## sdnea2024001

## Source details

**Canonical URL:** [sdnea2024001](https://www.imf.org/-/media/files/publications/sdn/2024/english/sdnea2024001.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/sdn/2024/english/sdnea2024001.pdf.md)
- [Structured JSON version](/-/media/files/publications/sdn/2024/english/sdnea2024001.pdf.json)

---

### Executive summary — AI exposure and cross-country patterns
- Almost 40 percent of global employment is exposed to AI.
- Advanced economies: about 60 percent of jobs are exposed to AI.
- Emerging market economies: 40 percent exposure.
- Low-income countries: 26 percent exposure.
- Advanced economies are both more susceptible to AI-related disruption and better poised to exploit AI benefits; emerging market and developing economies may experience less immediate disruptions but are less ready to seize AI’s advantages, potentially exacerbating the digital divide and cross-country income disparity.

### Complementarity versus substitution (measurement and occupational classification)
- Exposure measured with the AIOE index (Felten, Raj, and Seamans (2021)); complementarity adjusted using Pizzinelli and others (2023) to produce C-AIOE.
- Occupations categorized (thresholds by median values):
  - High exposure, high complementarity (HEHC): examples include surgeons, lawyers, judges.
  - High exposure, low complementarity (HELC): example: telemarketers.
  - Low exposure (LE): examples include dishwashers, performers.
- Key numeric patterns by exposure/complementarity (average employment shares):
  - Advanced economies: 27 percent in HEHC; 33 percent in HELC.
  - Emerging market economies: 16 percent in HEHC; 24 percent in HELC.
  - Low-income countries: 8 percent in HEHC; 18 percent in HELC.
- Heterogeneity ranges by country within groups:
  - Advanced economies:
    - HEHC range: 20.2–37.3 percent.
    - HELC range: 25.9–46.1 percent.
    - LE range: 22.5–53.6 percent.
  - Emerging market economies:
    - HEHC range: 5.7–28.2 percent.
    - HELC range: 10.4–34.7 percent.
    - LE range: 46.1–75.9 percent.
  - Low-income countries:
    - HEHC range: 2–35.3 percent.
    - HELC range: 1.4–33 percent.
    - LE range: 54–96.1 percent.
- Selected country-level observations:
  - UK: almost 70 percent of employment in high-exposure occupations (approximately equally distributed between high- and low-complementarity).
  - US: almost 60 percent of employment in high-exposure occupations (approximately equally distributed between high- and low-complementarity).
  - Brazil: high-exposure employment 41 percent.
  - India: high-exposure employment 26 percent.

### Inequality and distributional impacts
- Distinctive feature of AI: displacement risks extend to higher-wage earners, unlike past automation that hit middle-skilled workers most.
- Potential complementarity is positively correlated with income.
- Model simulations highlight distributional channels:
  - High complementarity can produce a more-than-proportional increase in higher-wage earners’ labor income, raising labor income inequality.
  - Enhanced capital returns from AI amplify increases in income and wealth inequality accruing to high earners.
- Empirical pattern in the UK:
  - Workers’ exposure to AI increases with income; potential complementarity also increases with income but peaks around the 75th percentile and declines slightly thereafter.

### Productivity and aggregate outcomes (model calibration and scenarios)
- Model calibration baseline: change in the capital share of 5.5 percentage points (based on UK change between 1980 and 2014).
- Scenario set (common features: same displacement via capital deepening; differ in complementarity and productivity):
  1. Low-complementarity:
     - Displacement dominates; labor income falls at the top.
     - Aggregate output increases by almost 10 percent (combination of capital deepening and a small increase in total factor productivity).
  2. High-complementarity:
     - AI strongly complements human labor in certain occupations.
     - Sectoral reallocation moves labor demand from low- to high-complementarity occupations.
     - Total income levels of low-income workers decline by 2 percent; gains at the top are almost 8 percent.
     - Aggregate output increase approximately similar to Scenario 1.
  3. High-complementarity and high-productivity:
     - Productivity calibrated to generate close to a 1.5 percentage point increase in workers’ average annual productivity growth rate in the first 10 years after AI adoption.
     - Output increases by 16 percent between steady states.
     - Total factor productivity (TFP) increases by almost 4 percent.
     - Total income rises for all workers: from 2 percent for low-income workers to almost 14 percent for high-income workers.
- Capital income / wealth effects:
  - Capital income and wealth inequality always increase with AI adoption across scenarios.
  - Interest rates increase by almost 0.4 percentage point in all scenarios.
  - High-income workers benefit more because they hold a large share of assets.

### Worker reallocation and heterogeneity (historical microdata evidence)
- Microdata analysis: rotating-panel labor force surveys for Brazil and the UK; occupation-switching probabilities (average yearly):
  - Brazil, college-educated: 43.7 percent.
  - United Kingdom, college-educated: 29.8 percent.
  - Brazil, non-college-educated: 38 percent.
  - United Kingdom, non-college-educated: 27 percent.
- Key behavioral patterns:
  - Workers tend to switch between similar occupations, but a significant fraction switch across occupations with different AI exposure/complementarity.
  - College-educated workers historically move more easily into high-complementarity jobs; non-college-educated workers are less likely to move into high-complementarity roles.
  - More than a third of those moving away from low-complementarity jobs shift toward roles with higher AI complementarity.
- Age dynamics:
  - College-educated workers typically transition from low- to high-complementarity jobs in their 20s and 30s; progression stabilizes by late 30s to early 50s.
  - Older workers face lower one-year reemployment probabilities and additional barriers (obsolete skills, mobility constraints, shorter career horizon).
- Wage implications:
  - Progressing to HEHC occupations is associated with higher wages in both Brazil and the UK.
  - In Brazil, switching from high-exposure to low-exposure occupations tends to contract hourly wages.

### AI Preparedness Index (AIPI) — structure and coverage
- Four aggregate dimensions:
  1. Digital infrastructure.
  2. Innovation and economic integration.
  3. Human capital and labor market policies.
  4. Regulation and ethics.
- Foundational vs second-generation preparedness:
  - Foundational: digital infrastructure and human capital.
  - Second-generation: innovation/economic integration and regulation/ethics.
- Coverage and normalization:
  - Index coverage: 32 advanced economies; 56 emerging market economies; 37 low-income countries.
  - Subindicators normalized to a 0–1 scale; each aggregate dimension is the simple average of its normalized subcomponents; the AIPI is the simple average of the four aggregate dimensions.
- Robustness:
  - PCA-based approach produced results indistinguishable from simple averaging.
- Cross-country implications:
  - Wealthier economies generally score higher on preparedness and are both more exposed to AI and better positioned to harness it.
  - Low-income countries are less exposed but underprepared, especially in digital infrastructure and digitally skilled labor force.

