Evidence (1441 claims)
Search and filter individual claims pulled from the papers. Looking for a specific finding ("what's the effect on wages?"), you're in the right place. Want to compare whole outcome categories against each other instead? Use the Evidence Explorer.
The board below groups claims two ways: by broad theme (nine paper-level topics) and by outcome category (the 34 claim-level outcomes that the Explorer and Syntheses also use).
Browse by theme
Nine broad, paper-level topics. Click one to filter the claims below.
Adoption
20058 claims
Filter claims →
Productivity
17184 claims
Filter claims →
Governance
16099 claims
Filter claims →
Human-AI Collaboration
16034 claims
Filter claims →
Innovation
10501 claims
Filter claims →
Org Design
10496 claims
Filter claims →
Labor Markets
6444 claims
Filter claims →
Skills & Training
5385 claims
Filter claims →
Inequality
4148 claims
Filter claims →
Claims by outcome category
Counts by direction of finding. These are the same 34 outcome categories the Explorer compares and the Syntheses are written for. A linked row has a published synthesis.
| Outcome | Positive | Negative | Mixed | Null | Total |
|---|---|---|---|---|---|
| Other | 1820 | 479 | 278 | 1820 | 4588 |
| Organizational Efficiency | 2711 | 616 | 401 | 173 | 3922 |
| Governance & Regulation | 2075 | 886 | 459 | 246 | 3714 |
| Technology Adoption Rate | 1467 | 530 | 258 | 206 | 2488 |
| Decision Quality | 1281 | 496 | 289 | 152 | 2228 |
| Output Quality | 1227 | 447 | 207 | 138 | 2025 |
| AI Safety & Ethics | 634 | 754 | 207 | 83 | 1688 |
| Research Productivity | 826 | 241 | 114 | 422 | 1624 |
| Firm Productivity | 1052 | 154 | 163 | 66 | 1441 |
| Task Allocation | 685 | 211 | 331 | 99 | 1335 |
| Market Structure | 433 | 423 | 242 | 46 | 1150 |
| Innovation Output | 639 | 91 | 105 | 34 | 871 |
| Task Completion Time | 476 | 113 | 43 | 36 | 672 |
| Firm Revenue | 445 | 126 | 58 | 25 | 656 |
| Skill Acquisition | 364 | 119 | 109 | 34 | 626 |
| Consumer Welfare | 288 | 167 | 104 | 31 | 592 |
| Employment Level | 214 | 140 | 174 | 50 | 582 |
| Error Rate | 230 | 251 | 35 | 16 | 535 |
| Fiscal & Macroeconomic | 268 | 136 | 71 | 50 | 532 |
| Inequality Measures | 100 | 307 | 96 | 12 | 515 |
| Worker Satisfaction | 221 | 173 | 60 | 30 | 484 |
| Automation Exposure | 155 | 138 | 65 | 36 | 398 |
| Regulatory Compliance | 171 | 120 | 30 | 13 | 335 |
| Developer Productivity | 222 | 58 | 27 | 13 | 321 |
| Team Performance | 188 | 56 | 50 | 24 | 320 |
| Wages & Compensation | 146 | 104 | 46 | 16 | 312 |
| Training Effectiveness | 207 | 41 | 21 | 26 | 298 |
| Job Displacement | 23 | 153 | 52 | 4 | 232 |
| Hiring & Recruitment | 102 | 57 | 30 | 11 | 202 |
| Skill Obsolescence | 16 | 102 | 24 | 6 | 148 |
| Creative Output | 71 | 42 | 23 | 6 | 143 |
| Social Protection | 57 | 30 | 11 | 3 | 101 |
| Labor Share of Income | 29 | 42 | 24 | 2 | 97 |
| Worker Turnover | 43 | 29 | 6 | 4 | 82 |
| Industry | — | — | — | 1 | 1 |
Twitter-derived signals and Google Trends signals provide complementary information for apparel demand forecasting; neither signal source uniformly dominates the other when used alone.
Pairwise head-to-head comparisons of signal sets, in which Twitter alone and Google Trends alone each won approximately 51%–57% of matchups.
Regions with higher human capital and adoption capacity capture more productivity gains, while disadvantaged regions face stagnation.
Presented as a regional heterogeneity claim; no regional panel, productivity measure, or comparative estimate is supplied.
