Evidence (3566 claims)
Adoption
8570 claims
Productivity
7631 claims
Governance
6869 claims
Human-AI Collaboration
6491 claims
Org Design
4175 claims
Innovation
4114 claims
Labor Markets
3566 claims
Skills & Training
2966 claims
Inequality
2066 claims
Evidence Matrix
Claim counts by outcome category and direction of finding.
| Outcome | Positive | Negative | Mixed | Null | Total |
|---|---|---|---|---|---|
| Other | 758 | 199 | 100 | 900 | 2007 |
| Governance & Regulation | 826 | 400 | 191 | 122 | 1563 |
| Organizational Efficiency | 777 | 193 | 124 | 84 | 1189 |
| Technology Adoption Rate | 635 | 233 | 124 | 97 | 1098 |
| Research Productivity | 422 | 128 | 57 | 336 | 954 |
| Output Quality | 476 | 179 | 59 | 47 | 761 |
| Decision Quality | 328 | 177 | 81 | 47 | 640 |
| Firm Productivity | 435 | 57 | 88 | 20 | 606 |
| AI Safety & Ethics | 218 | 277 | 65 | 33 | 599 |
| Market Structure | 180 | 170 | 123 | 24 | 502 |
| Task Allocation | 213 | 64 | 72 | 33 | 387 |
| Skill Acquisition | 170 | 61 | 61 | 17 | 309 |
| Innovation Output | 203 | 27 | 43 | 18 | 292 |
| Employment Level | 105 | 54 | 107 | 13 | 281 |
| Fiscal & Macroeconomic | 131 | 69 | 43 | 26 | 276 |
| Consumer Welfare | 117 | 63 | 42 | 11 | 233 |
| Firm Revenue | 153 | 48 | 26 | 3 | 230 |
| Task Completion Time | 173 | 31 | 8 | 12 | 225 |
| Inequality Measures | 44 | 122 | 49 | 6 | 221 |
| Worker Satisfaction | 89 | 65 | 22 | 12 | 188 |
| Error Rate | 69 | 92 | 10 | 2 | 173 |
| Regulatory Compliance | 77 | 69 | 14 | 5 | 165 |
| Automation Exposure | 56 | 56 | 26 | 13 | 154 |
| Training Effectiveness | 94 | 21 | 13 | 19 | 149 |
| Wages & Compensation | 77 | 36 | 25 | 6 | 144 |
| Team Performance | 86 | 17 | 27 | 10 | 141 |
| Developer Productivity | 95 | 17 | 14 | 6 | 133 |
| Job Displacement | 12 | 80 | 20 | 1 | 113 |
| Hiring & Recruitment | 52 | 7 | 8 | 3 | 70 |
| Creative Output | 31 | 18 | 8 | 3 | 61 |
| Skill Obsolescence | 5 | 46 | 6 | 1 | 58 |
| Social Protection | 27 | 16 | 8 | 2 | 53 |
| Labor Share of Income | 17 | 19 | 17 | — | 53 |
| Worker Turnover | 11 | 12 | — | 3 | 26 |
| Industry | — | — | — | 1 | 1 |
Labor Markets
Remove filter
High fixed costs may concentrate training capacity among a few providers, risking reduced competition.
Listed under Risks to Watch: the paper warns that high fixed costs could concentrate capacity. This is a theoretical market-concentration risk; no empirical market analysis is provided.
Upfront and maintenance costs are substantial; economic evaluation should compare these costs to downstream benefits such as placement rates and productivity gains.
Paper recommends economic evaluation, lists cost-per-curriculum and other cost metrics; presented as advice rather than results. No empirical cost–benefit data provided.
Complexity and lock-in to specific standards may raise barriers to innovation and increase switching costs.
Discussed in Regulation and compliance economics and Risks: claims that standardisation and embedded processes could produce vendor/standard lock-in. This is a theoretical risk flagged by the authors, not supported by empirical data in the paper.
Upfront governance costs (policy, tooling, staff) become a key part of adoption cost and affect ROI calculations and payback periods for automation investments.
Economic reasoning and implications discussed in the paper; no empirical cost data provided—recommendation based on practitioner experience and theoretical cost accounting.
