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
Routine, boilerplate, and debugging tasks are most automatable or complemented by LLMs, shifting value toward design, verification, and systems thinking.
Task-level analyses, observational studies, and synthesized findings showing larger gains on repetitive or templated tasks versus high-level design tasks.
Liability and intellectual-property ownership around AI-assisted code are unresolved practical and legal concerns.
Legal and policy analyses, practitioner reports, and qualitative interviews noting ambiguous legal frameworks and unresolved questions about ownership and liability for AI-assisted code.
Token taxes reduce some geographic tax arbitrage relative to input taxes but do not eliminate cross-border avoidance; international coordination and trade/regulatory levers are crucial.
Political-economy analysis and recommendations in the paper; no international case studies or empirical coordination outcomes provided.
A robust empirical pattern in the literature is that AI’s effects vary by skill level: displacement risk is concentrated among lower-skilled tasks while augmentation and wage gains are more likely for higher-skilled tasks.
Empirical findings and syntheses cited (Brynjolfsson et al., 2023; Chen et al., 2024) that report task- and skill-differentiated effects on employment and wages; evidence comprises cross-sectional exposure analyses and panel studies in the cited literature.
Short-run consumer gains from faster, cheaper service can be undermined by trust losses from hallucinations or perceived deception, reducing long-term consumer surplus.
Conceptual welfare analysis and cited case examples in the literature; no longitudinal consumer-surplus measurement provided in this review.
Conventional productivity metrics (e.g., handle time) may misstate value because they do not capture multi-dimensional impacts like quality and trust.
Conceptual critique and synthesis of measurement challenges discussed in the literature; no empirical measurement study presented in this review.
There is potential for substantial cost savings and throughput gains in repetitive, high-volume interactions, but these are offset by costs for integration, monitoring, and error remediation.
Industry case examples and conceptual cost–benefit reasoning aggregated in the review; the paper contains no new quantitative cost estimates or sample-based measurements.
Generative AI will substitute for routine service tasks while complementing skilled workers for escalations and complex problem solving, shifting labor demand toward supervisory and relationship-focused roles.
Economic and labor-market analyses synthesized in the review; projections are inferential and based on heterogeneous secondary sources, not primary labor-market experiments.
Full automation of customer service is suboptimal because persistent risks (hallucinations, contextual errors, lack of genuine empathy, integration complexity) remain; hybrid human–AI systems achieve the best outcomes.
Synthesis of documented failure modes and practitioner case examples from the literature; no primary experimental data or controlled trials in this review. Inference based on heterogeneous empirical reports and conceptual analyses.
Welfare effects of democratized access to AI-assisted ideation are ambiguous: access could democratize innovation but also amplify low-quality outputs and misinformation absent proper curation.
Theoretical discussion and empirical examples of misinformation/low-quality outputs from LLMs cited in the review; no comprehensive welfare accounting provided.
Net gains in innovation from increased idea volume depend on complementary human capacity for curation and development; raw increases in ideas do not automatically translate into higher-quality innovation.
Synthesis noting studies where idea quantity rose but downstream quality or successful development did not necessarily increase; review highlights heterogeneity across workflows and dependence on human integration.
The most effective deployment model is a 'cognitive co-pilot' in which AI expands and challenges the idea space while humans provide curation, strategic evaluation, and experiential judgment.
Prescriptive conclusion drawn from synthesis of studies where human-AI collaboration (human curation/selection) produced better downstream outcomes than AI-alone outputs; evidence heterogenous and largely short-term.
Generative AI functions as a dual-purpose cognitive tool: a high-volume catalyst for divergent idea generation and a structured assistant for decomposing complex problems.
Nano-review / synthesis of existing empirical literature on LLM-assisted creativity and problem-solving, drawing on experimental ideation tasks, design/ideation studies, and applied case evidence; no original dataset or new experiments in this paper.
Net value from generative AI is contingent: gains are largest where breadth of ideas and rapid iteration matter, and smaller or riskier where deep domain expertise, tacit knowledge, or high-stakes judgments are required.
Synthesis of heterogeneous empirical results showing task-dependent benefits; argument grounded in observed differences across lab and field contexts and documented limitations in domain-specific performance.
Generative AI raises measurable productivity (lower marginal cost per interaction) but introduces quality and trust externalities; optimal deployment balances these trade-offs.
Pilot cost analyses and operational reports showing lower marginal costs per interaction alongside documented quality/trust issues; primarily observational and model-based reasoning.
Full automation produces trade-offs unfavorable to complex service quality and trust; hybrid models with human-in-the-loop control are preferable.
Synthesis of case studies, pilot results, and conceptual reasoning comparing fully automated routing to hybrid/human-in-the-loop deployments; limited randomized comparisons.
Generative AI can materially improve customer service productivity through 24/7 automation, scalable personalization, and agent augmentation — but is not a substitute for humans.
Synthesis of deployments, pilot studies, vendor reports, and some experimental A/B tests described in the paper; no pooled sample size provided and much evidence is short-run or observational.
