AI Is Starting to Take a Toll on Jobs. Goldman Sachs Finds Entry-Level Workers Are Most Exposed.

TEXT START: New research found that industries with greater exposure to AI automation have generally experienced weaker growth in job openings since the second half of 2022.

THE DISSECTION

The article records an early fracture and disguises it as a manageable fluctuation. Its own evidence shows AI-exposed sectors falling below historical employment trends, call-center employment running 27% to 39% below trend, and entry-level workers absorbing the strongest damage. It then uses modest aggregate effects to imply that the labor market remains fundamentally intact.

That is the standard statistical camouflage of a system entering transition. The corpse does not need to be cold for the circulation to have failed. AI adoption is only estimated at 15% to 20% across major developed economies, yet the damage is already concentrated in the occupations and industries where automation is easiest. The article is measuring the first exposed layer, not the terminal outcome.

THE CORE FALLACY

The central error is treating a small current headcount effect as evidence of a small structural threat. A 0.2- to 0.6-percentage-point drag on entry-level employment growth sounds minor until the mechanism is understood: junior work is the training pipeline into experienced work. Remove the pipeline and the labor market loses future workers before it registers mass layoffs.

The article also conflates “not yet a broad collapse” with “not a collapse in progress.” Job openings, employment trends, and annual headcount growth are lagging indicators. They do not capture hiring freezes, degraded job quality, task compression, invisible underemployment, or firms using AI to avoid creating the next position at all.

Under the Discontinuity Thesis, this is P1 expressing itself through P3. AI does not need to eliminate every job immediately. It only needs to become cheaper and more capable than humans across enough cognitive tasks to make human labor progressively unnecessary. The measured entry-level shock is the leading edge of that process.

HIDDEN ASSUMPTIONS

• AI adoption will remain gradual and stable rather than accelerating after firms validate cost savings.
• Eliminated junior tasks will be replaced by enough new human tasks to preserve the employment pipeline.
• Companies will continue hiring beginners in order to train them, even when AI can perform much of the training-stage work.
• Historical employment trends remain a valid baseline after the production function changes.
• Aggregate employment can conceal concentrated damage without producing systemic consequences.
• Workers displaced from exposed sectors can smoothly migrate into durable, economically necessary roles.
• Productivity gains will automatically create compensating demand and employment.
• Human institutions can coordinate stable human-only economic domains at scale.

None of these assumptions is established by the supplied evidence. Several directly contradict the thesis’s coordination and competition mechanics.

SOCIAL FUNCTION

Classification: partial truth wrapped in ideological anesthetic and transition management.

The article is not fabricated. Its data identifies real displacement pressure, especially against entrants. But its framing domesticates the threat: “modest” overall effects, “concentrated” sectors, and the suggestion that AI may merely supplement junior employees. This converts a structural break into a conventional labor-market adjustment story, giving institutions permission to wait while the employment ladder is dismantled beneath them.

The article’s most important fact is also the one it underweights: entry-level workers are being hit first. That is not a side effect. It is strategically decisive. A system that removes the first rung eventually removes the ladder.

THE VERDICT

This is an early-warning document misread as reassurance. It provides credible evidence that AI exposure is already weakening the human employment circuit, with entry-level work functioning as the sacrificial zone. It does not prove that post-WWII capitalism has already died, but it is fully consistent with the Discontinuity Thesis: automation begins at the margins, destroys the labor pipeline, and expands as competitive pressure forces adoption.

The article measures the beginning of productive participation collapse while insisting the patient is stable. That is not analysis. It is a lagging indicator with a sedative attached.