BY Florent Herisson / エリソンフロー MARCH 2026 OSAKAWIRE INTELLIGENCE EN FR JP
INTELLIGENCE REPORT SERIES ISSUE 003 MARCH 2026 OPEN ACCESS

SERIES: WHAT IS ACTUALLY HAPPENING

AI & Employment
The Honest Evidence

A sourced, calibrated analysis of what AI is doing to work — separated by what is established fact, what is contested, and what is myth. Neither panic nor reassurance. Evidence.

Primary SourcesIMF, WEF, ILO, Goldman Sachs, OECD, Brookings, Stanford Digital Economy Lab, Harvard, Federal Reserve
Evidence TiersEvery factual claim is categorised and sourced inline
Reading TimeApprox. 3–4 hours comprehensive
Last VerifiedMarch 2026
Evidence Tier Key → ✓ Established Fact ◈ Strong Evidence ⚖ Contested ✕ Misinformation ? Unknown
01

Why This Topic
Generates More Heat Than Light

Three distinct camps are shouting past each other. None is entirely right. Understanding the structure of the debate is the prerequisite to understanding the evidence.

Few topics generate more confident, contradictory claims per column inch than artificial intelligence and jobs. In a single week, credible outlets will publish "AI is coming for half of all white-collar jobs" and "AI creates more jobs than it destroys — history proves it." Both headlines are technically defensible. Neither is the full story.

The confusion is structural, not accidental. It arises from three distinct debates being collapsed into one: what is happening now (empirical, measurable), what will happen by 2030 (contested projection), and what will happen over 20–50 years (genuinely unknown). Researchers who are pessimistic about long-run structural employment and researchers who are optimistic about near-term aggregate job creation are sometimes citing different time horizons — and both can be correct simultaneously.

There is also a political economy to the optimism. Companies building AI have a vested interest in the narrative that their technology creates more jobs than it destroys. Unions and displaced workers have a vested interest in documenting harm. Neither group is necessarily dishonest — both are applying selection pressure to which evidence gets amplified.

Camp A: Techno-Optimists

Core claim: History shows technology always creates more jobs than it destroys. The lump-of-labour fallacy is real.
Key figures: WEF (net +78M jobs by 2030), Goldman Sachs (15% productivity boost), Erik Brynjolfsson (MIT)
Best evidence: 60% of US jobs today didn't exist in 1940. Multiple waves of automation expanded total employment.
What they understate: Transition costs are real and long. Past patterns may not hold if AI is fundamentally different in speed and scope.

Camp B: Structural Pessimists

Core claim: This time is different. AI displaces cognitive work, not just manual tasks. Speed of change will outpace adaptation.
Key figures: Daron Acemoglu (MIT), Dario Amodei (Anthropic), Oxford's Frey & Osborne
Best evidence: New job creation has slowed since 1970. Labour's share of income is falling. Entry-level workers already being squeezed.
What they understate: Early predictions (47% automation risk, 2013) proved too dire. New tasks genuinely do emerge.
The third camp
Process realists (IMF, ILO, Brookings, most academic economists): Accept that net job creation is likely, but argue the distributional consequences are severe. Jobs being created are not going to the people whose jobs are being destroyed. The policy failure is not "too few jobs" but "wrong jobs, in wrong places, requiring wrong skills, at wrong speed." This report operates primarily in this camp.

This report attempts to be useful by being disciplined: every claim is explicitly categorised by evidence tier. Where the data is clear, we say so. Where it is contested, we say so and present both credible sides. Where claims are not supported by evidence — regardless of which camp makes them — we say that too.

02

What We Actually Know
The Data Landscape

Separating the numbers that are well-established from the projections that are contested — and being honest about the difference.

40%
of global employment exposed to AI in some capacity
IMF SDN 2024 · ◈ Strong Evidence
92M
jobs projected displaced by 2030 (global)
WEF Future of Jobs 2025 · ⚖ Contested
170M
new jobs projected created by 2030 (global)
WEF Future of Jobs 2025 · ⚖ Contested
6–7%
of US workforce at risk of direct displacement if widely adopted
Goldman Sachs Research 2025 · ◈ Strong Evidence
20%
decline in employment, software devs aged 22–25 from 2022 peak
Goldman Sachs 2025 · ✓ Established
75%
of knowledge workers already using AI tools at work
Microsoft/LinkedIn Work Trend Index 2025 · ◈ Strong Evidence
66%
productivity improvement reported by AI users in knowledge work
Nielsen Norman Group 2025 · ⚖ Contested
86%
of 6M most vulnerable displaced workers are women
Brookings/Lightcast 2026 · ◈ Strong Evidence
Critical caveat: models vs measurement
Almost all large-scale figures (92M displaced, 170M created, 47% at risk) are model outputs based on task-exposure analysis, not observed employment data. The WEF, IMF, Goldman Sachs, and McKinsey figures estimate what could happen if AI capabilities spread and are deployed optimally. Actual employment data measuring what has already happened is much more limited — and shows more modest (though real) effects so far. Do not treat projections as certainties.

What the Data Actually Confirms (Right Now)

✓ Established Fact Entry-level hiring in AI-exposed sectors is already contracting

Goldman Sachs Research (2025) documents a measurable, statistically significant drop in employment specifically among workers aged 22–25 in AI-exposed occupations. Unemployment among 20–30 year-olds in tech-exposed roles rose by ~3 percentage points since early 2025. Software developers aged 22–25 saw a ~20% drop in employment compared to their late-2022 peak. This is corroborated by Stanford Digital Economy Lab's "Canaries in the Coal Mine" (Brynjolfsson et al., 2025), which identified the same cohort as the leading indicator of AI labour market impact.

Importantly: overall employment continues to rise. This is not a macro employment collapse — it is a targeted compression of the entry-level hiring pipeline in specific AI-exposed roles. The signal is concentrated in who gets hired to start their career, not in mass layoffs.