### Policy priorities and sequencing (high-level recommendations)
- For advanced and more developed emerging market economies:
  - Invest in AI innovation and integration.
  - Advance regulatory frameworks to optimize benefits from increased AI use.
- For less prepared emerging market and developing economies:
  - Prioritize foundational infrastructure development.
  - Build a digitally skilled labor force.
  - Given limited fiscal space, focus on high-return projects that build foundational capacity.
- For all economies:
  - Strengthen social safety nets.
  - Provide retraining for AI-susceptible workers.
  - Define AI property rights and calibrate redistributive fiscal policies to shape distributional outcomes.
  - Pursue international cooperation on ethical standards and data security.

### Key empirical and historical numeric highlights (preserved)
- Almost 40 percent of global employment exposed to AI.
- Advanced economies: about 60 percent exposure.
- Emerging market economies: 40 percent exposure.
- Low-income countries: 26 percent exposure.
- Advanced-economy employment shares: 27 percent HEHC; 33 percent HELC.
- Calibrated change in the capital share used in model scenarios: 5.5 percentage points.
- Productivity calibration: close to a 1.5 percentage point increase in workers’ average annual productivity growth rate in the first 10 years (Scenario 3).
- Output increases by almost 10 percent in low-complementarity scenario; by 16 percent in high-complementarity and high-productivity scenario.
- Interest rates increase by almost 0.4 percentage point in all scenarios.
- Historically, only about 20 percent of workers moving from high-exposure to low-exposure occupations also entered the informal sector.

*Source: Staff Discussion Notes — "Gen-AI: Artificial Intelligence and the Future of Work" (IMF), sdnea2024001 (selected excerpts).*

### Executive Summary __________________________________________________________________ 2

### Executive Summary

### AI exposure and cross-country patterns
- Almost 40 percent of global employment is exposed to AI.
- In advanced economies, about 60 percent of jobs are exposed to AI.
- Overall exposure is 40 percent in emerging market economies and 26 percent in low-income countries.
- Advanced economies are both more susceptible to AI-related disruption and better poised to exploit AI benefits; emerging market and developing economies may experience less immediate disruptions but are less ready to seize AI’s advantages, which could exacerbate the digital divide and cross-country income disparity.

### Complementarity versus substitution
- A new measure of potential AI complementarity suggests that, of the jobs exposed to AI in advanced economies, about half may be negatively affected by AI, while the rest could benefit from enhanced productivity through AI integration.
- The note attempts to assess potential complementarity and substitution using an approach that considers the social, ethical, and physical context of occupations and required skill levels.

### Inequality and distributional impacts
- AI will affect income and wealth inequality.
- Unlike previous waves of automation that affected middle-skilled workers most, AI displacement risks extend to higher-wage earners.
- Potential AI complementarity is positively correlated with income.
- Model simulations suggest that, with high complementarity, higher-wage earners can expect a more-than-proportional increase in their labor income, leading to an increase in labor income inequality.
- Enhanced capital returns from AI can amplify increases in income and wealth inequality that accrue to high earners.
- Countries’ choices regarding the definition of AI property rights, as well as redistributive and other fiscal policies, will ultimately shape AI’s impact on income and wealth distribution.

### Productivity and aggregate outcomes
- The gains in productivity, if strong, could result in higher growth and higher incomes for most workers.
- Owing to capital deepening and a productivity surge, AI adoption is expected to boost total income.
- If AI strongly complements human labor in certain occupations and the productivity gains are sufficiently large, higher growth and labor demand could more than compensate for the partial replacement of labor tasks by AI, and incomes could increase along most of the income distribution.

### Worker reallocation and heterogeneity
- College-educated workers are better prepared to move from jobs at risk of displacement to high-complementarity jobs; older workers may be more vulnerable to the AI-driven transformation.
- Evidence from the UK and Brazil shows that college-educated individuals historically moved more easily from jobs now assessed to have high displacement potential to those with high complementarity, while workers without postsecondary education show reduced mobility.
- Younger workers who are adaptable and familiar with new technologies may be better able to leverage new opportunities; older workers may struggle with reemployment, adapting to technology, mobility, and training for new job skills.

### Policy priorities and AI preparedness
- A novel AI preparedness index indicates priorities differ by development level:
  - Advanced and more developed emerging market economies should invest in AI innovation and integration, while advancing adequate regulatory frameworks to optimize benefits from increased AI use.
  - Less prepared emerging market and developing economies should prioritize foundational infrastructural development and building a digitally skilled labor force.
- For all economies, social safety nets and retraining for AI-susceptible workers are crucial to ensure inclusivity.

---

### Introduction and analytical approach

### Framing and motivation
- AI promises to boost productivity and growth but its impact is uncertain and may vary by job roles and sectors, with the potential to amplify disparities.
- AI represents a wide spectrum of technologies designed to enable machines to perceive, interpret, act, and learn to emulate human cognitive abilities; generative AI (GenAI) includes systems such as sophisticated large language models that can create new content, while other AI models are more specialized.
- The field is rapidly evolving, especially with GenAI, expanding AI’s potential applications and the set of job functions it can affect.

### Research questions
The note aims to answer six questions:
1. Which countries are more exposed to AI adoption? Which countries are likely to benefit most?
2. How differently will AI affect workers within countries? Which segments of workers are likely to thrive and which face more risks?
3. Historically, how frequently did workers shift between roles now facing varying AI exposure? What insights do these shifts reveal about labor adaptability?
4. In what ways could AI reshape income and wealth inequality?
5. What is the potential impact for growth and productivity?
6. Which countries appear better prepared for the AI transition? How can policies maximize gains and mitigate likely AI-related challenges?

### Contributions relative to existing literature
- Distinguishes potential AI complementarity from substitutability using the approach developed by Pizzinelli and others (2023), adding contextual occupational features beyond pure task-based metrics.
- Offers initial insight into workers’ historical transitions between occupations with differing AI exposure, using microdata for one advanced and one emerging market economy.
- Analyzes how AI may affect income and wealth inequality within countries, combining exposure patterns and a model-based analysis of labor and capital income inequality and income levels.
- Examines AI preparedness across a large sample of advanced and emerging market and developing economies using a novel AI Preparedness Index.

### Important caveats
- Model analysis assumes activity grows in occupations with high AI complementarity and falls in low-complementarity occupations, but the AI exposure analysis holds sector sizes fixed and tasks in each occupation unchanged, making the results more pertinent for the short to medium term.
- Over longer horizons, workers may migrate across sectors, acquire new skills, and jobs will evolve; the analysis assumes homogenous task composition within occupations across countries, though significant cross-country variation can exist.
- The approach abstracts from linkages across occupations and countries (trade linkages) and from cross-border spillovers of AI exposure.
- Model-based results depend on calibration assumptions and the uncertain pace of AI adoption, investment in physical capital, reorganization needed to capitalize on AI, and timing of aggregate macroeconomic effects.

---

### Key findings, scenarios, and policy considerations

### Key empirical findings
- Employment exposure estimates:
  - Almost 40 percent of global employment exposed to AI.
  - About 60 percent exposure in advanced economies.
  - 40 percent exposure in emerging market economies.
  - 26 percent exposure in low-income countries.
- Complementarity: Of exposed jobs in advanced economies, about half may be negatively affected while the rest could benefit via productivity enhancement.