AI-driven automation accelerates occupational task reallocation, raising productivity but producing uneven wage effects.
Asserted in the supplied example contribution; no underlying paper, dataset, sample, or statistical analysis is provided.
Leadership effects in studies of AI adoption and productivity are endogenous to firm selection and internal processes, so causal studies should use instruments or quasi-experimental designs.
The paper's methodological implication concerning endogeneity and causal inference in AI adoption and productivity research.
The severity of economic losses from cybersecurity incidents depends on organizational exposure, technological preparedness, governance and risk management, detection and response speed, industry characteristics, and the regulatory environment.
Framework-based qualitative analysis identifying interacting moderators of incident-loss severity.
Cost stickiness and total asset turnover operate as positive indirect channels linking digital transformation to profitability, partially offsetting a larger negative direct implementation effect in the short run.
Bootstrap parallel mediation analysis with 5,000 firm-clustered resamples and bias-corrected 95% confidence intervals; the paper states that the primary inferential specifications use one-year-lagged mediators.
The review finds that technology investment alone is insufficient to generate enterprise value from big data platforms and cloud-based analytics.
Structured integrative review of 30 peer-reviewed and policy sources published from 2020 to 2025, using thematic coding across platform architecture, analytics capability, adoption, governance, innovation, and Vision 2030 alignment.
The potential productivity benefits of AI vary with industrial composition, knowledge intensity, digital infrastructure, human capital, and broader AI readiness.
The paper cites BIS analysis covering 56 countries and 16 industries.
The effect of DGCI on regional total-factor carbon efficiency varies according to regional innovation capacity and tertiary-sector development.
Heterogeneity tests in the city-level panel examine whether the DGCI–TFCE relationship differs across regional innovation-capacity and tertiary-sector-development contexts.
Comparative advantage in services trade is increasingly determined by AI-adjusted task productivity rather than nominal wage differences alone.
The paper's theoretical framework, which incorporates wages, AI-assisted productivity, supervision costs, service quality, models, data, computing power, and organizational integration.
Standard productivity statistics may fail to capture gains from AI concentrated in safety, quality, downtime reduction, or tacit-process improvements.
The paper identifies measurement challenges as an implication of the qualitative findings; it does not present a statistical comparison of productivity measures.
The productivity-protective effect of digitalization is primarily an ex-ante preparedness effect: contemporaneous digitalization does not significantly affect current agricultural total factor productivity, whereas digitalization lagged by two periods has a significant positive effect.
Temporal-structure analysis using the unbalanced panel of Chinese listed agricultural firms; the paper compares contemporaneous and two-period-lagged digitalization measures.
The effect of two-period lagged agricultural-firm digitalization on agricultural total factor productivity is nonlinear with respect to climate-risk intensity: below a climate-risk threshold of approximately 18.96, the marginal effect is negative, while above the threshold it becomes positive.
Unbalanced panel data of Chinese A-share listed agricultural firms from 2007 to 2023; firm-level digitalization was constructed from annual-report text analysis and matched to city-level climate physical risk data. The study estimates nonlinear marginal effects and identifies a threshold.
AI deployment is at an early and uneven stage across leading publicly listed firms, with interest and investment outpacing realized benefits.
CIO interviews across leading international publicly listed companies and abductive analysis of differences in deployment outcomes.
Responsible workplace data governance has a positive but less stable association with Tobin's Q than with ROA.
Tobin's Q was analyzed as a secondary outcome using alternative timing structures and panel specifications.
In warehouse settings, algorithmic routing can improve coordination and measurable throughput while reducing method autonomy and intensifying work pressure.
Review synthesis of warehouse and logistics studies, including Cheon and Erickson (2025).
Across 15 frontier models, mean final net worth ranged from $20,856 to $188,488, representing a 9.0-fold difference.
Evaluation of 15 frontier models in the Business Arena marketplace; final net worth was the primary terminal outcome.
AI applications enhance enterprise productivity primarily by upgrading human capital through increasing the proportion of highly educated employees while displacing low-skilled workers.
The review summarizes an analysis by Yu and Qi (2025) of 3,646 Chinese A-share companies.
AI usage intensity alone does not guarantee performance improvement; complementary dynamic capabilities, data governance, and human capital are necessary for realizing performance gains.