✓ Established Fact AI significantly boosts individual worker productivity in exposed roles

Multiple high-quality randomised experiments confirm that AI tools genuinely raise productivity. Brynjolfsson, Li & Raymond (2025, Quarterly Journal of Economics) — the most rigorous study available — found that generative AI raised customer support worker productivity by an average of 14%, with the largest gains for lowest-skilled workers. GitHub Copilot studies show coding speed improvements of ~55% for developers. The key finding is that AI currently functions as a productivity amplifier, not a replacement, for most active workers. The replacement dynamic is manifesting in hiring — fewer new people needed — rather than in layoffs of existing workers.

◈ Strong Evidence Labour's share of income was already declining before AI, and AI is likely to accelerate this

Labour's share of income in US nonfarm businesses fell from ~64% in 1980 to ~57% in 2017 (Acemoglu, Manera & Restrepo, 2020). This pre-AI trend reflects decades of capital-biased automation, where the effective tax rate on labour (~25–34%) far exceeds that on capital (~5–10%), incentivising substitution. IMF research (2024–2025) projects that AI is likely to further increase returns to capital at the expense of labour income — but this effect depends heavily on whether AI complements or substitutes high-income workers, and the degree to which productivity gains are captured by capital owners versus distributed as wages.

"The most widespread impact of generative AI is likely to be on job quality rather than job quantity."

— International Labour Organization (ILO/NASK Global Index, May 2025)
03

Occupation Risk Matrix
Which Jobs, By What Mechanism, On What Timeline

Risk level, exposure mechanism, and projected timeline — sorted by occupational category. Risk percentages are model-based estimates (see Section 02 caveat).

How to read this table
"Exposure" ≠ "certain displacement." A job with 70% task exposure means AI can technically perform 70% of its tasks today — not that 70% of those jobs will disappear. Whether that translates to job loss, job transformation, or wage suppression depends on adoption speed, organisational choices, regulation, and the emergence of new tasks. The mechanism column matters most.
Occupation Task Exposure Primary Mechanism Timeline Evidence
Customer Service Representatives
80%
LLM chatbots handle Tier-1 inquiries; human roles shrink to exception handling 2024–2026 (already active) ✓ Established
Data Entry Clerks
90%
Direct automation of repetitive data processing; high-accuracy OCR + AI 2024–2027 ✓ Established
Administrative / Secretarial Assistants
75%
Scheduling, drafting, document management, email — all AI-replicable 2024–2028 ◈ Strong Evidence
Translators / Interpreters
70%
LLMs now at near-human quality for standard commercial translation; documented employment decline 2023–2026 (ongoing) ✓ Established
Bookkeepers / Accounting Clerks
72%
Routine financial data processing fully automatable; AI accounting software scaling fast 2025–2028 ◈ Strong Evidence
Proofreaders / Copy Editors
65%
Grammar/style correction tasks now performed by AI at superior accuracy for standard content 2023–2026 ◈ Strong Evidence

Sources: ILO/NASK Global Index 2025; Goldman Sachs Research 2025; McKinsey Global Institute; IMF SDN 2024

Occupation Task Exposure Primary Mechanism Timeline Evidence
Junior Software Developers
60%
Code generation tools reduce entry-level role requirements; senior roles become more productive 2024–2028 ✓ Established
Paralegals / Legal Assistants
58%
Document review, legal research, contract analysis — all LLM-replicable at speed 2025–2029 ◈ Strong Evidence
Financial Analysts (entry level)
55%
Routine analysis, report generation, data synthesis now automated; senior judgment retained 2025–2030 ◈ Strong Evidence
Journalists / Content Writers
52%
Data journalism and standardised content generation automated; investigative/narrative less so 2024–2028 ◈ Strong Evidence
Radiologists (screening layer)
50%
AI outperforms humans on initial image screening; role shifts toward complex diagnosis, communication 2026–2032 ⚖ Contested
Retail Cashiers
65%
Self-checkout + frictionless retail scaling; Amazon Go model expanding 2024–2030 ◈ Strong Evidence
Occupation Task Exposure Primary Mechanism Notes Evidence
Teachers / Educators
35%
Administrative tasks + standardised content generation automated; core instruction and mentoring resilient Role transformation likely; volume reduction unlikely short-term ⚖ Contested
Accountants / Auditors (senior)
40%
Routine elements automatable; complex judgment, client relationship, liability-facing work resilient Bifurcation: junior roles squeezed, senior roles amplified ◈ Strong Evidence
Marketing Specialists
45%
Content creation, A/B testing, campaign analysis rapidly automating; creative strategy less so Productivity amplifier currently; displacement coming at entry level ◈ Strong Evidence
Human Resources
38%
Screening, scheduling, admin automated; culture, conflict resolution, judgment-heavy work not ATS (applicant tracking systems) already AI-driven ◈ Strong Evidence
Truck / Delivery Drivers
30%
Autonomous vehicles technically approaching viability; regulatory, insurance, last-mile delays remain High volume of affected workers (3.5M in US alone); timelines consistently delayed ⚖ Contested
Occupation Task Exposure Why Resilient Risk Level Evidence
Plumbers / Electricians / Skilled Trades
8%
Requires physical dexterity in unstructured environments; robots cannot yet perform reliably or cost-effectively Low — 10–20 year horizon minimum ✓ Established
Registered Nurses
12%
Physical care, patient communication, emotional labour, clinical judgment in unstructured situations Low for displacement; high for productivity augmentation (AI diagnostics) ✓ Established
Mental Health Therapists
10%
Therapeutic relationship, empathy, nuanced human judgement; AI tools as supplements, not replacements Low — regulatory and ethical barriers high even if AI improves ◈ Strong Evidence
Early Childhood Educators
7%
Physical care, relationship formation, developmental monitoring — not replicable by AI Very low ✓ Established
Senior Executives / CEOs
15%
Strategic judgment, relationship capital, accountability, ambiguous decision-making in novel situations Low — but AI will amplify productivity of those who adopt it ◈ Strong Evidence
Construction Workers
10%
Physical manipulation in unstructured, variable environments; robotics not yet viable at scale Low for 10+ years; potentially higher 2030–2040 ✓ Established
The "Resilient Jobs" Pattern
Four characteristics dominate jobs that are genuinely hard to automate: (1) physical dexterity in unstructured environments, (2) emotional and relational care, (3) ethical accountability that cannot be delegated to a machine, (4) real-time adaptation to unpredictable human situations. Jobs combining multiple of these factors are the most secure for the next 15+ years.
04