### Model scenarios and distributional outcomes
- High-complementarity scenario:
  - AI strongly complements human labor in certain occupations.
  - Productivity surge and capital deepening boost total income.
  - Higher-wage earners can see a more-than-proportional increase in labor income, widening labor income inequality.
  - Enhanced capital returns accrue to high earners, further amplifying income and wealth inequality.
- Low-complementarity / high-displacement scenario:
  - Partial replacement of labor tasks by AI reduces labor demand in affected occupations.
  - Without offsetting productivity-driven demand increases, incomes could stagnate or decline for displaced workers, raising inequality.

### Policy recommendations (high-level)
- For advanced and more developed emerging market economies:
  - Invest in AI innovation and integration.
  - Advance regulatory frameworks to optimize benefits from increased AI use.
- For less prepared emerging market and developing economies:
  - Prioritize foundational infrastructure development.
  - Build a digitally skilled labor force.
- For all economies:
  - Strengthen social safety nets.
  - Provide retraining for AI-susceptible workers.
  - Define AI property rights and calibrate redistributive fiscal policies to shape distributional outcomes.

---

*Source: Executive Summary and Introduction of "Staff Discussion Notes Gen-AI: Artificial Intelligence and the Future of Work" (IMF).*

### introductions of general-purpose technologies, such as electricity. This uncertainty applies also to the results of

### AI Exposure and Complementarity (SDN EA 2024/001)

### II.1 Conceptual framework
- Framework refines task-based occupational approach to measure:
  - "Exposure" to AI (degree of overlap between AI applications and required human abilities), based on Felten, Raj, and Seamans (2021).
  - "Complementarity" potential (degree of shielding from displacement and likelihood of AI complementing human work), augmenting with Pizzinelli and others’ (2023) index that leverages social, ethical, and physical context and required skill levels.
- Occupations categorized into three groups (thresholds represented by median values):
  - High exposure, high complementarity (HEHC): significant AI support potential, limited scope for unsupervised AI; examples: surgeons, lawyers, judges.
  - High exposure, low complementarity (HELC): high likelihood AI could replace human tasks; example: telemarketers.
  - Low exposure (LE): minimal or no potential for AI application; examples: dishwashers, performers.
- Caveats and limitations noted:
  - Measures are relative across occupations.
  - High complementarity can still yield displacement for workers lacking AI-related skills or in firms that do not invest.
  - Static view: does not incorporate IT infrastructure availability, workers’ ability to acquire skills or relocate, ongoing integration of AI and robotics, societal preference/regulatory changes, adoption speed, adoption costs, or macro feedback effects from productivity gains.
- Data coverage and classification:
  - Definitions applied to 142 countries using the online International Labour Organization (ILO) employment database.
  - Main analysis uses 72 sub-major occupation groups (2-digit ISCO)-08.
  - Microdata analysis: 130 minor groups (3-digit) for India and 436 unit groups (4-digit) for the other five countries examined.

### II.2 Cross-country differences (key findings and statistics)
- Global and country-group shares:
  - About 40 percent of workers worldwide are in high-exposure occupations.
  - Share in advanced economies: 60 percent of workers in high-exposure occupations.
- Average employment shares by exposure and complementarity (advanced economies, emerging market economies, low-income countries):
  - Advanced economies: 27 percent in high-exposure, high-complementarity occupations; 33 percent in high-exposure, low-complementarity occupations.
  - Emerging market economies: 16 percent in high-exposure, high-complementarity occupations; 24 percent in high-exposure, low-complementarity occupations.
  - Low-income countries: 8 percent in high-exposure, high-complementarity occupations; 18 percent in high-exposure, low-complementarity occupations.
- Selected country-level observations:
  - UK: almost 70 percent of employment is in high-exposure occupations (approximately equally distributed between high- and low-complementarity).
  - US: almost 60 percent of employment is in high-exposure occupations (approximately equally distributed between high- and low-complementarity).
  - High-exposure employment in emerging market economies: 41 percent in Brazil; 26 percent in India.
- Heterogeneity ranges (by country within groups):
  - Advanced economies:
    - HEHCs range between 20.2 and 37.3 percent.
    - HELCs range between 25.9 and 46.1 percent.
    - LEs range between 22.5 and 53.6 percent.
  - Emerging market economies:
    - HEHCs range between 5.7–28.2 percent.
    - HELCs range between 10.4–34.7 percent.
    - LEs range between 46.1–75.9 percent.
  - Low-income countries:
    - HEHCs range between 2–35.3 percent.
    - HELCs range between 1.4–33 percent.
    - LEs range between 54–96.1 percent.
- Driving factor:
  - Differences in employment composition by broad occupational groups (economic structure) explain most cross-country variation (examples: UK—large professional and managerial shares; India—large shares of craftspeople, skilled agricultural, and elementary workers).

### II.3 Within-country differences (distributional implications)
- Gender:
  - Women tend to be employed in high-exposure occupations more than men.
  - For women the high-exposure share is distributed approximately equally between high- and low-complementarity jobs—implying both greater risks and greater opportunities.
  - Exceptions: countries with high shares of women in agricultural jobs (for example, India) may deviate from this pattern.
- Education:
  - Higher education levels are associated with greater shares of employment in high-exposure occupations across all examined countries.
  - This association is especially pronounced for occupations with high complementarity.
  - Implication: AI could more strongly affect high-skilled workers relative to past automation waves, but complementarity may offset displacement for those with the relevant skills.
- Age:
  - No common pattern across countries because cohort composition (gender and education) differs markedly by country.
  - UK and US: younger groups contain more college-educated individuals due to increased university attendance over past 30 years.
  - Emerging market economies and low-income countries: younger groups have more women due to recent rises in female labor participation.
- Income distribution:
  - Share of employment in high-exposure, low-complementarity jobs (at risk of displacement) is broadly similar across income quantiles (with a mildly positive slope in emerging market economies).
  - Employment in high-exposure, high-complementarity jobs is more concentrated in upper-income quantiles.
  - Correlation between earnings and potential complementarity:
    - Stronger in emerging market economies (complementarity rises at the top of the distribution).
    - More muted in some advanced economies (for example, in the UK complementarity plateaus at the top).
  - Implication: AI gains likely to disproportionately accrue to higher-income earners, especially in countries such as India and, to a lesser extent, the US.

*Source: STAFF DISCUSSION NOTES — Gen-AI: Artificial Intelligence and the Future of Work (International Monetary Fund), excerpt covering Sections II–II.3.*

### 1. High-Exposure, Low-

### 1. High-Exposure, Low-

### III. Worker Reallocation in the AI-Induced Transformation
- Long-term adjustment: workers will adjust to changing skill demands and sector shifts, with some transitioning to high-AI-complementarity roles and some struggling to adapt.
- Distinction emphasized between jobs and workers: AI adoption may destroy some jobs and create or enhance others, but incumbents may not be the ones who reap benefits.
- Employment effects depend on worker characteristics (age, education), affecting adaptability; historical data suggest some workers may struggle to adapt.

