Evidence (9875 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
9875 claims
Filtered →
Productivity
8807 claims
Filter claims →
Governance
7870 claims
Filter claims →
Human-AI Collaboration
7560 claims
Filter claims →
Org Design
4892 claims
Filter claims →
Innovation
4781 claims
Filter claims →
Labor Markets
4004 claims
Filter claims →
Skills & Training
3308 claims
Filter claims →
Inequality
2332 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 | 870 | 233 | 116 | 1066 | 2363 |
| Governance & Regulation | 976 | 451 | 218 | 133 | 1809 |
| Organizational Efficiency | 949 | 224 | 144 | 88 | 1416 |
| Technology Adoption Rate | 764 | 287 | 141 | 122 | 1325 |
| Research Productivity | 501 | 152 | 74 | 362 | 1101 |
| Output Quality | 542 | 216 | 69 | 69 | 896 |
| Decision Quality | 387 | 198 | 94 | 54 | 740 |
| Firm Productivity | 513 | 67 | 101 | 27 | 714 |
| AI Safety & Ethics | 249 | 303 | 73 | 36 | 667 |
| Market Structure | 190 | 192 | 134 | 27 | 548 |
| Task Allocation | 243 | 77 | 91 | 36 | 452 |
| Innovation Output | 291 | 33 | 55 | 20 | 401 |
| Skill Acquisition | 206 | 72 | 65 | 21 | 364 |
| Employment Level | 133 | 63 | 115 | 22 | 335 |
| Fiscal & Macroeconomic | 153 | 79 | 52 | 32 | 323 |
| Task Completion Time | 206 | 37 | 12 | 15 | 272 |
| Firm Revenue | 179 | 52 | 29 | 5 | 266 |
| Consumer Welfare | 130 | 76 | 47 | 13 | 266 |
| Inequality Measures | 48 | 137 | 51 | 6 | 242 |
| Worker Satisfaction | 101 | 81 | 25 | 13 | 220 |
| Error Rate | 84 | 110 | 11 | 5 | 210 |
| Wages & Compensation | 98 | 47 | 30 | 10 | 185 |
| Regulatory Compliance | 88 | 73 | 17 | 7 | 185 |
| Automation Exposure | 66 | 64 | 33 | 16 | 182 |
| Team Performance | 105 | 29 | 30 | 11 | 176 |
| Training Effectiveness | 109 | 22 | 14 | 21 | 168 |
| Developer Productivity | 114 | 21 | 14 | 8 | 158 |
| Job Displacement | 12 | 90 | 24 | 1 | 127 |
| Hiring & Recruitment | 57 | 9 | 9 | 5 | 80 |
| Skill Obsolescence | 6 | 56 | 9 | 1 | 72 |
| Social Protection | 43 | 17 | 8 | 2 | 70 |
| Creative Output | 35 | 21 | 9 | 4 | 70 |
| Labor Share of Income | 18 | 21 | 17 | 1 | 57 |
| Worker Turnover | 15 | 16 | — | 4 | 35 |
| Industry | — | — | — | 1 | 1 |
Adoption
Remove filter
Policy design should be adaptive and sector-sensitive, balancing innovation with safeguards while targeting skills, infrastructure, and inclusive finance to maximize social returns from SME AI adoption.
Policy recommendations derived from the literature review and identified cross-cutting barriers/enablers; these are prescriptive rather than empirically validated within the review.
Innovative financing (blended finance, pay-per-use, outcome-linked financing) is critical to overcome upfront cost barriers and enable scalable, risk-sharing investments in AI for SMEs.
Policy reports and selective case studies in the review demonstrating these instruments can facilitate uptake; systematic evidence on scalability and impact remains limited.
Developing pragmatic, locally appropriate data governance arrangements (standards, privacy safeguards, data trusts) is necessary to build trust and enable SME participation in data-driven markets.
Policy literature and governance proposals reviewed; examples of data-governance models (e.g., data trusts, federated learning) discussed, but empirical evaluations in LMIC SME contexts are scarce.