Traditional automation governance is often ad hoc, underestimates security and compliance risks, and does not scale safely for mission-critical enterprise systems.
Synthesis of industry best practices and practitioner-sourced lessons (qualitative observations and case illustrations). No systematic survey or quantitative incidence rates provided.
Prompt fraud reduces the marginal cost of producing convincing fraudulent artifacts, which may increase fraud frequency and expected losses absent mitigations.
Economic reasoning and conceptual modeling of incentives; no empirical estimates of frequency or losses included.
Lack of prompt provenance, versioning, and validation practices increases organizational exposure to prompt fraud.
Conceptual analysis and recommended controls (provenance/versioning) drawn from audit-framework comparisons and threat modeling.
There is insufficient logging/traceability of prompts, responses, and model versions in many workflows, creating a control weakness for detecting prompt fraud.
Observations from literature/regulatory review and the paper's threat/control mapping; asserted as a common operational gap (no systematic measurement).
Shadow AI — unsanctioned, decentralized use of GenAI tools — amplifies prompt-fraud risk by bypassing central controls and audit trails.
Conceptual analysis and organizational risk reasoning; references to common practices of unsanctioned tool use (no empirical prevalence data).
External actors can commit prompt fraud via customer-facing systems or social-engineering prompt chains.
Conceptual threat scenarios and mapping of attack surfaces (customer-facing interfaces, input channels); illustrative examples provided.
Internal actors manipulating prompts within authorized AI workflows are a realistic and important threat vector for prompt fraud.
Threat modeling and scenario-based analysis highlighting insiders with authorized access who can craft prompts.
Prompt fraud can defeat controls that rely on plausibility, standard formatting, or human review that trusts model-like language.
Threat mapping and literature on automation bias; illustrative vignettes showing how machine-like outputs mimic authoritative formats.
Prompt fraud lowers the entry cost of producing convincing fraudulent artifacts, increasing the ease with which attackers can create plausible forgeries.
Economic reasoning and conceptual analysis based on GenAI behavior and illustrative scenarios (no empirical cost or frequency data).
Prompt fraud — the intentional manipulation of natural-language prompts to cause generative AI systems to produce misleading, fabricated, or deceptive artifacts that bypass internal controls — constitutes a novel, low-cost fraud vector that traditional IT- and process-focused controls are ill-equipped to detect or prevent.
Conceptual analysis and threat modeling grounded in literature/regulatory review and illustrative vignettes; no systematic empirical incidence data provided.
Secure infrastructure (including SECaaS-provided tools) affects the availability and trustworthiness of AI training data and models; breaches reduce returns to AI R&D via direct losses and reduced trust.
Conceptual linkage supported by case studies of data/model theft and technical literature on secure enclaves, differential privacy, federated learning; no broad quantitative estimate provided.
Security externalities (one firm's breach raising ecosystem risk) complicate private incentives and may justify policy interventions such as standards or mandatory reporting.
Economic theory on externalities, case studies showing spillovers from breaches, and policy analyses recommending interventions.
Concentration among large cloud/SECaaS providers can create market power, platform dependency, and affect competition in AI markets.
Market-structure theory, observed concentration patterns in industry reports, and qualitative case studies; no causal estimates provided in the chapter.
Latency and integration frictions can limit the suitability of SECaaS for specialized workloads, including some AI pipelines.
Technical evaluations and benchmarks that measure latency/resource overhead; reports and case studies noting integration challenges for high-throughput or low-latency workloads.
Reliance on a small set of major cloud/SECaaS providers creates vendor lock-in, concentration risk, and systemic vulnerability if a major provider is compromised.
Market-structure discussions, observed provider outages and incidents (case studies), and theoretical arguments about concentration; no single causally identified empirical estimate provided.
Resource-rich labs and firms are likely to adopt LLM orchestration faster, which could widen gaps in research capacity between institutions and countries unless mitigated by policy choices.
Equity and diffusion argument based on resource requirements (compute, data, validation); no adoption-rate data or cross-institution comparisons provided.
There is potential for 'winner-take-most' market outcomes if a few players combine superior models, instrument control software, and exclusive datasets.
