Amsterdam began not as a nation’s city but as a corporation’s outpost. The Dutch West India Company established New Amsterdam at the tip of Manhattan. Commerce prevailed over creed. At the colony’s northern edge stood a twelve-foot wooden wall, erected to secure property claims and defend against Lenape resistance. That wall would later lend its name to Wall Street. From its inception, the settlement embodied a paradox: pluralism sustained by trade, expansion secured through enclosure.
Capitalism functioned as a civil religion. It dissolved inherited hierarchies and accelerated innovation, yet it depended upon continuous growth. In a finite world, growth without limit becomes strain. Four centuries later, that strain has arrived.
On March 18, 2026, Federal Reserve Chair Jerome Powell stated plainly: “Effectively, there is zero net job creation in the private sector.”[1] The economy shed 92,000 jobs in February. Unemployment rose to 4.4 percent.[2] Total job creation for 2025 was the weakest outside a recession since 2003. Powell noted that labor force growth has stopped entirely. It is a condition, he said, “we’ve never had in our history.”
Artificial intelligence arrives into this stalled economy with unusual force. It is not just another commodity, but a technology that automates cognition itself. If industrialization mechanized muscle, AI mechanizes decision. Entry-level roles in writing, coding, data entry, and customer service are vanishing, being replaced by software that needs no sleep, no health insurance, no parental leave, and no pension.[3] AI arrives not into abundance but into scarcity, accelerating a zero-sum displacement of workers already navigating a collapsed labor market.
Social work emerged during the upheavals of industrial capitalism. Settlement houses and early welfare institutions responded to structural dislocation: migration, labor exploitation, and urban poverty.[4] The profession’s ethical mandate has never been confined to individual intervention. It extends to the systems that generate vulnerability. AI is now such a system.
The United States has over one million trained social workers, the largest such concentration on earth.[5] They serve fifty million people with mental health disorders annually, twenty million with substance use disorder, 580,000 experiencing homelessness, and 400,000 children in foster care.[6] They work in treatment centers, clinics, hospitals, schools, courts, and prisons. This workforce has no unified voice on the defining structural transformation of its era.
The NASW Code of Ethics obligates social workers to challenge social injustice and to act on behalf of people who are vulnerable and oppressed.[7] That obligation is not passive. It does not permit waiting until the damage is undeniable. AI-driven decision-making is already operating in child welfare screening, criminal sentencing, benefits eligibility, and mental health triage.[8] These are domains where social workers hold direct responsibility for human outcomes. When those systems are opaque, unauditable, and designed without clinical input, the profession’s ethical mandate is being violated.
What is required is not caution but serious structural opposition. Three actions are immediately within reach:
- The Council on Social Work Education should mandate AI ethics and algorithmic literacy in MSW curricula. A profession that cannot evaluate the tools reshaping its field cannot advocate for the people affected by them.
- NASW should establish a standing task force on AI, charged with developing enforceable practice standards for its use in clinical and institutional settings.
- Social workers in child welfare, criminal justice, healthcare, and housing should have standing to challenge algorithmic decisions affecting their clients. This is the same standing they exercise against any other harmful institutional process.
Participation in AI governance must become social justice advocacy. One million social workers, organized around a shared standard, would constitute the largest ethical counterweight to unregulated AI deployment in the Western world.
Amsterdam once stood at the frontier of commercial modernity. Artificial intelligence stands at the frontier of cognitive modernity. History tells us what happens when that frontier is governed only by profit: the wall goes up, and the people outside it are forgotten.
Notes
[1] Jerome H. Powell, FOMC Press Conference Transcript, Federal Reserve Board, March 18, 2026.
[2] U.S. Bureau of Labor Statistics, “The Employment Situation, February 2026.” Unemployment 4.4%; net job losses of 92,000.
[3] World Economic Forum, Future of Jobs Report 2025 (Geneva: WEF, January 2025): 40% of employers expect to reduce staff where AI can automate tasks; 92 million jobs projected displaced by 2030. See also SignalFire State of Talent Report (2024), documenting a roughly 25% decline in Big Tech new-graduate hiring; Cornell University research finding a ~13% reduction in junior hiring at U.S. firms adopting AI.
[4] Jane Addams, “The Subjective Necessity for Social Settlements” (1892), repr. in Twenty Years at Hull-House (New York: Macmillan, 1910); see also Mina Carson, Settlement Folk: Social Thought and the American Settlement Movement, 1885–1930 (Chicago: University of Chicago Press, 1990).
[5] NIH/American Community Survey (2024), est. 1,022,859 social workers in the United States.
[6] SAMHSA, 2023 National Survey on Drug Use and Health; U.S. Department of Housing and Urban Development, 2024 Annual Homeless Assessment Report to Congress; Administration for Children and Families, AFCARS Report #30, FY 2023.
[7] National Association of Social Workers, Code of Ethics (Washington, DC: NASW, 2021 rev.), Ethical Principles: “Social workers challenge social injustice” and pursue social change “with and on behalf of vulnerable and oppressed individuals and groups of people.”
[8] On child welfare screening, see Chouldechova et al., “A Case Study of Algorithm-Assisted Decision Making in Child Maltreatment Hotline Screening” (FAT* 2018), and ACLU, “Family Surveillance by Algorithm” (2021), documenting predictive risk models in at least 26 states, including the Allegheny Family Screening Tool. On criminal sentencing, see State v. Loomis, 881 N.W.2d 749 (Wis. 2016), on the COMPAS recidivism algorithm. On benefits eligibility, see V. Eubanks, Automating Inequality (New York: St. Martin’s, 2018). On mental health triage, see M. Sendak et al., “Presenting Machine Learning Model Information to Clinical End Users,” npj Digital Medicine (2020).