Who Gets Hit First
The Hidden Fault Lines

The aggregate numbers obscure radically different experiences across gender, age, education, and income. The same "AI creates more jobs" headline can be simultaneously true in the aggregate and catastrophic for specific populations.

Gender Warning — ILO/NASK Global Index 2025 (March 2026 update)
Female-dominated occupations are almost twice as likely to be exposed to generative AI as male-dominated ones: 29% vs 16%. At the highest exposure level (Gradient 4), the gap widens to 9.6% female vs 3.5% male in high-income countries. This is not a secondary concern — it is the most demographically concentrated finding in the entire AI employment literature.
79%
of working women employed in AI-exposed occupations
University of St. Thomas / Euronews analysis 2024 ·
66%
of men employed in AI-exposed occupations
Euronews analysis 2024 ·
86%
of most vulnerable AI-displaced workers (those least able to adapt) are women
Brookings/Lightcast 2026 ·
25%
lower rate of AI tool adoption among women vs men
Harvard Business School, Koning et al. 2025 ·

The structural reason: Between 93–97% of secretary and administrative assistant positions in the US were held by women between 2000–2019 (US Census Bureau). These are Tier-1 occupations for AI displacement. The ILO finds that the overrepresentation of women in clerical and administrative roles is the primary driver of the gender gap in AI exposure — not any inherent characteristic of women's work being uniquely automatable.

The compounding problem: Women are not only concentrated in higher-risk jobs — they are adopting AI tools at lower rates, making them less likely to shift from "AI replaces me" to "AI amplifies me." Research suggests women face additional social penalties for using AI tools (concerns about being perceived as "cheating" or less intelligent) that men do not face to the same degree.

The bias layer: AI systems trained on historical data reproduce and can amplify existing gender biases in recruitment, pay decisions, and credit scoring — creating risk in both the jobs lost and the jobs applied for. The ILO notes that women are underrepresented in AI development (only 22% of AI professionals globally per WEF 2025), making self-correction through diverse development teams structurally difficult.

~20%
Employment drop for software devs aged 22–25 from late-2022 peak
Goldman Sachs Research Aug 2025 ·
+3pp
Unemployment increase for ages 20–30 in AI-exposed tech roles since early 2025
Goldman Sachs Research Aug 2025 ·
1.4×
Millennials (35–44) more likely to report strong familiarity with gen AI tools than other age groups
McKinsey survey 2025 ·
129%
Gen Z workers more likely than over-65s to worry about AI making their job obsolete
PwC/SSRN 2025 ·

The entry-level compression: The clearest and most documented age effect is the compression of entry-level hiring. AI is reducing the need for junior workers precisely in the roles that traditionally served as the first rung of professional career ladders: junior developer, junior analyst, junior paralegal, customer service rep. The pipeline to senior roles is narrowing before those roles are themselves threatened.

The irony for Gen Z: The generation most worried about AI is not the generation losing jobs in the aggregate — overall employment is not crashing. They are the generation finding the door to their career ladder narrower than it was for previous cohorts. This is a real harm even if the macro numbers look fine.

Older displaced workers: Workers aged 50+ who lose AI-exposed jobs face the most severe transition challenges. The Boston Fed's research (December 2024) found that about 21% of surveyed workers expected AI to worsen their financial situation within 5 years, concentrated heavily in this older cohort. Retraining for new sectors is harder, takes longer, and has lower returns at this life stage — this group is the one identified by Brookings as most vulnerable.

The Education Paradox — IMF SDN 2024
Unlike previous automation waves (which primarily displaced low-educated workers), AI disproportionately exposes higher-educated workers to automation risk. However, higher-educated workers also have the highest AI complementarity — meaning they are also more likely to benefit. The same doctor, lawyer, or analyst whose work is most AI-exposed is also best positioned to use AI to become dramatically more productive. The risk is bifurcation within educated professions, not across education levels.

IMF Working Paper (Rockall, Tavares, Pizzinelli, 2025) distinguishes between three occupational groups: HELC (High Exposure, Low Complementarity — the danger zone), HEHC (High Exposure, High Complementarity — the amplified zone), and LE (Low Exposure — largely unaffected). The critical policy question is which workers fall where.

Low education (no college degree): Lower immediate exposure to AI (IMF: 26% for low-income country workers vs 60% for advanced economies), but also lower capacity to transition to AI-complementary roles. The "protection" of not being in the AI economy's crosshairs is partly an artefact of not yet having access to the digital infrastructure that enables both the risk and the opportunity.

College degree holders: 44% acknowledge AI can perform some of their tasks (vs 22% without college) — higher awareness, but also higher adaptive capacity. Studies confirm workers with post-secondary education experience AI more as a complement to their capabilities than a substitute.

The Brookings Institution confirmed: "Better-paid, better-educated workers face the most exposure." But exposure does not mean harm if complementarity is high. The real danger is the layer of workers with enough education to be in AI-exposed roles but without the seniority, adaptability, or resources to pivot to complementarity.

The class dimension of AI employment impact is the most politically combustible and analytically contested aspect of this topic. It requires careful separation of two different dynamics operating simultaneously.