Implementing scalable financing and procurement models (pay-as-you-go, leasing, blended finance) can overcome upfront cost barriers for SMEs adopting AI.
Policy and finance reports and a small number of case examples cited in the review showing such instruments enabling technology uptake; systematic evidence on effect sizes is limited.
Strengthening ecosystem linkages among academia, tech providers, financiers, and regulators enhances the prospects for inclusive, scalable AI adoption by SMEs.
Case studies and ecosystem analyses in the reviewed literature that document positive roles for partnerships and coordinated support; evidence is descriptive and context-dependent.
Incremental investment in human capital and development of dynamic capabilities (learning, adaptation) increases SMEs’ absorptive capacity and the likelihood of successful AI adoption.
Theoretical grounding in RBV and DC literature combined with illustrative case evidence from the review showing firms with stronger learning capabilities tend to adopt and benefit more from technology.
A phased adoption approach (assess needs → pilot low-risk use cases → scale modularly) is recommended to reduce risk and improve outcomes for SME AI projects.
Synthesis of best-practice guidance and pragmatic recommendations from case studies and policy literature; not empirically validated as a universal causal strategy in LMIC SMEs within the review.
External market pressures and customer demand often drive AI adoption decisions in SMEs.
Surveys and market analyses from the literature indicating demand-side pressures as adoption triggers; evidence mainly observational.
Access to finance, including scalable and blended financing models, is a key enabler for SME AI adoption.
Policy reports, case studies and financial analyses discussed in the review that identify financing availability and instrument design as central constraints/enablers; evidence is descriptive and context-dependent.
Local innovation ecosystems (universities, incubators, private-sector partnerships) support SME uptake of AI.
Case studies and ecosystem analyses in the reviewed literature documenting successful university–industry linkages and incubator support facilitating technology transfer and skills development.
Supportive government policy and adaptive regulation are important enablers of AI adoption among SMEs.
Synthesis of policy reports and governance literature included in the review identifying regulatory clarity and supportive policy as common enabling factors.
AI can improve market access for SMEs (e.g., via digital platforms and AI-enabled credit scoring) and enable potential value-chain upgrading.
Policy analyses and case-study evidence showing digital platforms and algorithmic credit assessment opening opportunities for SMEs; examples referenced from Botswana and similar LMIC contexts.
AI adoption supports new product/service innovation and faster time-to-market for SMEs.
Qualitative case studies and practitioner reports cited in the review showing instances of AI assisting R&D, prototyping, and launch processes; limited systematic quantitative measurement across sectors.
AI-enabled customer segmentation and personalization can increase sales and customer retention for SMEs.
Empirical examples and case studies from the literature and policy reports documenting improved targeting and retention in firms that adopted AI tools; evidence is largely observational and context-specific.
AI can generate productivity gains for SMEs through automation and process optimization.
Multiple case studies and firm-level surveys reported in the literature showing examples of automation-related efficiency improvements; no large-scale randomized or causal studies cited that uniformly quantify effect sizes across LMIC SMEs.
Anticipatory analytics and automated decision support can improve public resource allocation and reduce response lag, raising public sector productivity and potentially changing demand for private sector services.
Aggregate claims from empirical cases and theoretical pieces in the review that report or argue for efficiency/productivity gains from predictive systems; synthesis across several studies in the 103‑item corpus.
Realizing economic and social benefits from public‑sector AI requires interoperable, ethical‑by‑design systems combined with sustained investments in skills, infrastructure, and accountability mechanisms.
Prescriptive synthesis from the systematic review that aggregates recommendations across empirical studies and institutional reports within the 103‑item corpus.
Big Data and AI are enabling a shift in public governance from reactive to anticipatory decision-making and resource allocation.
Synthesis from a PRISMA-guided systematic review of 103 peer‑reviewed articles and institutional reports (2010–2024) mapping empirical cases of predictive analytics and AI deployment in public-sector domains.
RAG approaches (cloud or on-prem) outperform a zero-shot baseline (base model without retrieval) on retrieval/generation performance.
Empirical comparative experiments included a zero-shot base model baseline, GPT RAG cloud, and on-prem RAG; summary implies comparative superiority of RAG over zero-shot but does not provide exact metrics or sample sizes.
