No, AI is not killing jobs for everyone, studies say - Computerworld

TEXT START: While companies continue to seek people with advanced AI skills, younger workers remain in danger of seeing their jobs automated.

The Dissection

The article performs statistical containment. It takes a real early warning—young workers in AI-exposed occupations are down 19% relative to comparable workers in less-exposed jobs, with employment falling 11% since late 2022—and packages it as proof that the system is broadly intact because older workers and aggregate payrolls have not yet collapsed.

Its central maneuver is temporal: it treats the first visible fracture in the entry-level pipeline as an isolated demographic effect rather than the leading edge of a broader substitution process. It then points to demand for AI skills, human oversight, and aggregate job creation as if new coordination roles prove durable human economic necessity. They do not.

The Core Fallacy

The article confuses “AI is not eliminating every job immediately” with “AI is not destroying the mass employment system.” Those are different propositions.

The data presented are compatible with the Discontinuity Thesis. Young workers occupy the most standardized, cheapest, and least institutionally protected cognitive roles. They are therefore the first cohort exposed to automation. Older workers retain positions because of experience, institutional embeddedness, managerial authority, client relationships, and the lag between technical capability and organizational redesign. That is a delay mechanism, not a refutation.

The article also mistakes gross demand for AI labor for net demand for human labor. Companies may hire AI specialists while using AI to reduce thousands of ordinary cognitive roles. A growing market for people who deploy the replacement system does not preserve the people being replaced. It is the labor market installing its own demolition equipment.

Finally, “humans checking AI output” is treated as a permanent complement. Under P1, verification itself becomes a target for automation once systems become reliable enough, or once the cost of human review exceeds the cost of residual error. The night shift described in the article is not evidence of human indispensability. It is evidence that organizations are already moving production outside the human workday.

Hidden Assumptions

• Aggregate payroll stability means productive participation is stable. It does not account for cohort exclusion, wage suppression, underemployment, or declining access to the career ladder.
• Older workers are safe because their jobs are structurally durable. Their current protection may reflect lag, not immunity.
• AI-exposed entry-level jobs can disappear without damaging the wider system. In reality, removing entry points destroys the training pipeline that replenishes experienced workers.
• New AI-related roles will scale sufficiently to absorb displaced workers. The article provides no evidence for that.
• Human oversight remains necessary indefinitely. It assumes the verifier cannot itself be automated.
• Job counts measure economic health. They do not establish bargaining power, income distribution, job quality, or whether humans remain necessary to production.
• The current transition rate is the terminal rate. A slow initial rollout is treated as a permanent ceiling on substitution.
• Hiring plans for 2027 represent durable employment rather than temporary transition spending, competitive experimentation, or labor substitution infrastructure.
• Institutions can preserve human-only domains at scale. The article never addresses P2: whether firms facing competitive pressure can voluntarily refuse cheaper, faster machine labor.

Social Function

Primarily ideological anesthetic and transition management, with a partial truth at its core.

The partial truth is narrow: AI has not yet produced visible economy-wide payroll collapse, and some experienced workers remain employed while demand for AI implementation skills rises. But the article turns that lag into reassurance. It gives executives a vocabulary for calling early displacement “selective,” presents the endangered young cohort as a contained casualty, and treats the construction of AI deployment capacity as evidence that the old labor order is adapting.

Its most important omission is the productive-participation question. It counts jobs while avoiding the question of how many humans remain economically necessary once cognitive work is performed by agents. That omission converts a structural warning into a business-cycle story.

The Verdict

The article does not disprove AI-driven labor obsolescence. It documents its opening phase: entry-level cognitive workers are being removed first, wages are being compressed, and firms are hiring a smaller class of people to operate the machinery that replaces broader classes of workers.

The headline is therefore false in the only sense that matters. AI is not killing jobs for everyone yet. It is killing the access route by which “everyone” was supposed to become economically useful. Aggregate payroll data are the anesthesia; the missing entry-level pipeline is the wound.