### Historical job transition patterns (microdata evidence)
- Analysis uses rotating-panel labor force surveys for Brazil and the UK to track worker occupation transitions.
- Key empirical observations:
  - Workers tend to switch between similar occupations, indicating potentially limited flexibility, but a significant fraction switch across occupations with different AI exposure/complementarity.
  - College-educated workers have historically shown greater ability to transition into jobs with high AI-complementarity potential.
  - Average yearly occupation-switching probabilities:
    - Brazil, college-educated: 43.7 percent
    - United Kingdom, college-educated: 29.8 percent
    - Brazil, non-college-educated: 38 percent
    - United Kingdom, non-college-educated: 27 percent
  - College-educated individuals in AI-intensive or potentially AI-intensive jobs tend to stay within such environments when switching jobs.
  - More than a third of those moving away from low-complementarity jobs shift toward roles with higher AI complementarity.
  - Non-college-educated workers are predominantly in low-AI-exposure jobs and are less inclined to move to high-complementarity positions when switching from high-exposure, low-complementarity occupations.

### Age and education effects; life-cycle dynamics
- College-educated workers often transition from low- to high-complementarity jobs in their 20s and 30s; career progression stabilizes by late 30s to early 50s.
- Non-college-educated workers show similar patterns but less pronounced and occupy fewer high-exposure positions.
- Implications for young college-educated workers:
  - Vulnerable to disruption but more adaptable; could benefit from faster experience accumulation using AI.
  - If low-complementarity positions serve as stepping stones, reductions in such roles could hinder market entry for young high-skilled workers.
  - Generative AI and easier-to-use tools may amplify productivity gains especially for less experienced and low-skilled workers (citing study finding greatest productivity impact for less experienced and low-skilled customer support workers).
- Older workers face additional barriers:
  - Lower likelihood of reemployment within a year after termination compared with young and prime-age workers.
  - Factors: obsolete skills, geographic/emotional ties, financial obligations, reluctance or perceived barriers to transition, shorter career horizon reducing firm incentives to train.
  - Pension and unemployment insurance generosity can amplify reemployment frictions for older workers.

### Reemployment probabilities and mobility by age
- One-year reemployment probability studies (Brazil and UK) show:
  - After unemployment, older workers previously in high-exposure, high-complementarity occupations are less likely to find jobs in the same category than prime-age workers.
  - Differences reflect technological change, preference shifts, and age-related biases or hiring stereotypes.
- Technology’s role in earnings losses after unemployment:
  - Braxton and Taska (2023) find technology contributes 45 percent of earnings losses following unemployment.

### Occupational switches and wage effects
- Progressing to high-exposure, high-complementarity occupations is associated with higher wages in both Brazil and the UK.
- In Brazil, switching to low-exposure from high-exposure occupations tends to produce contraction in hourly wages (income losses).
- Occupational mobility as driver of wage dynamics:
  - Literature: occupational mobility is important for wage growth and wage inequality (Kambourov and Manovskii 2009 and others).
  - Prior estimates of occupation switching:
    - US yearly switching rate estimates: 21 percent (Kambourov and Manovskii 2009); 3.5 percent monthly equivalent to 34.7 percent annually (Moscarini and Vella 2008).
    - Brazil: 30 percent switching over a period of four months (Monsueto, Moreira Cunha, and da Silva Bichara 2014).

### Summary of reallocation section
- Young college-educated workers: most vulnerable yet most adaptable; frequently move between job types.
- High-exposure, high-complementarity roles offer wage premiums; switching to low-exposure roles can lower wages.
- Workers of all ages tend to return to similar roles after unemployment, suggesting some labor market inflexibility.
- Historical patterns are informative but structural transformation from AI adoption remains uncertain.

---

### IV. AI, Productivity, and Inequality (model-based analysis)
- Analytical approach: task-based model (described in Rockall, Pizzinelli, and Tavares (forthcoming); builds on Drozd, Taschereau-Dumouchel, and Tavares (2022) and Moll, Rachel, and Restrepo (2022)).
- Three critical channels through which AI may affect the economy:
  1. Labor displacement: tasks shift from labor to AI capital, reducing labor income.
  2. Complementarity: AI increases importance of tasks not displaced, boosting demand for occupations with high AI complementarity.
  3. Productivity gains: AI may generate broad-based productivity increases, boosting investment and overall labor demand, offsetting some labor-income decline.
- Capital income effects: AI adoption leads to increases in the return on capital, raising capital income and wealth inequality consistent with initial asset distributions.
- Model calibration: United Kingdom (a country highly exposed to AI adoption).
- Income decomposition in the model:
  1. Labor income (can be positively or negatively affected by AI depending on complementarity).
  2. Capital income (increases with AI adoption).
  3. Benefits and other income (government benefits, pensions, etc.).
- Empirical patterns in the UK:
  - High-income workers have a much larger share of capital income than middle- and low-income workers.
  - Middle- and low-income workers’ total income depends more on labor income.
  - Workers’ exposure to AI increases with income; potential complementarity also increases with income but, in the UK, peaks around the 75th percentile and declines slightly thereafter.

### Simulation scenarios and calibration details
- Three scenarios assume a labor share decline consistent with historical automation episodes; baseline calibration draws on UK change between 1980 and 2014:
  - Assumed labor share decline: 5.5 percentage points following AI introduction.
- Scenario differentiation (all embed same displacement via capital deepening, but differ in complementarity and productivity):
  1. Low-complementarity: AI only mildly increases demand for high-complementarity occupations.
  2. High-complementarity: AI strongly supports demand for high-complementarity occupations.
  3. High-complementarity and high productivity: same as (2) plus economy-wide productivity augmentation concentrated among workers in high-complementarity occupations.
- Productivity increase calibration in scenario (3):
  - Calibrated to generate close to a 1.5 percentage point increase in workers’ average annual productivity growth rate in the first 10 years after AI adoption.
  - This 1.5 percentage point figure is at the lower end of firm-level studies estimating AI’s potential impact on workers’ productivity (Briggs and Kodnani 2023).

*Sources: American Community Survey; Gran Encuesta Integrada de Hogares; India Periodic Labour Force Survey; Labour Market Dynamics in South Africa; Pesquisa Nacional por Amostra de Domicílios Contínua; Pizzinelli and others (2023); UK Labour Force Survey; and IMF staff calculations.*

### 1. Exposure of Income to AI

### 1. Exposure of Income to AI

### Measurement and data
- AI exposure measure: share of total hours worked in a job in the top 30 percent of AI Occupational Exposure scores (Felten, Raj, and Seamans (2021)), weighted by hours worked; threshold chosen for comparability with historical automation episodes.
- AI complementarity measure: based on work contexts and skills as discussed in Box 1 and Pizzinelli and others (2023).
- Income categories (Panel 1): (1) wage income; (2) benefits, pensions, and other income; (3) capital income (rents and estimated investment income).
- Data sources: UK Office for National Statistics, Wealth and Assets Survey; and IMF staff calculations.