The review reports findings from Rahmani et al. (2026) concerning AI usage intensity and organizational complements.
The systematic review found that AI adoption outcomes in SMEs are contingent on complementary organizational capabilities, knowledge-management infrastructure, human-capital quality, and institutional context rather than technology deployment alone.
Systematic review of 29 peer-reviewed Scopus-indexed articles using PRISMA-guided screening, bibliometric mapping, and qualitative content analysis.
Data platforms are likely necessary but not sufficient for capturing AI rents; returns also depend on organizational complementarities such as skills, governance, and processes.
Conceptual synthesis of findings on platform affordances, organizational capabilities, and uneven enterprise AI outcomes.
Evidence on the effects of modern data platforms on realized AI value is less consistent and less direct than evidence concerning analytics capability and organizational performance.
Comparative synthesis of reviewed studies and assessment of their empirical designs and outcome links.
Across seven examined cases, the paper argues that the capital gate is decisive because strong engineering results do not ensure that a firm can meet the self-financing test.
Comparative case analysis of seven frontier-technology cases using audited filings, prospectuses, and official industry statistics.
A Chinese quantum-computing firm reported 2025 net profit of 5.39 million yuan after four consecutive loss years, but net profit after removing non-recurring items was negative 43.80 million yuan.
Decomposition of the firm's reported profit using its financial disclosures; the company attributed part of the improvement to government grants and investment income.
Returns to AI investment are expected to be heterogeneous and path-dependent across institutions because they depend on organizational complementarities such as skills, leadership, and data architecture.
Implication derived from the capability-conversion framework; the paper distinguishes AI's technological affordances from the organizational HRA capability required to realize them.
The effect of digital-intelligent policy collaboration is strongest when green technological innovation exceeds a threshold and environmental regulation is of moderate intensity.
A staggered DID panel threshold model examining green innovation and environmental regulation as threshold or moderating conditions.
The green co-empowerment effect of digital-intelligent policies varies with regional characteristics and is shaped by talent agglomeration and intellectual property protection.
Moderation and heterogeneity analyses conducted within the staggered DID framework.
AI and generative AI may lower the costs of productization, personalization, and experimentation for digital ventures, but the gains depend on data access, human capital, and platform relationships.
Qualitative synthesis of literature on AI, generative AI, firm capabilities, data assets, human capital, and platform ecosystems; the paper identifies this as a research and economic implication rather than reporting a causal estimate.
AI investment is associated with higher firm sales, employment, market valuation, and product innovation, but AI-powered growth is concentrated among larger firms.
The paper summarizes firm-level evidence from Babina et al. (2024) and related organizational research.
The model represents organizational software output as Q(N,T) = A N^(α+β) T^β e^(−λT), where the exponential term captures cognitive friction from reviewing AI-generated output.
Theoretical production-function specification derived by substituting total token use with per-engineer token intensity; no empirical estimation is reported.
Raw proprietary data is a relatively weak competitive moat, whereas data embedded in a proprietary, continuously generated customer-feedback loop can be a stronger and more difficult-to-replicate complement.
Synthesis of competing theoretical and empirical arguments in the literature; the paper explicitly presents the raw-data versus feedback-loop distinction as its favored reconciliation.
The AI capability-to-competitive-advantage pathway is expected to be strong when complementary resources are available and weak or absent when those resources are scarce.
The proposed moderated-mediation framework treats complementary resources as gating mechanisms on the prior capability-to-advantage pathway rather than merely as antecedents of AI capability.
Gemini 3.1 Pro achieves a higher win rate than Opus 4.6 but earns lower total profit because its margin per win is lower.
Comparison of leaderboard performance metrics for Gemini 3.1 Pro and Opus 4.6.
AI-enabled process redesign could produce larger, non-marginal changes to task boundaries and work organization, increasing the potential for productivity gains while also increasing implementation risk and heterogeneity in realized returns across firms.
The paper's conceptual implications for firm behavior and productivity; not supported by a measured productivity effect or firm-level dataset.
The policy effect exhibits significant regional heterogeneity: it is most pronounced in the eastern region, followed by the western region, while the impact in the central and northeastern regions is not statistically significant.
Heterogeneity analysis by region reported in the paper using the 30-province panel (2012–2022); region-specific DID estimates compared across eastern, central, western, and northeastern regions.