Capital Concentration Risk

Mechanism: AI raises productivity, productivity gains flow to capital owners (shareholders, IP holders) rather than workers. Labour's share of income falls further from its already-reduced 57% of US nonfarm income.
Evidence: Acemoglu & Johnson (IMF Finance & Development, 2023): "On our current trajectory, the first-order impact is likely to be increased inequality within industrial countries."
IMF modelling: With high AI-capital complementarity, higher-wage earners see more-than-proportional income increase, amplifying both labour income inequality and wealth inequality through enhanced capital returns.

Wage Compression Offset

Mechanism: AI primarily displaces high-income workers' tasks, potentially reducing wage inequality by making top-earner tasks more contestable.
Evidence: Brynjolfsson et al. (2025, QJE) found largest productivity gains for lowest-skilled workers using AI tools — AI as a skill leveller in the short term.
Caveat: Brookings warns this near-term productivity gain for low-skill workers is likely transitional. As technology matures, the same workers face displacement risk rather than augmentation.
The Klarna Test Case
Klarna reported their AI system performed the work of 700 customer service agents approximately one year after laying off 700 employees. Brookings's analysis: the 700 displaced agents were not the ones promoted into supervisory or AI-complementary roles — they were the ones automated out entirely. This is the "hollowing out" risk: AI boosts productivity of the firm while concentrating gains at the top. The Klarna case is a single data point, not a proof of general pattern — but it illustrates the distributional risk that makes aggregate "net positive" job creation statistics feel hollow to those in the displaced cohort.
05

The Global South Problem
Where "New Jobs Will Emerge" Fails

The most common rebuttal to AI displacement fears is that new jobs will emerge, as they have in previous technological transitions. In wealthy countries, this is a contested but plausible case. In the developing world, it is much harder to make.

High-Income Countries
34%
Highest exposure; also highest capacity to benefit. White-collar service concentration. Digital infrastructure strong. New AI-complementary jobs most likely to emerge here.
Emerging Market Economies
~21%
Moderate exposure, moderate capacity to benefit. Risk of "double exposure": outsourced service jobs (call centres, data entry) being automated from developed-world employers' end without replacement local options.
Low-Income Countries
11–26%
Lower exposure due to digital divide, but lower capacity to benefit from AI either. ILO warns: "lower exposure does not equal lower risk" — weak labour protections amplify any disruption that does occur.
Global South Outsourcing Hubs
HIGH
Philippines, India (call centres, BPO), Kenya (content moderation, data annotation): these workers face displacement decisions made by corporations in other countries. No local political accountability.
The Historical Precedent Failure — LSE Media@LSE, Nov 2025
Optimists cite the Industrial Revolution as proof that new jobs emerge to replace automated ones. The LSE analysis identifies a critical asymmetry: those new jobs emerged in the same geographic regions where the old jobs disappeared. A factory worker in 19th-century Lancashire could become a factory worker in a new industry nearby. A call centre worker in Manila whose job is automated by AI in 2026 cannot become an AI developer — the new jobs require different skills, different infrastructure, and are created in different countries. Geographic continuity does not hold for AI-driven outsourcing displacement.

India's paradox: India aspires to become a major AI hub, with its AI market projected to grow at 25–35% CAGR by 2027. Yet India's ~$250 billion IT and business services sector — which employs millions in English-speaking, outsourced cognitive roles — is precisely the sector most exposed to AI automation from Western corporations cutting costs. The workers who benefit from India's AI ambitions and those who lose jobs to it are entirely different populations, separated by education, language, location, and income.

The data annotation trap: A significant share of "new jobs" in AI for the Global South consists of data labelling, content moderation, and AI training work — often paying $1–2.50/hour in Kenya, and similar in Bangladesh and India. These workers are doing the unglamorous labour that makes AI systems function, with minimal protections, no career pathway, and exposure to psychologically harmful content. The UNCTAD warned that AI could reduce the competitive advantage of low-cost labour in developing countries — the one economic lever they have — without creating equivalent alternative opportunity.

◈ Strong Evidence The global displacement–creation timing mismatch is most severe in developing economies

IMF research (2024) and ResearchGate analysis (2025) confirm: displacement concentrates in 2024–2027 while job creation spreads across longer timelines. In advanced economies, the institutions, safety nets, and educational systems to manage this transition exist (imperfectly). In developing economies experiencing displacement of outsourced roles, those institutional buffers are absent. The result: developing economies experience displacement without offsetting creation, widening international inequality.

In Latin America: ~25% of jobs in Brazil, Chile, Colombia, Mexico, and Peru exhibit high exposure to AI yet low task complementarity — rendering them highly vulnerable to substitution. For workers in call centres and outsourced services specifically, this risk is characterised as "acute" by ILO researchers.

06

What History Says
And Where the Analogy Breaks Down

The most powerful argument for labour market optimism is 200 years of evidence that technology creates more jobs than it destroys. This argument deserves serious engagement — and serious scrutiny.

The Historical Case for Optimism

1940–2025
60% of US jobs today did not exist in 1940 (Autor, Chin, Salomons, Seegmiller — MIT, 2024). More than 85% of employment growth since 1940 came from technology-driven job creation. The agricultural sector, which employed 40%+ of US workers in 1900, now employs <2% — yet overall employment rates are higher. ✓ Established
1990–2017
The Great Stability. Harvard economists Deming and Summers (2025, published) found occupational churn data showed a slowdown in technological disruption from 1990–2017 — the period of maximum "automation anxiety" paradoxically had the lowest disruption rate. ◈ Strong Evidence
2013
Frey & Osborne "47% at risk" (Oxford Martin Institute) — widely cited, over 17,000 academic citations, 44M Google results. Received near-apocalyptic coverage. Follow-up data showed the predicted displacement largely did not materialise on the predicted timeline. ✓ Established
2019→
Deming & Summers detect real shift. "From 2019 onward, it looks like things were changing quite a lot." Harvard data identifies AI as a genuine general-purpose technology disruption comparable to electrification and computing. STEM jobs grew from 6.5% to nearly 10% of US employment 2010–2024. ◈ Strong Evidence

Where the Historical Argument Fails

⚖ Contested "Technology always creates more jobs than it destroys" — the lump-of-labour counterargument

The standard economic rebuttal: The "lump of labour fallacy" — the mistaken belief that there is a fixed amount of work to be done — is a real logical error. New technologies create new demand, new industries, new occupations we cannot predict in advance. Federal Reserve Governor Barr (May 2025): economists have long been sceptical of the assumption that automation leads to permanent unemployment.