On-prem solutions simplify compliance with data sovereignty and privacy regulations (e.g., GDPR) and reduce legal risk for firms handling sensitive IP.
Policy-relevant assessment in environment/security evaluation arguing on-prem architectures ease regulatory compliance; no legal-case study evidence provided in summary.
Converting variable token/API costs into fixed on-prem costs can lower marginal cost per query for sustained, high-volume usage typical of some SMEs.
Economic/cost-structure analysis in the paper arguing that capex + ops converts variable to fixed costs and reduces marginal cost per query for sustained usage; no numeric break-even analyses reported in summary.
On-prem deployment materially improves data sovereignty and reduces risk of external data leakage.
Environment/security evaluations including threat/surface analysis and policy assessment arguing that on-prem architectures prevent external transmission of sensitive data; no empirical breach incidence data provided.
On-Premise RAG eliminates recurring token/API costs associated with cloud LLMs, reducing long-run OPEX.
Organizational cost accounting comparison between recurring cloud/API expenses and on-prem capital and operational costs presented in the TOE-grounded analysis; no dollar amounts or time horizons reported in summary.
On-Premise RAG outperforms commercial RAG on qualitative dimensions (usefulness and relevance) in specialized manufacturing domains.
Human evaluation by domain experts (human-in-the-loop judgments) assessing usefulness and relevance using the on-prem pipeline with a curated knowledge base; sample size and scoring protocol not specified in summary.
Market failures—data externalities, coordination failures, and large fixed costs for sensorization/computing—likely lead to underinvestment by private actors and justify targeted public interventions (data platforms, co-financing, standards).
Economic reasoning informed by observed underinvestment patterns in investment datasets and the structure of costs for sensorization/computing; institutional review indicating coordination gaps.
Institutional determinants (data governance, standards, public infrastructure) materially influence AI diffusion and should be incorporated explicitly into diffusion models alongside human capital and capital-cost channels.
Cross-country trend comparisons and institutional analysis demonstrating correlations between institutional variables and adoption/diffusion patterns; theoretical synthesis.
Workers are increasingly treating AI adoption as a collective bargaining and political issue, using strikes, bargaining demands, and internal organizing to contest deployments.
Synthesis of reports, case studies and contributions to the AIPOWW symposium documenting worker organizing episodes and demands related to AI deployments; no systematic dataset or sample size reported.
Policy recommendations include investing in workforce reskilling, promoting interoperability and data portability, designing proportional risk-based regulation, using regulatory sandboxes and staged deployment, and supporting capacity building for low- and middle-income countries to avoid an AI divide.
Synthesis of policy analysis, sectoral findings and normative recommendations derived from the comparative review and gap analysis.
AI adoption can raise firm- and sector-level productivity, potentially lifting aggregate output; measuring AI’s contribution requires new indicators of 'AI intensity'.
Economic reasoning and review of literature; recommendation for measurement approaches (software/hardware investment, AI talent, use of AI services). No primary empirical measurement provided.
Regulatory design should be context-sensitive and ethics-grounded rather than one-size-fits-all.
Normative evaluation and synthesis of governance frameworks and identified gaps across jurisdictions; policy recommendations grounded in ethical principles (transparency, fairness, accountability, human rights).
AI capabilities (learning, reasoning, perception, NLP) are being integrated rapidly across healthcare, finance, education, transportation, security and justice, producing major efficiency and service-quality gains.
Sectoral case studies and documented examples cited in policy/regulatory texts and secondary literature; comparative analysis of deployments across the listed sectors.
AI is driving large productivity and capability gains across sectors.
Synthesis of sectoral case studies and secondary literature across healthcare, finance, education, transportation, security and justice; comparative policy and regulatory analysis of documented AI deployments. No large-scale primary quantitative impact evaluation reported.
Environmental-performance labeling and user opt-outs could create demand for 'eco-optimized' models and influence competition among providers.
Market analysis in implications section (theoretical consumer preference/differentiation effects).