Governance reduces downside risk (compliance fines, outages) but raises implementation costs; economic assessments must weigh risk-adjusted returns.
Conceptual economic argument in the paper; supported by reasoning and practitioner experience but not by empirical cost–benefit studies within the article.
Safer scaling of automation may increase substitution of routine ERP/CRM tasks while governance and oversight roles create complementary high-skill positions (e.g., compliance engineers, auditors, prompt engineers).
Labor-market implications presented as theoretical reasoning based on how governance and automation interact; informed by practitioner observation but not empirically tested in the paper.
Overall, secure and resilient cloud infrastructure supported by SECaaS facilitates broader and safer diffusion of AI but creates economic trade-offs (market concentration, externalities, liability) that require empirical study and policy responses.
Synthesis of the chapter's literature review, case studies, and theoretical arguments; calls for empirical methods (regressions, event studies, structural models) to quantify effects.
Outsourcing via SECaaS shifts demand from in-house security labor to vendor-side security professionals, altering labor market composition and geographic distribution of expertise.
Labor-market reasoning and some survey evidence on outsourcing trends; chapter recommends empirical study (e.g., labor data, regional analyses) but does not present a specific dataset.
Tools such as secure enclaves, differential privacy, federated learning, and MPC influence the feasibility and cost of privacy-preserving AI; SECaaS providers offering these capabilities can change competitive dynamics.
Technical literature and vendor feature sets describing these technologies; theoretical implications for cost and competition discussed in the chapter.
Cyber insurance markets interact with SECaaS adoption; insurers may incentivize or require specific controls, altering firms’ security choices and underwriting practices.
Industry reports on cyber insurance requirements, surveys of insurer underwriting practices, and theoretical interaction effects; empirical analyses proposed (linking adoption to premiums).
Network effects in threat intelligence and telemetry can lead to winner-take-most outcomes but also increase the social value of shared defenses.
Theoretical arguments about network effects and empirical observation of aggregation benefits in threat-sharing initiatives; literature on public-good aspects of shared threat intelligence.
Pricing and contract design of SECaaS shape firm investment in complementary capabilities (data governance, secure model deployment).
Theoretical economic arguments and structural market models suggested in the chapter; empirical tests proposed (e.g., regressions, structural estimation) but no definitive empirical sample presented.
Emerging AI-driven strain optimization reduces design costs and may concentrate advantage with firms holding large proprietary datasets and compute resources, creating platform effects.
Economic argument supported by observed uses of proprietary datasets and ML in reviewed technical studies, and conceptual analysis of platform economics and data-driven advantage discussed in the paper.
'De-organized Growth' represents a structural shift toward decentralized, less formalized cultural work instead of firm-based expansion.
Synthesis of empirical findings: positive employment change without enterprise-count growth, plus evidence of increased platform-mediated gigs and procurement-driven work; derived from DID estimates and descriptive analyses of work organization patterns across cities (280 cities, 2008–2021).
Workforce transitions induced by AI imply distributional consequences (winners and losers), so policies should anticipate transitional unemployment and reskilling needs.
Inference from documented labor-market compositional changes (decline in routine tasks, growth in green occupations) combined with policy discussion in the paper; not a direct causal estimate of unemployment outcomes.
AI-enabled macro and fiscal models can improve policy testing and contingency planning but require transparency, validation, and safeguards against overreliance.
Conceptual argument and illustrative examples; no empirical trials or model performance metrics reported.
AI shifts the locus of economic governance from static rules to living systems that anticipate shocks and adapt in real time.
Policy-analytic framing and scenario-based reasoning within the book; supported by illustrative examples rather than empirical measurement.
International spillovers of AI-driven productivity depend on trade linkages and cross-border data flows; they are weaker when such linkages are limited.
Cross-country comparisons using trade flow data and measures of cross-border data policy/infrastructure; heterogeneous treatment effects in firm-level panels and country aggregates conditional on trade openness and data flow indices.
Emerging and low- and middle-income economies show smaller productivity gains (roughly 2–6%) and larger short-run job losses in routine occupations after AI adoption.
Estimates from worker-level microdata and firm panels in emerging economy samples, event studies of employment by occupation, and occupational task classification (ISCO/ISCO-08) to identify routine jobs.
Land-transfer effects on AGTFP are positive but constrained: institutional frictions limit the contribution of land transfer to green transformation.
Mediation results indicating a positive but limited indirect effect via land transfer/scale expansion, supplemented by discussion of institutional barriers in the paper.
Widening cross-country divergence in labor costs implies heterogeneous pathways for AI adoption and labor-market impacts across the region (high-cost countries may see faster automation and different skill-demand shifts than lower-cost ones).
Observed increased divergence in the 2013–2023 comparison across the 19-country sample plus theoretical mapping from cost levels to likely automation incentives; no direct panel evidence linking country-level cost divergence to differential AI adoption rates is provided.