The Acemoglu counterpunch (from MIT/IMF, December 2023): "There is no guarantee that, on its current path, AI will generate more jobs than it destroys." The historical pattern of new job creation relied on a balance between automation and new task creation. Sometime after approximately 1970, this balance was lost. Labour's share of income has been falling for 50 years. New task creation has slowed, particularly for workers without four-year college degrees. AI may accelerate an already-broken dynamic, not reverse it.

The speed argument: Historical transitions took generations. The loom displaced weavers over 50–100 years; workers' children adapted. AI is potentially compressing equivalent transitions to 5–10 years. Even if the long-run outcome is net positive, transition costs measured in human lives — income loss, psychological distress, family disruption — are real and concentrated in specific populations who cannot simply "wait for the new jobs."

⚖ Contested Is this time genuinely different? The "GPT is a GPT" thesis

The term "General Purpose Technology" (GPT in the economic sense) refers to technologies that reshape multiple sectors simultaneously — electricity, computing, the internet. Deming and Summers (2025) concluded that AI qualifies as a GPT of this magnitude.

What is arguably different about AI vs previous GPTs:

1. Previous GPTs automated physical or narrow cognitive tasks. AI is the first technology capable of performing general reasoning, language, and creative tasks — the work previously considered uniquely human and automation-proof. 2. Previous GPTs created new tasks that required human labour to perform. The new tasks AI creates (AI trainer, AI ethics officer, AI product manager) require far fewer workers relative to the tasks they replace. Prompt engineers — once predicted to be a large occupation — comprise less than 0.5% of LinkedIn job postings. 3. The capital-to-labour substitution incentive is structurally embedded in the US tax code (labour taxed at ~30%, capital at ~8%), making replacement the rational choice for any corporate actor.

"The US economy had 2.5 industrial robots per thousand workers in manufacturing in 1993. This number rose to 20 by 2019. Excessive automation has caused a decline in labour's share of income from 64% in 1980 to 57% in 2017."

— Acemoglu, Manera & Restrepo, cited in Chicago Booth Review
07

The Productivity Paradox
AI Boosts Output. Who Captures the Gains?

AI is raising productivity in AI-exposed sectors. This is well-evidenced and not seriously disputed. The crucial, contested question is whether those productivity gains translate into broader prosperity or concentrate further at the top.

14%
productivity increase for customer support workers using AI tools (largest gains for lowest-skill workers)
Brynjolfsson, Li & Raymond, QJE 2025 — gold-standard RCT ·
55%
coding speed improvement for developers using GitHub Copilot
Peng et al., MIT/Microsoft 2023 ·
15%
projected increase in labour productivity in US/developed markets when AI fully adopted
Goldman Sachs Research 2025 ·
+7%
projected increase in global GDP from AI over the next decade
Goldman Sachs Research 2025 ·

The productivity evidence is real. Randomised controlled experiments — the gold standard of social science — confirm AI tools raise output in professional settings. The question is not whether productivity increases, but who captures that increase.

⚖ Contested The "productivity bandwagon" — does productivity growth reach workers?

Acemoglu and Johnson (Power and Progress, 2023) introduce the concept of the "productivity bandwagon": the idea that for the majority of people to benefit from productivity growth, that productivity must be "anchored" to improved efficiency of human labour — raising workers' marginal productivity — rather than simply automating human tasks and capturing the gains as capital income.

The EPI (Economic Policy Institute) analysis adds that the effective tax rate on labour is approximately double that on capital in the US, meaning companies are structurally incentivised to substitute capital for labour even when it is not the most economically efficient choice. Brynjolfsson (MIT) recommends equalising effective tax rates on labour and capital as the most direct intervention to change this incentive structure.

The 1990s counter-evidence: EPI research shows the 1990s — which saw massive technology-driven productivity growth from the internet — resulted in broad-based wage growth and declining unemployment, not concentrated gains. The explanation: unemployment was driven sufficiently low to generate genuine bargaining power for workers. The policy lesson is that macro employment conditions matter as much as technology itself for whether productivity gains get distributed.

The J-Curve Effect — Brynjolfsson, Rock & Syverson (AEJ 2021)
General purpose technology transitions typically follow a "productivity J-curve": measured productivity falls in the short run as organisations reorganise, workers train, and workflows are redesigned — then rises dramatically once the transition is complete. We may currently be in the trough of the J-curve for AI — seeing costs and disruption before the full productivity gains materialise. This is a reason for calibrated optimism about the long-run outcome while being clear-eyed about near-term disruption.
08

The Misinformation Layer
Claims That Are Not Supported by Evidence

Both the catastrophist and the dismissive camps produce widely-shared claims that are not supported by the evidence. This section identifies the most common ones on both sides.

Catastrophist Myths

"47% of US jobs will be automated within 20 years" — widely repeated since 2013
Evidence Tier: Misinformation (as stated)

Frey & Osborne (Oxford Martin, 2013) produced a highly cited model predicting 47% of US occupations were at high risk. Harvard Data Science Review (Fall 2025) documents that this was a task-level analysis incorrectly extended to whole jobs. The OECD replication applying their own methodology arrived at 9% — five times lower. More critically: the occupations flagged as "at risk" in 2013 (tax preparers, telemarketers, insurance underwriters) have not, in fact, disappeared at scale over the subsequent 12 years. The 47% figure is technically a model output from 2013 projected with significant methodological caveats — presenting it as fact is misinformation.