### Key mechanisms linking AI to income distribution
- Net effect depends on the race between:
  - Degree of exposure to AI, and
  - Degree of complementarity between AI and labor,
  - And AI’s boost to productivity.
- Displacement effect (low complementarity): AI adoption can reduce labor income at the top of the distribution when complementarity gains are small relative to displacement.
- Complementarity effect (high complementarity): AI can increase earnings for workers in occupations with high complementarity, especially in the upper half of the income distribution.
- Productivity channel: When AI raises productivity, labor income rises for all workers because higher productivity increases demand for all factors of production; however, the increase is larger for workers with high AI complementarity, raising labor income inequality.

### Model scenarios and distributional outcomes (Figure 10)
- Common calibration: change in the capital share is 5.5 percentage points (based on 1980–2014 change).
- Scenario 1 — Low Complementarity:
  - Displacement dominates; labor income falls at the top.
  - Aggregate output increases by almost 10 percent (combination of capital deepening and a small increase in total factor productivity).
- Scenario 2 — High Complementarity:
  - Complementarity reduces the share of high-income workers negatively affected: the share at the top drops from almost 15 percent to less than 5 percent.
  - Sectoral reallocation moves labor demand from low- to high-complementarity occupations.
  - Total income levels of low-income workers decline by 2 percent; gains at the top are almost 8 percent.
  - Overall increase in national income approximately similar to Scenario 1, but labor income inequality increases.
- Scenario 3 — High Complementarity and High Productivity:
  - Output increases by 16 percent between steady states.
  - Total factor productivity (TFP) increases by almost 4 percent.
  - Despite increased labor income inequality, total income rises for all workers: from 2 percent for low-income workers to almost 14 percent for high-income workers.
  - These gains occur primarily in the first 10 years of the transition.
- Capital income and wealth inequality:
  - Always increase with AI adoption across scenarios.
  - Mechanism: labor displacement plus increased demand for AI capital → higher capital returns and asset values.
  - Interest rates increase by almost 0.4 percentage point in all scenarios (potential to partially offset the decline in the natural rate of interest in the UK and advanced economies in general).
  - High-income workers benefit more because they hold a large share of assets.

### Aggregate implications (Figure 11)
- Low-complementarity scenario: output ≈ almost 10 percent increase; small TFP increase.
- High-complementarity scenario: similar aggregate output and TFP impact as low complementarity when capital deepening is the same, but with redistributive sectoral shifts.
- High-complementarity plus high-productivity scenario: output +16 percent; TFP +almost 4 percent; distributional gains skewed to top earners.

### Implications for emerging market and developing economies
- Two relevant aspects: (1) higher initial income and wealth inequality, and (2) lower exposure to AI.
- Simulations indicate:
  - Higher initial inequality could exacerbate wealth disparity because AI gains accrue predominantly to top earners.
  - Labor income inequality could decrease more if AI-exposed workers are concentrated at the top.
  - In economies with fewer AI-exposed workers, direct distributional impacts may be less pronounced.
- AI can potentially boost inclusion by enhancing public services and sectors like agriculture and health care, but these aspects are outside the model scope.
- Cross-border effects:
  - Potential reshoring to advanced economies could reallocate capital and labor away from less prepared regions, reducing equilibrium interest rates in advanced economies and exerting downward pressure on capital income in origin countries.
  - Alternatively, with sufficient investment, AI could allow emerging economies to leapfrog in certain sectors and reduce cross-country inequality.
- Caveat: model abstracts from changes in property rights definitions and fiscal/redistributive policy responses that could reshape outcomes.

### AI Preparedness overview (AIPI) and implications
- Proposed AI Preparedness Index (AIPI) organized under four categories:
  1. Digital infrastructure
  2. Innovation and economic integration
  3. Human capital and labor market policies
  4. Regulation and ethics
- Preparedness is cumulative: foundational preparedness (digital infrastructure and human capital) enables adoption; second-generation preparedness (innovation and legal frameworks) supports diffusion, trust, and scaling.
- Cross-country patterns:
  - Wealthier economies (advanced and some emerging market economies) generally score higher on preparedness and are both more exposed to AI and better positioned to harness it.
  - Low-income countries are relatively less exposed but underprepared across all dimensions, especially digital infrastructure and digitally skilled labor force.
- Correlations (Figure 13):
  - In advanced economies, regulatory frameworks and innovation/integration are more strongly correlated with the size of the digital sector.
  - In low-income countries, digital infrastructure and human capital are strongly associated with the digital sector size.
- Policy sequencing:
  - For economies with high exposure and strong foundational preparedness: prioritize strengthening digital innovation capacity and adapting legal and ethical frameworks.
  - For economies with weak foundational preparedness: prioritize investment in digital infrastructure and human capital before emphasizing second-generation preparedness.
  - Given limited fiscal space in many low-income countries, focus spending on high-return projects that build foundational capacity.

### Distributional and labor-market considerations for policy
- Exposure patterns by demographic:
  - Women and highly educated workers: consistently more exposed to AI but also more likely to benefit from complementarity.
  - Older workers: more likely to struggle in reemployment and skill acquisition.
- Distinctive feature of AI relative to past automation:
  - Displacement risks span the entire income spectrum, including high-income earners and skilled professionals, whereas past automation hit middle-skilled jobs more.
  - Complementarity with high-income professionals can amplify labor income inequality if complementarity is strong.
- Policy imperatives:
  - Strengthen digital infrastructure and human capital where foundational gaps exist.
  - Adapt regulatory and ethical frameworks to build trust and govern AI diffusion.
  - Protect and retrain workers at risk of displacement; promote labor reallocation policies and social safety nets.
  - Pursue international cooperation on ethical standards and data security given the cross-border nature of AI.

*Source: IMF staff discussion notes — sdnea2024001: "1. Exposure of Income to AI".*

### Box 1. Artificial Intelligence Occupational Exposure and Potential

### Box 1. Artificial Intelligence Occupational Exposure and Potential Complementarity

### Definitions and measurement of AI occupational exposure
- The most common measure is the AI Occupational Exposure (AIOE) index of Felten, Raj, and Seamans (2021), measuring the correspondence between 10 AI applications and 52 human skills.
- The overlap between AI and human abilities is weighted by the degree of importance and complexity of such skills in each job.
- The AIOE index is interpreted in relative terms and reported as normalized or rescaled between 0 and 1.
- The AIOE measure is agnostic about whether exposure implies complementarity or substitution; it focuses on the relative likelihood of AI’s integration into the functions of a given job.