The note provides 2025 projections that incorporate recent legal reforms in six countries, changing future cost estimates.
Projection exercise using the 19-country baseline (2023) and explicitly incorporating known legislative/reform changes enacted in six countries to update NWC, MCSL and CFIL projections to 2025.
Automation reshapes job tasks — reducing demand for some routine manual roles while increasing demand for technical, supervisory, logistics-planning, and service roles — implying substantial reskilling needs rather than outright net job collapse.
Labor-market analysis using occupational employment and job-posting data (task content), supplemented by qualitative interviews and surveys tracing task changes and reskilling needs; scenario sensitivity checks on net employment under alternative adoption paths.
Broader conclusion: AI has the potential to raise productivity and create value, but without proactive policy the benefits risk being concentrated among skilled workers and firms, exacerbating inequality and regional disparities.
Integrative interpretation drawing on productivity and distributional findings from the 17 studies and theoretical considerations about differential complementarities and adoption patterns.
Whether AI is net job‑creating depends on context (sector, country, policy environment, and workforce skill composition).
Observed heterogeneity across the 17 studies by sectoral setting, country context, and policy environment; studies report differing net employment outcomes depending on these factors.
AI contributes to labor‑market polarization: growth in high‑skill opportunities alongside contraction in many middle- and low‑skill roles.
Comparative synthesis of occupational and wage-composition findings across the 17 studies shows recurring patterns of expansion at the high-skill end and reductions in middle/low-skill employment.
Cross-country variation in demand versus supply of new skills is large, and this variation is captured by a Skill Imbalance Index.
Construction of a Skill Imbalance Index at the country level that compares skill demand (vacancies requesting new skills) to proxies for skill supply (worker skill endowments or related measures); country-level comparisons show wide variation in the index.
Labor-market polarization intensifies: gains are concentrated among high-skilled workers.
Occupation-level analyses of employment and wage changes showing larger positive effects for high-skilled occupations following adoption of new skills.
Overall employment and wages rise where new skills are adopted, but these gains are uneven across workers and occupations.
Cross-sectional and panel analyses relating diffusion of new skills (measured from vacancies) to changes in employment and wages across occupations and demographic groups.
Expected differential wage pressure: wages are likely to fall for routine/low‑skill occupations and rise or remain stable for high‑skill workers who possess complementary AI skills.
Econometric studies summarized in the review (cross‑sectional and panel regressions) and theoretical consistency with SBTC; the review highlights heterogeneity in findings and limited long‑run causal certainty.
AI contributes to skills polarization: demand rises for advanced cognitive, digital, and socio‑emotional skills while routine cognitive and manual task demand declines.
Theoretical integration (SBTC), task decomposition studies showing shifts in task demand by skill content, and labour‑market analyses reporting changes in occupational skill mixes; evidence comes from cross‑sectional and panel studies summarized in the review.
AI/ML has a dual, sector- and skill-dependent effect on labor: widespread displacement of routine and lower-skilled tasks coexists with augmentation of professional and cognitive work and the creation of new labor forms (gig, platform-mediated, and human–AI hybrid roles).
Systematic synthesis of peer‑reviewed empirical studies, industry and policy reports, task‑based analyses, and firm/establishment case studies across cross‑country and sectoral analyses; empirical approaches include econometric (cross‑sectional and panel) studies linking automation/AI adoption to employment and wages, task decomposition analyses, and surveys of firm adoption and restructuring. The review notes heterogeneity across studies and limited long‑run causal evidence.
AI technical capability in the U.S. labor market is substantially larger and far more geographically diffuse than visible adoption suggests.
Agent-based simulation that maps thousands of AI tools to a skills taxonomy and a synthetic population representing the U.S. workforce (151 million agents), covering 32,000+ skills and ~3,000 counties; comparison of the Iceberg Index (skills-based exposure) to a visible-adoption wage-share metric.
The paper presents hypothesis tests assessing whether university status (and Alliance ranking) and the presence of specialized AI programs affect graduate employment effectiveness, and reports identification of key/high-performing universities.
Statement of empirical approach: hypothesis testing on effects of university status/Alliance ranking and specialized programs using the monitoring dataset; results and significance levels are reported in the full article.
Heterogeneity across universities implies that targeting high-performing institutions and diffusing their practices could be more effective than uniform expansion of AI training.
Observed variation in employment effectiveness, placement outcomes, and wages across the 191 universities; policy implication drawn from comparative performance patterns.
Labor market institutions (unions, collective bargaining), education and training systems, social safety nets, and regulations substantially mediate distributional and aggregate outcomes of AI adoption.
Comparative institutional analysis and equilibrium models linking institutional settings to wage-setting and reallocation dynamics, supported by empirical cross-jurisdiction comparisons where available.
Developing economies face different trade-offs from AI adoption than advanced economies, due to different occupational structures and complementarities.
Comparative analyses and sectoral studies drawing on cross-country microdata and institutional comparisons; theoretical models highlighting differences in task composition and absorptive capacity.