to the skill-axis J-curve. As labor market structures undergo profound changes, governments must first establish effective social protection.

Several policy options deserve consideration, including the negative income tax, the earned income tax credit and universal basic income. Each distributes government support differently across income groups. Under a negative income tax, government assistance gradually declines as earned income rises. Under an earned income tax credit, subsidies increase as employment income grows, thereby strengthening incentives to work. Universal basic income occupies a middle position because everyone receives the same payment regardless of income. The optimal design of these systems naturally depends on each country's fiscal capacity and labor market conditions. Any sustainable safety net must simultaneously protect lower-income households while preserving incentives for labor force participation.

The skill-axis J-curve also points toward a new direction for education policy. If rewards increasingly concentrate among highly skilled workers, educational goals cannot simply focus on raising average achievement. The priority should instead be cultivating excellence.

Not everyone can stand at the top under a single uniform standard. Education should therefore move away from standardized expectations and instead encourage individual choice based on personal abilities and interests. Rather than directing every student along the same path, schools should help individuals achieve advanced expertise within their chosen fields. Personalized education will become far more valuable than one-size-fits-all instruction.

Ultimately, AI policy requires two distinct but complementary responses. The time-axis J-curve calls for institutional reforms that enable societies to adopt new technologies quickly and effectively. The skill-axis J-curve requires stronger social protection alongside fundamental innovation in education so workers can adapt to changing labor markets.

Technological innovation is advancing faster than many anticipated. The decisive question is no longer the pace of technological development itself but whether society is prepared to meet it. Success in the AI era will not be determined solely by superiority in algorithms or computing power. The real divide will emerge between societies that prepare their institutions, education systems and labor markets for technological transformation and those that fail to do so. That difference, more than technology alone, will determine the winners and losers of the AI revolution.