### Complementarity and the complementarity-adjusted measure (C-AIOE)
- Pizzinelli and others (2023) propose a potential complementarity index to adjust the original AIOE measure.
- Greater potential complementarity reduces exposure; hence, a higher complementarity-adjusted AIOE (C-AIOE) more explicitly reflects a higher chance of labor substitution.
- The complementarity index draws from O*NET work contexts and skills, using case-by-case judgment on contexts where societies may be less likely to allow unsupervised use of AI (for example, the criticality of decisions and gravity of consequences).
- Illustrative examples: judges and doctors, despite high AI exposure, would still likely be human beings due to decision criticality and consequences.

### Conceptual framework and implications
- Two dimensions: exposure (x-axis) defines the scope for applying AI to carry out the main functions of a job; complementarity (θ) determines, given exposure, whether AI is more likely to substitute for or complement human labor.
- Outcomes by quadrant:
  - High exposure + low complementarity: relatively higher likelihood of AI replacing key tasks; in acute cases, a decrease in demand for the occupation, leading to reduced employment prospects, lower wages, and higher risk of displacement.
  - High exposure + high complementarity: greater likelihood of workers experiencing productivity growth and wage gains from adopting AI-driven technologies, contingent on possessing skills needed to use AI. Without such skills, workers may face lower compensation and reduced employment prospects.
  - Low exposure (regardless of complementarity): complementarity is less relevant because tasks likely affected by AI are less integral to the occupation.
- Box Figure 1.1: conceptual diagram showing AIOE on x-axis and complementarity (θ) on y-axis; red reference lines denote the median of AIOE and complementarity.

### Formal adjustment formula
- Given potential complementarity, θ, the complementarity-adjusted AI occupational exposure (C-AIOE) is constructed as:
  - C-AIOE = AIOE *(1– θ –  θMIN))
- The adjustment lowers exposure for occupations with higher values of θ relative to the occupation with the lowest complementarity (θMIN).

### Key takeaways
- AIOE quantifies the scope for AI to perform job functions but does not by itself determine whether AI will be a substitute or complement.
- Incorporating a complementarity dimension (θ) alters vulnerability rankings: professions and managers see lower average exposure after adjustment; clerical occupations show relatively high complementarity-adjusted exposure and vulnerability to disruption.
- Complementarity is influenced by societal and technical concerns (for example, decision criticality) and varies substantially within occupational groups as well as across them.
- The benefits of AI adoption for workers in high-exposure, high-complementarity jobs are conditional on acquiring the relevant skills to use AI effectively.

*This box was prepared by Carlo Pizzinelli.*

### Annex Table 3.1. Quarterly Transition Probabilities across Occupation Types and Labor Market

### Annex Table 3.1. Quarterly Transition Probabilities across Occupation Types and Labor Market Statuses for Brazil and the United Kingdom

### Table scope and definition
- Each cell reports the percentage of workers who transition from the occupation or labor market status listed in the respective row to that listed in the respective column between two quarters.
- Each row adds up to 100 percent; that is, the totality of workers in the occupation or labor market status listed in the respective row in the first quarter.
- Definitions preserved from source:
  - U2N = a transition from unemployment to inactivity.
  - HE2LE = a worker employed in an occupation code with AI exposure above the median in the current quarter but employed in an occupation code with AI exposure below the median in the previous quarter.
- Data sources: Pesquisa Nacional por Amostra de Domicílios Contínua; UK Labour Force Survey; and IMF staff calculations.

### Key empirical findings (panel summaries and notes)
- Brazil and United Kingdom quarterly transition probabilities are presented in percent (table cells not reproduced here).
- Note on figure interpretation: Transition probabilities represent the average share of workers in the “from” category who move to the category listed on the column in the following quarter.
- For Brazil, Annex Figure 3.1 panels summarize:
  - Panel 1: Share of employment in total employment by formality and exposure category.
  - Panel 2: Transition probabilities for formal workers moving to a low-exposure occupation. “From” indicates the exposure category of the person’s occupation in the preceding quarter.
  - Blue bars represent probability of a formal worker moving to a formal job; orange bars represent probability of a formal worker moving to an informal job.
  - LE = low exposure.

### AI and informality — empirical conclusions
- In many emerging market and developing economies, despite high labor informality, AI-induced labor reallocation is unlikely to affect the size of the formal labor force significantly because growth in high-exposure, high-complementarity occupations will likely be in the formal sector.
- Workers displaced from high-exposure, low-complementarity occupations may face job loss and move to informality.
- Evidence from Brazil indicates a limited risk of such a double blow (Annex Figure 3.1):
  - A large share of employment in low-exposure occupations is in formal work arrangements (panel 1).
  - Most occupational switches of formal workers have not involved movement into the informal sector (panel 2).
- Historic statistic: "Historically, only about 20 percent of workers moving from high-exposure to low-exposure occupations also entered the informal sector."

### Annex 4 — Model overview (main features)
- Time is continuous.
- Final consumption good produced using intermediate goods obtained from a continuum of tasks aggregated according to a Cobb-Douglas production function.
- Tasks can be produced using labor or capital; capital is assumed more productive than labor for tasks taken over by AI (displacement is productivity-enhancing).
- Agents are heterogeneous in skills and ability to invest in capital markets; provide inelastic labor supply across sectors and face dissipation shocks.
- Different sectors pay different wages; bond investors receive risk-free rate, capital market investors receive return on capital.
- Agents maximize standard preferences over utility flows from consumption subject to a budget constraint and a natural debt limit.
- Heterogeneity in skill types and investment replicates income and wealth inequality.
- Three channels through which AI adoption affects the economy:
  - Displacement: tasks performed by labor are performed by capital (characterized by changes in 훼훼_z).
  - Complementarity: reallocates value added and labor demand from workers with less AI complementarity to workers with high AI complementarity (characterized by changes in 휂휂_z); assumed not to affect overall labor share.
  - Productivity: increases output and wages of workers with high AI complementarity (characterized by changes in 휓휓_z).
- In the model, the displacement channel is changes in 훼훼_z, the complementarity channel is changes in 휂휂_z, and the productivity channel is changes in 휓휓_z.

### Annex 4 — Additional scenarios and implications
- Two hypothetical scenarios illustrate displacement vs complementarity channels:
  - Scenario 1: Displacement effect affects all workers equally; complementarity follows data shown in Figure 9, panel 2.
    - Outcome: All workers lose labor income due to fewer tasks performed; workers with high AI complementarity gain demand for non-displaced tasks and accrue most productivity gains from AI adoption.
    - Consequence: More significant labor income and wealth inequality; high-AI-complementarity (high-income) workers capture most gains.
  - Scenario 2: Complementarity channel deactivated; displacement occurs according to data in Figure 9, panel 2 (i.e., exposure increases with income).
    - Outcome: Income gains from AI adoption are higher at the bottom of the income distribution because bottom workers are less exposed and suffer less task displacement.
    - Consequence: AI adoption leads to lower income inequality since capital income gains do not fully offset larger top-end labor income losses from task displacement.
- Calibration detail common to scenarios:
  - The calibrated change in the capital share is the same: 5.5 percentage points, in line with the change in the capital share observed in 1980–2014.
- Figure interpretation note: Plots show change in total income by income percentile decomposed into change in labor income (blue) and change in capital income (orange). P = percentile.