Economics reasoning about network effects and data concentration; no empirical market concentration metrics specific to microscopy provided.
Upfront investments required for compute, data labeling, validation, and safety testing may raise entry costs and favor incumbents.
Economic logic about fixed costs and scale advantages; no measured entry-cost or firm-dynamics data provided.
There is a risk of deskilling for some technical roles, creating implications for training and workforce development.
Theoretical reasoning about automation-induced deskilling; no empirical study or measured skill changes provided.
There is a nonlinear 'Digital Exclusion Trap': fiscal support is ineffective or harmful in places below a critical level of digital infrastructure.
Nonlinear/threshold tests and heterogeneous-effect analyses in the DID framework showing that treatment effects on cultural employment vary by digital infrastructure level, with null or negative effects below an estimated threshold (analysis on 280 cities, 2008–2021).
AI reshapes local labor markets by automating routine tasks.
Micro-level analysis of occupations and task content using granular online job-posting data (decomposition of occupational and task changes); panel and IV analyses link higher AI exposure to declines in routine-task employment shares.
Multipolar competition in AI increases risks of fragmented regulations, export control cascades, and inefficient duplication of standards, producing large economic coordination and collective‑action costs.
Theoretical argument and literature synthesis on international political economy of standards and controls; no novel quantitative cost estimates, though the paper recommends empirical research avenues to quantify these costs.
AI‑driven information operations, recommendation systems, and content economies alter market incentives, advertising revenues, and the political economy of attention—creating externalities not priced in markets.
Interpretive synthesis of literature on digital platforms, misinformation, and attention economics; supported by cited secondary studies and policy examples rather than new empirical measurement.
Competition over AI standards, data governance norms, and platform rules is an economic contest with long‑run market structure implications (network effects, winner‑take‑most outcomes).
Theoretical synthesis drawing on platform economics and standards literature; supported by qualitative examples of standard‑setting contests but without new quantitative market structure analysis.
Export controls, sanctions, investment screening, and tech diplomacy function as economic levers of smart power and reshape global AI supply chains, FDI flows, and comparative advantage.
Policy‑focused evidence and examples cited in the literature review and case studies; proposed policy event‑study approaches are suggested but no original empirical event study is presented.
The digital/AI era changes both the tools (new technological instruments of influence) and the targets (information environments, data infrastructures), creating novel governance and collective‑action problems.
Conceptual analysis supported by literature synthesis on digital platforms, AI, surveillance, and information operations; illustrative examples from policy and secondary studies rather than original empirical measurement.
Short-run displacement risks from AI adoption create distributional concerns that warrant active labor market policies (retraining, wage insurance) and portable social protections.
Worker-level evidence of short-run employment losses in routine occupations, particularly in emerging economies, and literature synthesis on displacement effects motivating policy recommendations.
Human-in-the-loop controls formalize supervisory labor and create persistent oversight costs even after automation scales.
Pattern design and governance lifecycle recommendations highlighting human checkpoints; qualitative reasoning without measurement of oversight hours or costs.
Higher non-wage costs and higher formalization costs create barriers to creating formal salaried employment and alter firms’ hiring and investment decisions.
Theoretical and policy interpretation based on measured NWC and CFIL levels in the 19-country sample and economic reasoning about how employer cost structure affects hiring and investment incentives; no firm-level causal estimation reported.
Labor costs in Latin America and the Caribbean have risen since 2013, and divergence in labor costs across countries has widened over that period.
Comparison of the updated 2023 indicator estimates with prior IDB estimates (2013) across the 19-country sample; temporal comparison of country-level indicators and summary statistics showing increased dispersion.
AI-enabled platforms can increase market concentration and platform power, creating competition and data-governance risks and uneven distributional effects across regions and worker skill levels.
Observational platform-concentration indicators and distributional analyses in the case material; scenario and sensitivity checks on distributional outcomes under alternative adoption/policy regimes.
AI substitutes for and displaces many routine and low-skill occupations, increasing automation risk for those roles.
Multiple empirical studies in the reviewed sample document higher automation/substitution risk and observed employment declines in routine/low-skill tasks and occupations.
Young workers experience pronounced negative effects in occupations exposed to AI.
