Artificial Intelligence and the Labor Market
LABOR, TECHNOLOGY, AND SOCIAL TRANSFORMATION
From Job Displacement to a Conscious Reorganization of
Work
Reframed in accordance with the central thesis of A New Social Order
in the Age of Consciousness
| Central thesis AI does not determine the future of work by itself. Ownership, power, and conscious social participation determine whether productivity becomes concentrated profit and insecurity or shared time, security, and freedom. |
|---|
Introduction:
the wrong question and the necessary question
Artificial intelligence is already changing the labor market, but the
change cannot be understood simply by counting jobs that disappear and
jobs that appear. The common question – “Will AI replace workers?” – is
incomplete. Technology does not enter society in a vacuum. Its
consequences are shaped by ownership, political power, workplace
organization, public institutions, and the purposes for which it is
deployed.
The more important question is therefore: who controls AI, who
benefits from its productivity, and who bears the cost of transition?
Under private ownership, the same technology that could shorten the
working day, remove exhausting tasks, and expand access to knowledge may
instead be used to reduce payrolls, intensify surveillance, weaken
bargaining power, and concentrate wealth. Under democratic and socially
accountable control, however, higher productivity could become the
material basis for more free time, universal security, meaningful
participation, and richer forms of human activity.
This distinction is central to the argument of A New Social Order in
the Age of Consciousness. AI is not an independent historical subject
and will not create a new social order by itself. It is a rapidly
developing productive force. Whether it produces insecurity or
liberation depends on the level of collective consciousness,
organization, and participation through which society governs it.
What AI is actually changing
Most current evidence does not show the immediate disappearance of
whole occupations. It shows the reorganization of tasks within
occupations. AI is especially effective at activities that can be
digitized, standardized, repeated, predicted, or evaluated through
recognizable patterns. Data entry, document classification, routine
customer inquiries, scheduling, transcription, basic bookkeeping,
standardized drafting, and some forms of coding or analysis can now be
completed with fewer labor hours [1, 2].
This task-based transformation matters because an occupation is
rarely a single activity. A teacher does not merely transmit
information; a nurse does not merely record symptoms; a lawyer does not
merely search documents. Each occupation combines routine tasks with
judgment, responsibility, social interaction, physical presence,
creativity, care, and accountability. AI may automate one layer while
increasing the value of another. The near-term pattern is therefore not
a clean division between “jobs eliminated” and “jobs created,” but a
struggle over how tasks, authority, income, and time are
redistributed.
Three mechanisms operate at once. Substitution occurs when an AI
system performs a task previously assigned to a worker. Augmentation
occurs when it helps a worker perform a task faster or better. Expansion
can occur when lower costs or new capabilities increase demand and
create additional activity. None of these mechanisms determines the
social result on its own. A productivity gain may finance higher
profits, lower prices, shorter hours, improved services, or new
employment. The distribution is a matter of social power and conscious
choice [2, 5].
Occupations
under the greatest displacement pressure
The strongest near-term pressure falls on occupations dominated by
structured digital routines. These roles are not necessarily vanishing
at once, but openings may decline, entry-level pathways may narrow, and
the remaining jobs may contain fewer routine duties and more monitoring,
exception-handling, or customer-facing work [1, 4].
| Occupational group | Tasks most exposed | Likely form of pressure |
|---|---|---|
| Clerical and administrative | Data entry, filing, scheduling, transcription, form processing | Fewer openings; consolidation of roles; monitoring and exception-handling remain |
| Customer support and sales | Scripted inquiries, lead qualification, telemarketing | Automation of first contact; escalation and relationship work remain |
| Finance and legal support | Document review, standard reports, basic research | Entry-level tasks narrow; verification, judgment, and accountability gain importance |
| Media, translation, and design | Routine drafting, adaptation, captioning, basic visual variations |
Downward pressure on commodity work; stronger premium on originality and trust |
| Software and data work | Boilerplate code, testing assistance, documentation, routine analysis |
Higher output expectations; fewer junior tasks; more integration and review |
The danger is
not confined to low-skilled work
Generative AI reaches into cognitive labor that earlier automation
could not easily touch. Paralegal research, financial analysis,
translation, design iteration, software development, journalism,
marketing, and administrative decision support are all exposed at the
task level. This does not mean that expertise becomes unnecessary. It
means that parts of expert work can be separated, standardized, and
transferred to machines or to lower-paid workers equipped with AI tools
[7].
Jobs and fields likely to
expand
AI adoption is also creating or expanding occupations. Some are
directly technical: machine-learning engineers, data engineers, model
evaluators, AI security specialists, infrastructure engineers, robotics
technicians, and specialists in deployment and maintenance. Other roles
arise because AI systems require human judgment and social oversight:
domain-specific trainers, auditors, safety and bias evaluators, privacy
specialists, labor-impact analysts, accessibility designers, and experts
in human-machine collaboration.