"AI is about to eliminate 300 million full-time jobs globally"
Evidence Tier: Misinformation (as stated)

This Goldman Sachs figure (2023) is frequently misquoted. The original report stated that 300 million full-time job equivalents could be exposed to automation if AI were widely adopted — a task-exposure estimate under an optimistic AI deployment scenario. The same report projected that the most likely displacement scenario is 6–7% of the US workforce, with unemployment rising by just 0.5 percentage points above trend during the transition period, before recovering within approximately two years. The 300M figure is real; presenting it as a near-term mass unemployment forecast is not.

"AI will replace half of all white-collar entry-level jobs within five years" — Dario Amodei (2025)
Evidence Tier: Contested (stated with more certainty than evidence warrants)

Anthropic CEO Dario Amodei stated in 2025 that AI could eliminate roughly 50% of white-collar entry-level positions within five years. Nvidia CEO Jensen Huang explicitly pushed back. The evidence shows real and documented compression of entry-level hiring in AI-exposed sectors — especially tech. However, "50% of white-collar entry-level jobs" across all industries within 5 years would require adoption speed and scope that current data does not confirm. The underlying concern is legitimate; the specific number and timeline is not well-evidenced.

Dismissive Myths

"Technology always creates more jobs than it destroys — AI is no different"
Evidence Tier: Contested (historically valid but not necessarily predictive)

The historical pattern is real: 60% of today's US jobs didn't exist in 1940. But Acemoglu and Johnson document that new task creation has slowed since 1970, the balance between automation and job creation is already off-kilter, and the speed of AI adoption may compress transitions that historically took generations. The pattern's past validity does not guarantee future validity — particularly when AI is the first technology to threaten general reasoning tasks rather than just specific manual or narrow cognitive ones. The lump-of-labour fallacy is a real economic error; dismissing AI risk entirely by invoking it is also an error.

"Workers just need to reskill and they'll be fine"
Evidence Tier: Misinformation (as a complete policy prescription)

The evidence on retraining programmes is sobering. "The China Shock" (Autor, Dorn & Hanson, 2016) — the most impactful US economics paper of the last decade — demonstrated that import competition from China devastated large parts of the American workforce, and that retraining programmes largely failed to produce successful transitions. The US Workforce Investment and Opportunity Act (WIOA): as of 2023–24, fewer than 10% of training participants received on-the-job training; just 2% received apprenticeships. The successful retraining examples are rare. Telling displaced 55-year-old manufacturing or clerical workers to "reskill" without addressing structural barriers of cost, time, psychological difficulty, and age discrimination is not a policy — it is a reassurance that fails the most vulnerable workers.

"Prompt engineering will be the hot job of the future"
Evidence Tier: Misinformation (as stated)

Harvard Data Science Review (Fall 2025) documents: prompt engineers comprise less than 0.5% of a recent sample of advertised jobs on LinkedIn (Vu & Oppenlaender, 2025). The specific "new jobs from AI" predictions that circulated widely in 2022–2023 (prompt engineer, AI ethicist as a mass employer) have largely failed to materialise at the predicted scale. This does not mean no new jobs will emerge from AI — it means specific predictions about which jobs are systematically unreliable, and the total volume of net new jobs created is much harder to predict than the jobs being displaced.

09

What Governments Are Doing
Policy Reality vs Policy Need

A landscape of what is actually being tried, what the evidence says about each intervention, and the structural gap between the scale of potential disruption and the scale of policy response.

↑ Evidence of Effectiveness
Sector-Based Retraining (US WorkAdvance, Project QUEST, Year Up)
Harvard/Brown research (Katz, Roth et al.): sector-focused programmes produced earnings gains of 14–38% in the year following training completion, with effects persisting for several years. The key: pairing reskilling with employer relationships, not just classroom instruction. Evidence: Strong
↑ Evidence of Effectiveness
Wage Insurance for Displaced Workers
Supplements income when displaced workers take lower-paying bridge jobs — reducing the incentive to hold out for equivalent pay. Reduces long-term unemployment duration. Recommended by multiple economists as an underused transition tool. Evidence: Moderate
↑ Evidence of Effectiveness
Labour–Capital Tax Rebalancing
Brynjolfsson (MIT) and Acemoglu, Manera & Restrepo: equalising effective tax rates on labour and capital (currently ~30% vs ~8% in US) could reduce automated tasks and raise employment by up to 4%. Politically difficult; economically well-supported. Evidence: Strong (theoretical)
⚖ Contested
Robot / Automation Tax
Gates proposed (2017); Acemoglu et al. modelled optimal rate at 10.15% (raises employment 1.14%, labour share ~1%). MIT economists suggest 1–3.7% to avoid stifling innovation. Key problem: defining "robot" or "AI" in a digitised economy is legally complex. Evidence: Theoretical, no live trials
⚖ Contested
Universal Basic Income (UBI)
Stockton SEED experiment (2019–2021) and OpenResearch (2020–2023) showed cash transfers improve mental health, employment rates, and stability. Systematic reviews confirm poverty reduction. But cost at scale is enormous: a meaningful US UBI (~$1,000/month) requires ~19% consumption tax increase. Evidence: Effective in pilots; scaling fiscally unclear
⚖ Contested
Lifelong Learning Accounts (LLAs)
Portable, individual training accounts funded by government/employer/worker contributions — enabling continuous reskilling without requiring employer sponsorship. Currently proposed but not widely implemented. Singapore's SkillsFuture programme is the most advanced version. Evidence: Promising; limited at scale
↓ Limited Evidence of Effectiveness
Generic Retraining / "Just Reskill" Programmes
"The China Shock" research showed that broad retraining programmes for trade-displaced workers largely failed to produce equivalent employment outcomes. WIOA data (2023–24): <10% on-the-job training, 2% apprenticeships. The model works when tied to specific employer demand; it fails as a general solution. Evidence: Weak in aggregate
↓ Limited Evidence of Effectiveness
AI Moratoriums / Development Freezes
Proposed by some researchers and NGOs. The practical challenge: AI development is globally distributed. A freeze in one jurisdiction drives development to others with weaker safety/labour standards. Economically costly without guaranteed benefit. The EU AI Act takes a regulation-not-moratorium approach, the more evidenced-based path. Evidence: Not supported
The Scale Mismatch
The policy tools that exist are largely designed for the previous wave of automation — manufacturing and trade displacement. The institutional response to AI-driven cognitive work displacement is still being designed. Erik Brynjolfsson (Stanford): "It's astonishing how little seriousness business leaders and policymakers are approaching the coming decade with." The gap between the pace of AI capability development and the pace of institutional policy development is, itself, a major risk factor.
10

