AI and labor in New York - The Chief

TEXT START: A few of our stories and columns are now in front of the paywall.

The Dissection

This is not really an article about whether AI will transform New York labor. It is an inventory of defensive reactions as labor discovers it no longer controls the production function. Unions are trying to convert technological replacement into bargaining language—transparency, accountability, consultation, safety, human connection, contract enforcement, public polling, and regulation. The common move is to recast AI from an ownership weapon into something that can be domesticated by procedure.

The article’s most revealing sequence is layoff followed by AI deployment, especially in the Montefiore and Business Insider examples. It shows that firms do not need AI to be perfect. They need it to be cheaper, scalable, and good enough to pressure labor. Employers decide adoption; workers litigate afterward or organize preemptively under shrinking leverage.

The Core Fallacy

The core error is treating AI as a governance problem rather than a competitive displacement mechanism. If a firm can produce acceptable output with fewer workers, a union contract or ethical rule may slow adoption, raise its cost, or preserve selected roles. It does not restore the mass employment-to-wage-to-consumption circuit. It shifts the timing and distribution of the losses.

Human connection, professional expertise, public trust, and safety can justify retaining human labor in bounded areas, especially nursing, teaching, and transit. They do not prove those jobs remain economically indispensable at current staffing levels. A system can preserve a human interface while automating the underlying cognitive and administrative work. The human becomes a thin legitimacy layer over an automated core.

The proposed worker “control” is also mostly control over deployment conditions, not control over AI capital. A partnership with Microsoft, OpenAI, and Anthropic may give workers training infrastructure; it does not make them Sovereigns. Ownership, compute, models, data, and deployment decisions remain elsewhere.

Hidden Assumptions

• Regulation can freeze human labor requirements despite firms competing on cost and speed.
• Bargaining power remains intact while the employer’s credible substitute improves.
• Transparency and bias correction address the central problem, when the central problem is labor redundancy.
• Jobs protected in one institution will remain viable across the wider market.
• Public opposition to automation can override fiscal pressure, hospital finances, transit economics, or media revenue.
• Frontline work is insulated because it is “human,” ignoring the administrative, diagnostic, scheduling, documentation, and decision-support layers that can be automated first.
• AI-assisted work will remain supplementary rather than becoming a headcount-reduction mechanism once management learns the workflow.
• The state can tax or regulate the gains before capital routes around the jurisdiction.
• Displaced workers will move into new jobs at sufficient scale and with a usable wage path.

The article partially admits the last assumption is failing: it describes mid-level compression and entry-level exclusion. It does not follow that admission to its systemic conclusion.

Social Function

Dominant classification: transition management, with a substantial partial-truth component and an ideological-anesthetic layer.

The text records real harms—biased systems, layoffs, contract violations, degraded trust, and safety concerns. That is the partial truth. But its structure turns a systemic rupture into a sequence of negotiable local disputes. Readers are offered hearings, partnerships, taxes, safeguards, contracts, and “wise use” as the available control surfaces. Those tools may govern the corpse’s posture; they do not revive the labor circuit.

The article also provides institutional self-exoneration. Officials can say they are preparing. Executives can say deployment is responsible. Unions can say they are taking control. Each position is respectable and legible. None answers who owns the productive system after human labor ceases to be necessary.

The Verdict

This is a competent field report on the first defensive formations around an advancing machine economy, not a theory of containment. Its evidence supports the Discontinuity Thesis more than its procedural framing: cognitive tasks are being separated from the workers who historically performed them, entry points into the labor hierarchy are being cut off, and every response remains downstream of owners who control the technology.

The likely outcome is not instant replacement of every human. It is more durable: AI makes human labor selectively optional, then institutions preserve a few visible human roles while compressing headcount, wages, training ladders, and bargaining power. Unions may win delays, severance, staffing floors, liability rules, and protected enclaves. Those are lag defenses and distributional fights. Unless workers acquire control of AI capital—or become indispensable Servitors to those who own it—the postwar labor order is not being saved. It is being administered through its decline.