Demand may also grow in fields whose essential content is embodied,
relational, or socially accountable: nursing, therapy, elder care,
skilled trades, education, community services, maintenance, and many
forms of creative and scientific work. AI can support these occupations
without replacing the human relationship or public responsibility at
their center.
| Expanding field | Illustrative roles and functions |
|---|---|
| AI development and infrastructure | ML engineering, data engineering, compute and deployment |
| Evaluation, safety, and accountability | Auditing, model evaluation, security, privacy, bias and impact assessment |
| Human-machine integration | Workflow design, domain implementation, training, change support |
| Robotics and technical maintenance | Installation, repair, calibration, field operations |
| Human-centered services | Care, therapy, education, skilled trades, community and accessibility services |
Why new job creation is not
enough
The phrase “new jobs” must nevertheless be treated carefully. A small
number of highly paid technical positions does not automatically
compensate for a large number of disrupted clerical or service jobs. New
work may appear in different regions, demand qualifications that
displaced workers were never given an opportunity to acquire, or offer
insecure contract arrangements. Gross job creation and gross job
destruction therefore cannot be compared as if every worker can move
frictionlessly from one side of the ledger to the other.
The
hidden transformation: job quality, surveillance, and entry-level
opportunity
Employment totals alone conceal major changes in the quality of work.
AI can remove drudgery, but it can also enable continuous measurement of
pace, behavior, communication, and output. When management controls the
system and workers have little voice, augmentation can become work
intensification: the employee produces more in the same time while the
employer captures the gain. Algorithmic scheduling and performance
scoring can reduce autonomy even when no job is formally eliminated [2,
8].
Entry-level employment deserves particular attention. Routine
assignments have traditionally allowed new workers to learn a
profession, acquire judgment, and progress toward greater
responsibility. If these tasks are automated without creating a new
pathway for apprenticeship and supervised practice, organizations may
save money today while weakening the future supply of experienced
workers. Early evidence of slower hiring in some AI-exposed occupations
should therefore be read not only as a short-term labor-market
adjustment but as a warning about the reproduction of knowledge across
generations [6].
The burden is also unequal. Women are highly represented in many
clerical and administrative occupations. Workers with fewer financial
reserves cannot endure long retraining periods. Regions dependent on a
narrow range of employers may lose both income and community stability.
Existing racial, gender, regional, disability, and class inequalities
can be reproduced in new technological forms unless transition policies
are deliberately universal and redistributive.
Why retraining alone is not
enough
Education and continuous learning are necessary, but “reskilling” is
often presented as though displacement were an individual failure to
adapt. That framing hides the institutional question. Workers did not
decide the pace of automation, the design of the system, the allocation
of investment, or the distribution of the resulting savings. Training
cannot solve a shortage of good jobs, unequal bargaining power,
geographic dislocation, or the private capture of socially produced
knowledge.
AI itself is produced socially. Models depend on generations of
science, publicly financed research, global digital infrastructure,
accumulated culture, the labor of engineers and data workers, and the
daily activity of millions of users. Yet ownership of the systems and
the income they generate is concentrated in a small number of
corporations and investors. This sharpens the contradiction between
increasingly socialized production and private appropriation. The
labor-market crisis is one expression of that deeper contradiction.
A serious response must therefore combine access to education with
income security, worker power, democratic governance, and changes in the
ownership and control of productive systems. Otherwise society
repeatedly prepares people to compete for a shrinking share of work
while productivity gains flow upward.
From
defensive adjustment to a conscious social program
A humane transition requires more than slowing technology or
preserving every existing job. Some work is dangerous, degrading,
monotonous, or unnecessary and should disappear. The objective is not to
defend labor as endless compulsion; it is to prevent the elimination of
tasks from becoming the elimination of human security and social
participation.
The following principles translate AI productivity into shared social
progress:
-
Reduce working hours without reducing living standards when
productivity rises. -
Guarantee income, healthcare, housing, and education during
transitions, so technological change does not become personal
catastrophe. -
Give workers an enforceable voice before AI systems alter staffing,
evaluation, scheduling, safety, or wages. -
Require transparent labor-impact reporting, independent audits, and a
right to challenge consequential algorithmic decisions. -
Direct a share of AI-generated productivity gains into public funds,
universal services, and worker-controlled transition programs. -
Build permanent, publicly supported education rather than temporary
training tied to one employer or product. -
Protect the right to organize, bargain collectively, strike, speak,
and participate in decisions about technology. -
Develop public, cooperative, and commons-based AI infrastructure so
essential knowledge systems are not governed exclusively by private
monopolies.
Participation
in power, not consultation after the decision
These measures should not be imposed by a remote bureaucracy.