### Annex 5 — AI Preparedness Index (indicators and aggregation)
- Conceptual framework:
  - Four aggregate dimensions: digital infrastructure; human capital and labor market policies; digital innovation and economic integration; regulation and ethics.
  - Supplementary indicators: sustained human capital investment, inclusive STEM expertise, labor and capital mobility within and across countries, adaptability of legal frameworks to new (digital) business models.
  - Digital infrastructure and human capital and labor market policies are "foundational" elements; innovation and economic integration and regulation and ethics are "second-generation" elements.
- Coverage:
  - Index is computed for 32 advanced economies, 56 emerging market economies, and 37 low-income countries.
- Normalization (as presented in source):
  - Within each aggregate dimension, subindicators (푥푥)—for the latest year with available data—are normalized on a 0–1 scale as follows:
    - 푥푥−푥푥_푚푖푛
      ───────────────────
      푥푥_푚푎푥−푥푥_푚푖푛
- Aggregation rules:
  - Each aggregate dimension is the simple average of its normalized subcomponents.
  - The AI Preparedness Index is the simple average of the four aggregate dimensions.
- Robustness checks and caveats:
  - Two shortcomings of simple averaging:
    - Equal weighting may undervalue key components and overemphasize minor ones.
    - Sensitivity to outliers and extreme values.
  - Robustness via principal component analysis (PCA):
    - For each aggregate dimension, the first principal component (PC) of subindicators is extracted, normalized between 0 and 1, and the index is then computed as the sum of these normalized PCs.
    - Results based on PCA are indistinguishable from those obtained with simple averaging.
- Empirical corroboration:
  - Section V shows the index’s components are correlated with information and communications technology employment, with conditional correlations that make intuitive sense by development level.

### Numeric and factual highlights (preserved exact values)
- Historically, only about 20 percent of workers moving from high-exposure to low-exposure occupations also entered the informal sector.
- Calibrated change in the capital share used in model scenarios: 5.5 percentage points, in line with the change observed in 1980–2014.
- Index coverage: 32 advanced economies; 56 emerging market economies; 37 low-income countries.

*Source: Annex Table 3.1 and accompanying annexes (Annex Figures and Tables) from the IMF Staff Discussion Notes "Gen-AI: Artificial Intelligence and the Future of Work" (selected excerpts provided).*

### References

### References

### Artificial Intelligence, Automation, and the Future of Work
- Acemoglu, D., and P. Restrepo. 2018. “Artificial Intelligence, Automation, and Work.” In The Economics of Artificial Intelligence: An Agenda, edited by A. Agrawal, J. Gans, and A. Goldfarb, 197–236. Chicago: University of Chicago Press.
- Acemoglu, D., and P. Restrepo. 2022. “Tasks, Automation, and the Rise in US Wage Inequality.” Econometrica 90 (5): 1973–2016.
- Alonso, C., A. Berg, S. Kothari, C. Papageorgiou, and S. Rehman. 2022. “Will the AI Revolution Cause a Great Divergence?” Journal of Monetary Economics 127: 18–37.
- Brynjolfsson, E., L. Danielle, and L. R. Raymond. 2023. "Generative AI at Work." NBER Working Paper 31161, National Bureau of Economic Research, Cambridge, MA.
- Eloundou, T., S. Manning, P. Mishkin, and D. Rock. 2023. “GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models.” arXiv.org working paper.
- Felten, E., M. Raj, and R. Seamans. 2021. "Occupational, Industry, and Geographic Exposure to Artificial Intelligence: A Novel Dataset and Its Potential Uses." Strategic Management Journal 42 (12): 2195–217.
- Felten, E., M. Raj, and R. Seamans. 2023. "How Will Language Modelers Like ChatGPT Affect Occupations and Industries?" arXiv.org working paper.
- Gmyrek, P., J. Berg, and D. Bescond. 2023. Generative AI and Jobs: A Global Analysis of Potential Effects on Job Quantity and Quality. ILO Working Paper 96. International Labour Organization, Geneva, Switzerland.
- Moll, B., L. Rachel, and P. Restrepo. 2022. "Uneven Growth: Automation's Impact on Income and Wealth Inequality." Econometrica 90 (6): 2645–683.
- Webb, M. 2020. “The Impact of Artificial Intelligence on the Labor Market.” Stanford University Working Paper, Stanford, CA.

### Technology Adoption, Productivity, and Firm-Level Effects
- Autor, D. H., and D. Dorn. 2013. “The Growth of Low-S kill Service Jobs and the Polarization of the US Labor Market.” American Economic Review 103 (5): 1553–597.
- Autor, D. H., F. Levy, and R. J. Murnane. 2003. “The Skill Content of Recent Technological Change: An Empirical Exploration.” Quarterly Journal of Economics 118 (4): 1279–333.
- Bartel, A., C. Ichniowski, and K. Shaw. 2007. “How Does Information Technology Affect Productivity? Plant-Level Comparisons of Product Innovation, Process Improvement, and Worker Skills*.” Quarterly Journal of Economics 122 (4): 1721–758.
- Bloom, N., M. Draca, and J. Van Reenen. 2015. “Trade Induced Technical Change? The Impact of Chinese Imports on Innovation, IT and Productivity.” Review of Economic Studies 83 (1): 87–117.
- Braxton, J. Carter, and B. Taska. 2023. "Technological Change and the Consequences of Job Loss." American Economic Review 113 (2): 279–316.
- Babina, T., A. Fedyk, A. He, and J. Hodson. Forthcoming. “Artificial Intelligence, Firm Growth, and Product Innovation.” Journal of Financial Economics.
- Drozd, L. A., M. Taschereau-Dumouchel, and M. M. Tavares. 2022. "Understanding Growth through Automation." Research Department, Federal Reserve Bank of Philadelphia, Philadelphia, PA.
- Briggs, J., and D. Kodnani. 2023. “The Potentially Large Effects of Artificial Intelligence on Economic Growth.” Goldman Sachs - Global Economics Analyst, New York.