What You Can Actually Do
Evidence-Based Individual Actions by Age & Context

Structural problems require structural solutions. But while waiting for policy, individuals can take actions that the evidence supports. Filtered by your life stage and sector.

Before reading this section
No individual-level action can fully compensate for structural displacement. The framing "what can I do" places responsibility on workers for failures that are primarily systemic. This section provides evidence-based personal strategy — not a substitute for demanding structural policy change.
Critical Priority
Develop AI Fluency — Now
Workers using AI tools report 25–66% productivity increases. The adoption gap between AI-fluent and non-fluent workers in the same role is already creating two-tier outcomes. This is not about becoming an AI developer — it is about integrating AI tools into whatever professional work you already do.
Critical Priority
Avoid Pure AI-Automatable Entry Roles if Possible
Goldman Sachs data shows entry-level hiring in AI-exposed roles already compressing by 16–20% in highest-risk occupations. If choosing between two otherwise equal paths, the one requiring physical presence, human judgment, and client relationships is structurally safer for the next 5–10 years.
High Priority
Build "T-Shaped" Skills
Depth in one domain + broad AI and data literacy. David Autor (MIT, 2024 NBER Working Paper): AI presents an opportunity to revitalise middle-skill work by enabling workers with complementary domain knowledge to access tasks previously done only by elite professionals.
High Priority
Build Financial Runway
A 6–12 month emergency fund buys the time needed for the "reskilling window" research identifies as critical (6–18 months). Workers who must take any available job immediately when displaced cannot make optimal transitions. The financial buffer is a career option.
Consider
Geographic Flexibility
AI-complementary roles are concentrated in specific cities and countries. If possible, maintain the option of geographic mobility — the new jobs are not evenly distributed, and the workers who can move to where demand is will have better outcomes than those who cannot.
Consider
Research Ancestral / Second Citizenship
Multiple citizenships broaden labour market options across regulatory environments. Some countries (Portugal, Ireland, Italy) have ancestry-based citizenship pathways. This is a long-term optionality play, not a crisis response.
Critical Priority
Audit Your Role for AI Exposure
Go through your daily tasks honestly. What percentage could be performed by an AI tool today? What percentage requires your physical presence, long-term relationship capital, or accountability structure? The audit itself is the beginning of positioning toward the latter.
Critical Priority
Become the AI-Amplified Version of Your Role
The workers who fare best in AI transitions are those who use AI to become dramatically more productive in their existing role, rather than waiting to be displaced and then retraining. The productivity evidence (14–55% gains) is real — capturing it yourself protects you better than any other single action.
High Priority
Diversify Income Sources
The transition period is the risk period. Having a second income stream — consulting, freelancing, a small business — reduces the cost of an involuntary career change. The research on "reskilling windows" shows the greatest vulnerability comes when people have no financial buffer for transition.
High Priority
Prioritise Relationship Capital
AI cannot replicate your professional reputation, your network, your client trust relationships, or your deep contextual knowledge of a specific organisation's culture. These are compounding assets that become relatively more valuable as AI commoditises generic task execution.
Critical Priority
Document and Position Your Irreplaceable Experience
Decades of contextual, domain, and organisational knowledge has genuine economic value — but only if it is visible. Ensure your specific expertise is clearly articulated to employers and clients. The risk is being seen as a "generic" job title (easily replaced) rather than a specific expert (hard to replace).
Critical Priority
Do Not Wait to Engage with AI Tools
Boston Fed research: optimism about AI rises with education level. Workers who engage with AI tools and integrate them into their practice have significantly better outcomes than those who wait and resist. The learning curve is lower than you think — and the reputational benefit of being "the senior person who leads AI adoption" is real.
High Priority
Financial Portfolio Review
IMF research confirms AI increases returns to capital relative to labour. Workers at 50+ with capital assets (pension, property, investments) should review whether those assets benefit from AI-driven productivity gains. The structural shift from labour to capital income makes capital ownership more important for this age cohort.
High Priority
Identify Bridge Roles Early
If your current role is in the high-risk matrix, identifying and moving toward complementary roles before displacement is far easier than recovering after it. Wage insurance makes accepting a lower-paying bridge role less catastrophic. Planning now is not pessimism — it is the option that the research shows produces the best outcomes.

If your occupation appears in the "Critical Risk" or "High Risk" tables in Section 03 — or if your role is primarily administrative, data-entry, or routine customer service.

Critical Priority
Accelerate Transition Planning
Do not wait for displacement to plan. The research-identified 6–18 month "reskilling window" is much more effective when used proactively than reactively. The workers with the worst outcomes are those who only begin transition planning after losing their job.
Critical Priority
Use AI Tools Immediately in Your Current Role
Counterintuitive but evidence-backed: using AI to become highly productive in your current high-risk role buys time and builds adjacent skills. The most likely first-wave automation is of the routine sub-tasks within your role; humans who manage the AI-assisted version of the role have a longer runway.
High Priority
Identify Sector-Specific Retraining Programmes
Generic community college programmes have weak outcomes. Sector-based programmes with employer partnerships (WorkAdvance model) show 14–38% earnings gains. Research what exists specifically for your occupational category — the quality difference is large.
High Priority
Prioritise Physical-Presence / Relationship Roles
Within your current organisation, actively seek to shift toward tasks that require physical presence, client relationships, or accountability structures. Even in high-risk occupations, these elements are the last to be automated.