A distinction must be made between the intensity of AI use and the maturity of AI use: while high intensity can facilitate short-term operational effects, sustainable differentiation typically emerges only at a high level of maturity (broad process integration, standardization, and scaling).
Argumentation grounded in literature review and conceptual reasoning; the paper cites empirical findings from various contexts but does not report a unified sample size.
The effectiveness of data & AI investments is critically dependent on context-specific, cluster-specific management strategy.
Study findings showing intercluster differentiation and varying outputs of management models, interpreted to imply cluster-specific management strategies are critical (methodological details not provided in excerpt).
Analysis of the digital maturity level with financial and operational key performance indicators of airlines has identified a considerable intercluster differentiation.
Comparative analysis of airlines' digital maturity levels alongside financial and operational KPIs (paper reports cluster analysis / intercluster comparison). Sample size not stated in excerpt.
Analysis of the digital maturity level with financial and operational key performance indicators of airlines has identified a considerable intercluster differentiation.
Comparative analysis of digital maturity levels and airlines' financial and operational KPIs reported in the paper (method details and sample size not provided in the excerpt).
These Big Data and Machine Learning applications have economic implications for productivity, revenue management, and new business formation in tourism.
Argument and synthesis drawn from the reviewed literature linking technical applications to economic outcomes (literature-based inference; no single study or sample size cited in the abstract).
Global survey evidence documents productivity gains alongside risks related to labor disruption, compute concentration, and uneven governance readiness.
Triangulation of case findings with secondary global survey evidence (surveys not specified in detail in the paper).
Effects of AI adoption on profitability vary systematically across ownership types, bank sizes, and policy cycles.
Heterogeneity analyses in the DID framework applied to the 17-bank panel (2009–2022) evaluating differences by ownership (e.g., state vs non-state), size, and policy-cycle periods.
Data-driven culture plays a dual role: it enables AI value creation as a mediator but, beyond a certain threshold, it constrains dynamic reconfiguration by limiting managerial discretion and strategic flexibility (the 'dark side' of DDC).
Interpretation of empirical mediation and nonlinear moderation results from PLS-SEM on the 254-respondent survey; authors explicitly describe this dual/enabling-and-constraining role.
Organizational data-driven culture (DDC) moderates the AIDC–performance relationship in a nonlinear way: excessive reliance on data weakens the marginal performance benefits of AIDC (threshold/diminishing-returns effect).
Moderation analysis reported in the paper using PLS-SEM on the n=254 survey sample; authors describe a contingent nonlinear moderating role of DDC with a weakening effect beyond a threshold.
Technological asymmetries (differences in access to advanced digital tools, AI capabilities and IT infrastructure) shape the financial stability and market performance of enterprises of various sizes.
Comparative analysis of 100 industrial joint-stock companies from multiple countries using multivariate regression models and index-based financial metrics (MC, EV, P/E, PEG, P/S, P/B, EV/R, EV/EBITDA).
By reframing advantage as architecture-dependent, SME performance becomes ecosystem-conditioned in AI-driven markets.
Conceptual conclusion drawn from the integrated model; authors argue theoretically that SME outcomes depend on ecosystem architectures rather than solely firm-level resources (no empirical validation reported).
The strategic advantages of the AI boom are not ubiquitous – they depend on other forms of 'institutional legitimacy and social license.'
Conceptual conclusion / synthesis in the paper arguing that institutional legitimacy and social license condition the distribution of AI advantages. No empirical sample or effect estimates provided in the excerpt.
The value of artificial intelligence (AIs) depends not just on firm-level capabilities but more so on national labour institutional compatibility (resource–institutional compatibility framework).
Theoretical development of a resource–institutional compatibility framework presented in the paper (conceptual argument). No empirical sample or quantitative test reported in the excerpt.
AI's effect on urban GTFE is heterogeneous across regions, city sizes, urban hierarchies, transportation-hub status, and old industrial base status.
Heterogeneity analysis conducted on the 2012–2021 panel of 279 Chinese cities comparing subgroups by region, city size, urban hierarchy, transportation hub/non-hub, and old industrial base/non-base.
Productivity gains vary widely across scenarios and countries and are substantially larger in countries with higher incomes.
Heterogeneity analysis across the compiled scenarios and the 31-country sample; authors report cross-country and cross-scenario variation and an income gradient (higher-income countries see larger gains).