### Labor Markets, Occupational Mobility, and Inequality
- Cortes, G. M., N. Jaimovich, and H. E. Siu. 2017. "Disappearing Routine Jobs: Who, How, and Why?" Journal of Monetary Economics 91: 69–87.
- Cortes, G. M., N. Jaimovich, C. J. Nekarda, and H. E. Siu. 2020. "The Dynamics of Disappearing Routine Jobs: A Flows Approach." Labour Economics 65: 101823.
- Kambourov, G., and I. Manovskii. 2009. "Occupational Mobility and Wage Inequality." Review of Economic Studies 50: 731–59.
- Moscarini, G., and F. G. Vella. 2008. “Occupational Mobility and the Business Cycle.” NBER Working Paper 13819, National Bureau of Economic Research, Cambridge, MA.
- Monsueto, S. E., A. Moreira Cunha, and J. da Silva Bichara. 2014. "Occupational Mobility and Income Differentials: The experience of Brazil between 2002 and 2010." Cepal Review 113: 139–55.
- Pizzinelli, C., A. Panton, M. M. Tavares, M. Cazzaniga, and L. Li. 2023. "Labor Market Exposure to AI: Cross-Country Differences and Distributional Implications." IMF Working Paper 2023/216, International Monetary Fund, Washington, DC.
- Dabla-Norris, E., C. Pizzinelli, and J. Rappaport. 2023. "Job Polarization and the Declining Wages of Young Female Workers in the United Kingdom." Oxford Bulletin of Economics and Statistics 85 (6): 1185-1210.
- Yashiro, N., T. Kyyrä, H. Hwang, and J. Tuomala. 2022. “ Technology, Labour Market Institutions and Early Retirement.” Economic Policy 37 (112): 811–49.
- Das, M., and B. Hilgenstock. 2022. "The Exposure to Routinization: Labor Market Implications for Developed and Developing Economies." Structural Change and Economic Dynamics 60 (C): 99–113.
- Bergholt, D., F. Furlanetto, and N. Maffei-Faccioli. 2022. "The Decline of the Labor Share: New Empirical Evidence." American Economic Journal: Macroeconomics 14 (3): 163–98.
- Klenova, K., and A. Korinek. 2021. "AI and Shared Prosperity." In Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society, 645–51. [Note: as spelled in source: Klinova, K., and A. Korinek. 2021.]

### Digital Economy, Data, and Finance
- Amaglobeli, D,. R. de Mooij, A. Mengistu, M. Moszoro, M. Nose, S. Nunhuck, S. Pattanayak, L. Rivero del Paso, F. Solomon, R. Sparkman, H. Tourpe, and G. Uña. 2023. “Transforming Public Finance through GovTech.” IMF Staff Discussion Note 2023/004, International Monetary Fund, Washington, DC.
- Boukherouaa, E., G. Shabsigh, K. AlAjmi, J. Deodoro, A. Farias, E. S. Iskender, A. T. Mirestean, and R. Ravikumar. 2021. “Powering the Digital Economy: Opportunities and Risks of Artificial Intelligence in Finance.” IMF Departmental Paper 2021/024, International Monetary Fund, Washington, DC.
- Carriere-Swallow, Y., and V. Haksar. 2019. "The Economics and Implications of Data: An Integrated Perspective," IMF Departmental Paper 2019/013, International Monetary Fund, Washington, DC.
- Haksar, V., Y. Carriere-Swallow, E. Islam, A. Giddings, K. Kao, E. Kopp, and G. Quiros-Romero. 2021. "Toward a Global Approach to Data in the Digital Age," IMF Staff Discussion Note 2021/005, International Monetary Fund, Washington, DC.
- Shabsigh, G., and E. B. Boukherouaa. 2023. “Generative Artificial Intelligence in Finance.” Fintech Note 2023/006, International Monetary Fund, Washington, DC.
- Bank of America. 2023. “Cybersecurity Report: Landscape, Trends & What Comes Next.” Bank of America Global Research, New York.
- Jamilov, R., H. Rey, and A. Tahoun. 2023. "The Anatomy of Cyber Risk." NBER Working Paper 28906, National Bureau of Economic Research, Cambridge, MA.
- Financial Conduct Authority (FCA). 2022. “Machine Learning in UK Financial Services.” Bank of England and Financial Conduct Authority, London, UK.
- IFC. 2020. “Artificial Intelligence in Emerging Markets: Opportunities, Trends and Emerging Business Models.” International Finance Corporation. World Bank, Washington, DC.

### Cross-Country, Development, and Policy Analyses
- Albanesi, S., A. D. da Silva, J. F. Jimeno, A. Lamo, and A. Wabitsch. 2023. “New Technologies and Jobs in Europe.” CEPR Discussion Paper DP18220, Centre for Economic Policy Research, London, UK.
- Cirera, X., D. Comin, and M. Cruz. 2022. "Bridging the Technological Divide: Technology Adoption by Firms in Developing Countries." World Bank, Washington, DC.
- Ilzetzki, E., and S. Jain. 2023. The Impact of Artificial Intelligence on Growth and Employment. VoxEU.org, June 20.
- OECD. 2023. “OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market.” Paris, France.
- Oxford Insights. 2022. “Government AI Readiness Index.” Malvern, UK.
- UNESCO. 2021. “AI and Education: Guidance for Policy-Makers.” Paris, France.
- USAID. 2019. “Artificial Intelligence in Global Health: Defining a Collective Path Forward.” Washington, DC.
- Wahl, B., A. Cossy-Gantner, S. Germann, and N. R. Schwalbe. 2018. "Artificial Intelligence (AI) and Global Health: How Can AI Contribute to Health in Resource-Poor Settings?" BMJ Global Health 3 ( 4): e000798.
- Rockall, E., C. Pizzinelli, and M. Mendes Tavares. Forthcoming. "Artificial Intelligence Adoption and Inequality." IMF Working Paper, International Monetary Fund, Washington, DC.

### Data, Measurement, and Historical Perspectives
- Chinn, M. D., and R. W. Fairlie. 2007. “The Determinants of the Global Digital Divide: A Cross-Country Analysis of Computer and Internet Penetration.” Oxford Economic Papers 59 (1): 16–44.
- Felten, E., M. Raj, and R. Seamans. 2021. "Occupational, Industry, and Geographic Exposure to Artificial Intelligence: A Novel Dataset and Its Potential Uses." Strategic Management Journal 42 (12): 2195–217.
- Ribas, R. P., and S. S. D. Soares. 2008. “The IBGE Monthly Employment Survey (PME) Panel.” Discussion Paper 1348, Institute of Applied Economic Research, Brasília, Brazil.
- Wootton, C. W., and B. E. Kemmerer. 2007. "The Emergence of Mechanical Accounting in the US, 1880–1930." Accounting Historians Journal 34 (1): 91–124.
- Keller, W. “International Technology Diffusion.” 2004. Journal of Economic Literature 42 (3): 752–82.
- Nicoletti, G., C. V. Rueden, and D. Andrews. 2020. “Digital Technology Diffusion: A Matter of Capabilities, Incentives or Both?” European Economic Review 128: 103513.
- Pizzinelli, C., A. Panton, M. M. Tavares, M. Cazzaniga, and L. Li. 2023. "Labor Market Exposure to AI: Cross-Country Differences and Distributional Implications." IMF Working Paper 2023/216, International Monetary Fund, Washington, DC.
- Carriere-Swallow, Y., and V. Haksar. 2019. "The Economics and Implications of Data: An Integrated Perspective," IMF Departmental Paper 2019/013, International Monetary Fund, Washington, DC.

*Gen-AI: Artificial Intelligence and the Future of Work — Staff Discussion Note No. SDN/2024/001*

---


_Source: https://www.imf.org/-/media/files/publications/sdn/2024/english/sdnea2024001.pdf_