Demographic breakdowns in occupation-level analyses showing larger employment declines (or weaker employment growth) for younger cohorts in AI-exposed occupations.
Diffusion of AI skills is associated with lower employment in occupations that are both highly exposed to AI and have low complementarity with it.
Panel/cross-sectional analyses linking occupation-level AI exposure and measured worker–AI complementarity to employment changes, using occupation classifications of exposure and complementarity.
Middle-skilled occupations are most at risk, contributing to a shrinking middle class (declines in middle-skilled employment).
Occupation-level analyses showing employment declines concentrated in middle-skilled occupations as new skills (IT/AI) diffuse.
AI adoption can reinforce winner‑take‑most market dynamics and increase market concentration due to data‑ and AI‑driven advantages.
Theoretical arguments and industry analyses on platform markets and data economies; empirical market‑structure studies and descriptive evidence cited in the review; the claim is derived from synthesis rather than a single causal identification design.
Impacts of AI on labor are uneven globally: developing regions face larger risks due to digital infrastructure gaps, limited reskilling capacity, and weaker social protections.
Cross‑country comparative analyses, policy and industry reports highlighting infrastructure and institutional differences, and sectoral case studies; review notes geographic bias toward advanced economies in the empirical literature, making some cross‑region inference provisional.
There is widespread displacement of routine and lower‑skilled tasks associated with AI and automation.
Task‑based analyses decomposing occupations into automatable vs augmentable tasks, econometric studies correlating measures of automation/AI exposure with declines in employment and/or hours in routine occupations, and industry reports documenting automation of routine tasks; evidence is largely from cross‑country and country‑specific empirical work summarized in the review.
Traditional macro indicators (GDP, income, unemployment) explain less than 5% of the state- and county-level variation in skills-based exposure.
Statistical analysis/regressions relating the Iceberg Index to standard macro indicators at state and county levels (reported explained variance R^2 < 0.05); sample includes all U.S. states and ~3,000 counties.
The broader cognitive automation potential is roughly five times larger than visible adoption and is geographically widespread (present across all states, not only coastal hubs).
Direct comparison of the two model-derived aggregates (11.7% vs 2.2%) and spatial analysis of the Iceberg Index across ~3,000 counties and all states in the simulation.
Broader cognitive automation potential across administrative, financial, and professional services amounts to 11.7% (~$1.2 trillion).
Iceberg Index computation summing the wage-value contributions of skills that current AI capabilities can perform; based on mapping of thousands of AI tools to ~32,000 skills and the simulated 151M-agent workforce across ~3,000 counties.
Visible AI adoption concentrated in computing/technology represents about 2.2% of U.S. wage value (~$211 billion).
Model-derived visible-adoption metric computed from mapped AI tool usage in technology/computing occupations, applied to the simulated 151M-worker population and national wage data to estimate percentage and dollar value.
Prevailing reskilling strategies assume access to stable employment, time and funds for training, certification systems, and institutional support — conditions that are weak or absent for informal platform workers; therefore standard reskilling policies are poorly suited to this context.
Qualitative synthesis of policy analyses and literature on reskilling programs and labour-market institutions; conceptual critique rather than new empirical testing.
Algorithmic management (opaque algorithms for assignment, pricing, and performance metrics) restructures platform work in ways that both change task composition and intensify precarity, reducing workers' ability to adapt to automation.
Draws on prior empirical studies and policy analyses of algorithmic management cited in the literature review; no new empirical data collected in this paper.
Task versus job displacement operate differently across institutional contexts: in formal labour markets, task automation can be accommodated through reallocation or protections, while in informal platform work task loss typically becomes outright job loss.
Argument built from secondary literature comparing formal and informal labour-market institutions and existing empirical studies on reallocation mechanisms; conceptual analysis in the paper (qualitative synthesis only).
AI-driven automation in platform-based informal work in India primarily displaces tasks, but because workers lack job security, institutional protections, and access to alternative labour tracks, task-level automation often manifests as full job displacement.
Synthesis of prior empirical studies, policy analyses, and theoretical work on platform-based labour and automation focused on India and comparable developing-country settings; conceptual framing distinguishing task-level vs job-level effects; no primary data or new empirical analysis in this paper.