If your occupation appears in the "Resilient" category — skilled trades, healthcare, education, complex professional services.

Still Important
Adopt AI as a Productivity Amplifier
Low displacement risk does not mean zero AI impact. AI tools that assist with documentation, research, scheduling, and administrative elements of resilient roles free up human time for the high-value work that is genuinely irreplaceable. Adopters will become significantly more productive than non-adopters — which matters for compensation and career advancement.
Still Important
Recognise the Indirect Effects
Your job may be safe, but your income and working conditions depend on the broader economy. AI-driven wage suppression in adjacent sectors reduces consumer spending power; financial instability among displaced workers affects every service sector. Resilient workers are not insulated from the macroeconomic effects of displacement.
Consider
Capital Accumulation Strategy
IMF modelling confirms AI raises returns to capital relative to labour. As a worker in a resilient labour-income role, deliberately accumulating capital assets (investment portfolio, property, equity stakes) positions you to benefit from AI-driven productivity growth that you would otherwise only experience as a consumer of cheaper goods and services.
SRC

Primary Sources

All factual claims in this report are sourced to specific, verifiable publications. Projections are clearly distinguished from empirical findings.

IMF SDN/2024/001
Gen-AI: Artificial Intelligence and the Future of Work
Cazzaniga et al., IMF Staff Discussion Note, January 2024. Core quantitative framework: 40% global exposure, 60% advanced economies. imf.org
IMF WP/2025/068
AI Adoption and Inequality
Rockall, Tavares, Pizzinelli. IMF Working Paper, April 2025. HELC/HEHC framework, household microdata analysis. imf.org
WEF FoJ 2025
Future of Jobs Report 2025
World Economic Forum. Survey of 1,000+ employers, 14M workers, 55 economies. 92M displaced / 170M created by 2030 projections. weforum.org
GS Research 2025
How Will AI Affect the Global Workforce?
Goldman Sachs Research, August 2025. 22–25 age cohort employment data; 6–7% US displacement estimate; 300M job equivalents global exposure. goldmansachs.com
ILO/NASK 2025
Generative AI and Jobs: A Refined Global Index of Occupational Exposure
International Labour Organization and Poland's National Research Institute. May 2025. 1 in 4 jobs globally exposed; gender differential quantification. ilo.org
ILO 2026
Gen AI, Occupational Segregation and Gender Equality in the World of Work
ILO Research Brief, March 2026. Female-dominated occupations 2x more likely to be exposed (29% vs 16%). ilo.org
Brynjolfsson et al. QJE 2025
Generative AI at Work
Brynjolfsson, Li, Raymond. Quarterly Journal of Economics 140(2), 2025. Gold-standard RCT: 14% productivity gain, largest for lowest-skilled workers. Flagship empirical study.
Stanford DEL 2025
Canaries in the Coal Mine: Six Facts About AI Employment Effects
Brynjolfsson, Horton, Li, Raymond. Stanford Digital Economy Lab Working Paper, 2025. Entry-level employment contraction in AI-exposed roles. digitaleconomy.stanford.edu
Acemoglu & Johnson 2023
Rebalancing AI
Daron Acemoglu and Simon Johnson. IMF Finance & Development, December 2023. "No guarantee AI will generate more jobs than it destroys." imf.org/en/publications/fandd
Brookings 2024
AI's Impact on Income Inequality in the US
Mark Muro, Brookings Institution, July 2024. Klarna case analysis; "hollowing out" of middle-wage jobs; 6M vulnerable workers (86% women). brookings.edu
Brookings/Lightcast 2026
AI Poses Bigger Threat in Jobs with More Women
Brookings/Lightcast analysis, February 2026. 86% of 6M most vulnerable AI-displaced workers are women. CBS News / Brookings.edu
HDSR Fall 2025
Can We Predict What Jobs AI Will Take?
Harvard Data Science Review, Issue 7.4, Fall 2025. Comprehensive methodological review of all major job-risk models; critique of Frey & Osborne. hdsr.mitpress.mit.edu
Deming & Summers 2025
Technical Disruption in the Labor Market (Aspen)
David Deming and Lawrence Summers, Harvard Kennedy School, 2025. 124 years US Census data; occupational churn metric; 2019 inflection point. Harvard Gazette
OECD 2024
Algorithm and Eve: How AI Will Impact Women at Work
OECD Policy Brief, December 2024. OECD AI surveys of employers and workers; gender gap in AI adoption (20 percentage points in Denmark). oecd.org
LSE Media 2025
The Perilous Future of AI Work in the Global South
Media@LSE, November 2025. Geographic continuity argument; India paradox; call centre displacement in Philippines. blogs.lse.ac.uk/medialse
Autor et al. MIT 2024
New Tasks and New Frontiers (NBER Working Paper)
David Autor et al., MIT/NBER, February 2024. AI as opportunity to revitalise middle class; T-shaped skill argument; 60% of today's jobs didn't exist in 1940. nber.org
Frey & Osborne 2013/2017
The Future of Employment
Frey and Osborne, Oxford Martin Institute / Technological Forecasting and Social Change 114 (2017). The seminal 47% US automation risk study. 17,000+ citations. technologyreview.mit.edu
Boston Fed 2025
Workers' Fears and Hopes About AI
Federal Reserve Bank of Boston, December 2025. National survey, December 2024. 21% expect AI to worsen financial situation within 5 years. bostonfed.org