U.S. starting to see heavy job losses in roles exposed to AI - The Detroit News

TEXT START: Several U.S. occupations expected to be impacted by artificial intelligence saw heavy job losses for a second year in 2025, led by customer service representatives and certain types of secretaries and salespeople.

THE DISSECTION

This article is a confirmation signal dressed as news. It catalogs documented employment declines across 18 BLS-flagged occupations and treats the numbers as a trend worth watching. The framing is clinical, data-forward, and studiously neutral. It quotes Goldman Sachs economists. It cites BLS caveats. It notes healthcare secretaries as a growth outlier. It gestures toward theChatGPT-era inflection point without naming it as the structural rupture it is.

The prose performs the work of normalization: these are "early signs," a "large-scale rearrangement," an "AI-driven impact." The language of weather. Things that happen. No agency assigned. No system verdict rendered.

THE CORE FALLACY

The article assumes the decline is measurable, sequential, and containable—that you can track AI's labor market footprint by counting secretaries and customer service reps and treating that count as the story.

The DT framework identifies this as a category error of sequence. These losses are not the impact of AI. They are the prodromal symptoms of the structural collapse of the mass employment->wage->consumption circuit. The article treats current job losses as a problem to be tracked. DT treats them as autopsy data from a body still technically alive.

The Goldman Sachs finding—that openings have fallen below pre-pandemic levels for AI-exposed occupations—means the pipeline is already compromised. Hiring has slowed before mass layoffs even arrive. That is not "early signs." That is acceleration.

HIDDEN ASSUMPTIONS

1. The list is exhaustive. The BLS cautions it isn't, but the article proceeds as if 18 occupations captures the scope. It doesn't. The same disruption is rewriting legal research, financial analysis, coding, content creation, logistics coordination, and diagnostics—none of which appear on this list.

2. Healthcare secretaries are a real exception. Medical administrative work is being automated at the backend (AI medical scribes, coding software, scheduling systems) while front-end growth persists. This is a lag artifact, not a moat. The exception proves the rule is dying, not that it survives.

3. 0.2% overall decline among 10 million workers is "heavy." In historical terms, perhaps. Under DT logic, this is the opening act of a sustained demolition. 0.2% followed by 0.2% followed by 0.2% is not a cyclical dip. It is the beginning of a long, accelerating hollowing.

4. "Rearrangement" implies reassignment. The article implies these workers will find other work. There is zero evidence of this. The occupations gaining ground do not absorb displaced administrative, clerical, and sales workers at anywhere near the scale required. The DT framing: these are not rearrangements. They are eliminations.

SOCIAL FUNCTION

This article is transition management theater. It acknowledges the phenomenon, provides enough data to appear serious, and buries the structural implications in neutral-voice caveats. Its function is to let readers process the information without confronting the arithmetic: 10 million jobs in visible decline, healthcare growth masking broader erosion, openings falling before layoffs arrive.

It reads like a pulse monitor attached to a patient in septic shock. The numbers are real. The interpretation is fatally inadequate.

THE VERDICT

Mechanical Death: The data confirms P3 of the DT framework—productive participation collapse is not theoretical. It is measured. Customer service (130,180 lost in one year), clerical/administrative (31,030), wholesale sales (28,670). These are not boutique categories. They are the functional backbone of routine cognitive labor, and they are contracting.

Lag Assessment: The 2.6-year window since ChatGPT's November 2022 release maps precisely to the time lag DT predicts between AI capability maturation and employment impact manifestation. What we are seeing is not the impact. It is the first visible tissue damage from a process already deep in the arterial system.

The Goldman Sachs signal is the most important data in the article: Job openings falling below pre-pandemic levels in AI-exposed occupations. This means the labor market is not just shedding incumbents—it is ceasing to create new positions. The replacement pipeline has been severed before the current workforce is even fully displaced. That is the signature of structural, not cyclical, collapse.

Survival verdict on the workers: Hyena's Gambit territory. Retraining within these occupational categories is cannibalizing each other—everyone is simultaneously trying to skill-up into the remaining slots. The arbitrage window is collapsing as the information spreads.

The article's function: It will be cited by those who want to appear informed about AI's economic effects. It will not be cited by those who understand what the data actually means. That asymmetry is itself diagnostic.