"AI Does It Better Than You": The Hidden Crisis Behind Slashing Entry-Level Jobs - SBS뉴스

TEXT START: According to an analysis by the Bank of Korea, 94% of the 285,000 youth jobs lost between June 2022 and June 2026 were concentrated in industries with high exposure to AI.

The Dissection

The article documents the early amputation of the career ladder, then tries to reclassify it as a design problem. Its strongest evidence is not that AI has already destroyed every youth job, but that entry doors are narrowing while exits are widening: youth employment contracted sharply in AI-exposed industries, inflows fell by roughly 11%, and exits into unemployment rose by roughly 32%.

The article also exposes the central contradiction. AI raises novice productivity by 34%, yet that same productivity gain allows firms to produce the same output with fewer novices. The text sees the contradiction, names it, and then retreats into a managerial appeal for “augmentation,” redesigned roles, and policy incentives.

The Core Fallacy

The article treats automation versus augmentation as a voluntary corporate choice. Under the Discontinuity Thesis, it is a competitive constraint. If one firm can replace ten entry-level workers with three AI-assisted workers, every rival faces pressure to do the same. Voluntary preservation of redundant labor is not a stable equilibrium.

The article’s deeper error is assuming that the career ladder can be rebuilt merely by assigning juniors verification, contextualization, and responsibility. But if AI can perform the drafting, research, coding, and routine analysis, those verification roles become smaller targets rather than permanent shelters. “Human oversight” is often the last wrapper around an automated process, not a new mass employment base.

It correctly identifies P1—cognitive automation—but refuses to follow it through P2 and P3. Institutions cannot coordinate a durable human-only labor domain when firms compete globally on AI-enabled cost and speed. Once productive participation is no longer necessary for the majority, preserving training inefficiency becomes a subsidy for obsolete labor, not an economic law.

Hidden Assumptions

• Firms will retain surplus entry-level workers because society needs them to learn, despite direct competitive incentives to reduce headcount.
• Future employers will need enough human-trained professionals to justify maintaining large junior cohorts.
• Verification, judgment, contextual adaptation, and responsibility will remain labor-intensive rather than becoming progressively automated.
• Policy subsidies can create durable career ladders instead of temporarily paying firms to preserve economically redundant positions.
• Productivity gains will expand total labor demand rather than permitting the same output with fewer workers.
• The older workers gaining employment in AI-exposed sectors possess a permanent moat through tacit knowledge, rather than a temporary lead while AI systems absorb more organizational context.
• Historical technological transitions are a reliable guide, despite generative AI attacking the cognitive training layer itself.
• Individual AI advantage can aggregate into collective employment security. The article’s own figures contradict this: personal productivity rises while aggregate entry opportunities shrink.

The text is cautious about causality, correctly noting that post-pandemic over-hiring and employer preferences may explain part of the decline. But uncertainty over the exact share caused by AI does not neutralize the mechanism. The disappearance of stepping-stone work is precisely the structural signal that matters.

Social Function

Primary classification: transition management, with strong elements of ideological anesthetic and partial truth.

The partial truth is valuable: entry-level work was not merely cheap labor; it was the training infrastructure that converted inexperienced people into experienced workers. The article accurately identifies that AI can destroy this infrastructure before society has built a replacement.

The anesthetic arrives at the conclusion. Responsibility is shifted to “businesses and society,” as though competition can be persuaded to preserve a labor-intensive apprenticeship system. The article converts a conflict over ownership, necessity, and distribution into a question of better workplace design. It lets firms automate the ladder while asking them politely to keep manufacturing climbers.

The Verdict

This is a lucid autopsy sabotaged by a ceremonial resurrection. The article proves that AI is narrowing the first rung of the labor market and may sever the pipeline that produces experienced workers. It then mistakes the need for a social replacement mechanism for evidence that firms can economically recreate the old ladder.

The career ladder is not being accidentally damaged. It is being rendered noncompetitive. Augmentation may create high-value niches and improve the prospects of selected AI-capable juniors, but it cannot preserve mass productive participation once AI delivers superior cognitive output at lower marginal cost. The article sees the guillotine descending and recommends repainting the staircase.