Workplaces, professional associations, unions, neighborhood councils,
public institutions, and affected communities need direct and continuing
participation in decisions about deployment. Transparency must include
what a system does, what data it uses, how it affects staffing and
wages, who can challenge its decisions, and how productivity gains are
allocated. Participation in power – not merely consultation after
decisions are made – is the democratic core of the transition.
Sectoral
differences and the limits of prediction
AI will not affect every sector at the same speed.
Information-intensive fields such as administration, finance, law,
media, software, and parts of engineering have high task exposure
because much of their work already exists in digital form. Care,
construction, maintenance, hospitality, transportation, and
manufacturing contain more physical or interpersonal tasks, although
robotics and autonomous systems may expand automation there as well.
Healthcare and education illustrate the central choice particularly
clearly: AI can reduce paperwork and give professionals more time with
people, or it can become a pretext for understaffing and standardized,
impersonal service.
Quantitative forecasts should be treated as scenarios, not destiny.
The source report cites a World Economic Forum projection of 92 million
jobs displaced and 170 million roles created by 2030. Such figures
describe modeled global churn, not guaranteed outcomes [3], and they do
not establish the quality, location, accessibility, or ownership
structure of the new work. Exposure measures likewise indicate that a
task could be affected; they do not prove that employers will automate
it or that a whole occupation will disappear.
The evidence available through 2026 remains mixed. Aggregate
unemployment has not yet shown a simple, universal AI shock, while some
studies identify slower hiring for younger workers and reductions in
openings in exposed occupations. Causation is difficult to isolate
because interest rates, outsourcing, post-pandemic adjustment, corporate
restructuring, and other technologies operate simultaneously. Honest
analysis must preserve this uncertainty while recognizing the direction
of structural pressure [6].
Conclusion:
the future of work is a question of social order
AI has made visible a possibility that earlier societies could only
partially imagine: the production of abundance with far less compulsory
human labor. Under the logic of private accumulation, that possibility
appears as unemployment, insecurity, surveillance, and a race among
workers to remain economically necessary. Under a new social logic, it
could mean shorter working time, secure access to life’s necessities,
expanded education, creative activity, care, and direct participation in
collective decisions.
The decisive conflict is therefore not between humans and machines.
It is between two ways of organizing the relationship among technology,
labor, and society. One treats AI as private capital whose purpose is to
reduce costs and maximize returns. The other recognizes AI as a socially
produced capacity that should be governed for human development and
common need.
Technology creates possibilities; consciousness and organization
determine which possibilities become reality. The task before us is not
merely to adapt workers to AI. It is to bring AI under conscious social
direction, so that liberation from unnecessary labor becomes freedom for
everyone rather than privilege for a few.
References and source note
The attached Stanford-generated report was the starting point, but
its abbreviated bibliography included mixed-quality sources. The revised
article relies chiefly on the primary institutional reports, official
labor statistics, and research papers listed below. Bracketed numbers in
the article correspond to this list. Forecasts remain scenarios rather
than certainties, and exposure is not treated as proof of job
elimination.
[1] Gmyrek, P., Berg, J., Kaminski, K., Konopczynski, F., Ladna, A.,
Nafradi, B., Roslaniec, K., & Troszynski, M. (2025). Generative AI
and Jobs: A Refined Global Index of Occupational Exposure. ILO Working
Paper 140. International Labour Organization. DOI: 10.54394/HETP0387. ILO
publication
[2] OECD. (2023). OECD Employment Outlook 2023: Artificial
Intelligence and the Labour Market. OECD Publishing. OECD
report
[3] World Economic Forum. (2025). The Future of Jobs Report 2025. WEF
report
[4] U.S. Bureau of Labor Statistics. (2026). Occupations with the
Largest Job Declines: Employment Projections, 2025-2035. BLS
table
[5] Brynjolfsson, E., Li, D., & Raymond, L. R. (2023; published
2025). Generative AI at Work. NBER Working Paper 31161; Quarterly
Journal of Economics, 140(2), 889-942. NBER paper
[6] Brynjolfsson, E., Chandar, B., & Chen, R. (2026). Canaries in
the Coal Mine? Six Facts about the Recent Employment Effects of
Artificial Intelligence. Stanford Digital Economy Lab. Stanford
paper
[7] Stanford Institute for Human-Centered Artificial Intelligence.
(2025). Assessing the Real Impact of Automation on Jobs, summarizing
David Autor’s Stanford Digital Economy Lab presentation. Stanford
HAI article
[8] Hayes, M., & Northup, J. (2026). AI Powers into the
Workplace. ADP Research, based on the 2025 Global Workforce Survey. ADP
Research
[9] Akhavi, H. (2026). A New Social Order in the Age of
Consciousness, Volume I. Conceptual framework used for the article’s
interpretation and conclusions.