Mandatory inference benchmarks and public reporting would create market and regulatory incentives to optimize models for energy efficiency (e.g., compression, routing, edge inference).
Policy implications / market design analysis describing likely provider responses to benchmarking and public reporting.
Mandatory model-level disclosure and user-choice rights would help internalize negative environmental externalities, shifting costs into firms’ deployment and pricing decisions.
Economic-policy analysis in the implications section (conceptual/incentive reasoning based on disclosure->price/internalization mechanisms).
The paper recommends international coordination to prevent regulatory arbitrage and ensure consistent standards for model-level environmental governance.
Policy design and cross-jurisdictional analysis arguing for harmonization to avoid compute relocation/obfuscation and regulatory gaps.
Investors and regional planners can use the Hub to identify emerging opportunity hubs and prioritize economic development or infrastructure to support skill formation.
Implications and use-case examples in the paper proposing the Hub's application for regional strategy and investment decisions; empirical evidence for realized investment outcomes is not provided.
Policy-simulation features make it possible to compare labor-market effects of alternative interventions (subsidies, regulations, training programs) before deployment.
Description of policy simulation dashboards and scenario-analysis capabilities in Methods and Implications sections; no quantitative validation details provided in the summary.
Geospatial hotspot identification enables region-specific training investments and curricula alignment with projected demand.
Implications section connects geospatial hotspot outputs to targeted reskilling/education policy; empirical effectiveness of doing this is implied by experimental claims but not quantitatively substantiated in the summary.
The Hub supports more targeted, data-driven workforce and policy decisions by producing actionable, interpretable outputs and scenario comparisons.
Paper's Main Finding and Implications sections arguing that outputs enable targeted reskilling, policy design, and regional strategy. Empirical support is claimed via an experimental evaluation but detailed results are not reported in the summary.
Experimental evaluation shows the Hub can quantify how automation and policy interventions alter future workforce readiness.
Paper describes scenario analysis and reports that the system quantifies impacts of automation and policy in experiments, but does not provide numeric results, evaluation methodology, or datasets in the provided summary.
Experimental evaluation shows the platform can pinpoint high-potential regional opportunity hubs.
Paper claims experimental results demonstrate ability to highlight regional opportunity hubs; evaluation details (data sources, sample size, metrics) are not provided in the summary.
Experimental evaluation shows the system can identify critical talent shortages.
Paper reports an experimental evaluation that the platform can surface critical shortages; no datasets, sample sizes, numerical metrics, or evaluation design details are reported in the abstract/summary.
Hybrid approaches may deliver the best economic return by reducing need for large-scale primary data collection while maintaining acceptable performance, but they require modest real-data collection costs for fine-tuning and validation.
Inferred from comparative evaluations and economic reasoning in the reviewed literature that contrast synthetic-only, real-only, and hybrid strategies; evidence is suggestive rather than pooled quantitative analysis.
Hybrid datasets (synthetic data combined with real patient data) consistently yield better model performance than synthetic-only training across reviewed studies.
Critical literature review and thematic synthesis of machine-learning evaluation studies reported in peer-reviewed articles, technical reports and policy analyses across searched databases (Scopus, Web of Science, PubMed, Google Scholar). The review reports a recurring pattern across multiple studies, though the number of studies and exact effect sizes are heterogeneous and not enumerated in the paper; limitations noted include publication bias and heterogeneity in outcome metrics.
International certification protocols tied to explainability and safety standards would influence investment incentives and market structure.
Policy and economic analyses in the literature synthesis arguing how standards/certification shape firm behavior and investment; no empirical causal estimation provided.
A tiered risk-management framework that allocates governance intensity to interventions by clinical criticality and autonomy is recommended to maximize benefits while containing harms.
Authors' policy recommendation derived from literature synthesis of governance frameworks, risk analyses, and implementation studies; prescriptive rather than empirically validated in large-scale trials.
Federated learning and privacy-preserving collaboration can combine data advantages without centralizing sensitive records and may reduce duplicated validation costs over time.
Technical literature and pilot studies on federated learning and privacy-preserving methods summarized in the paper; limited large-scale, long-term deployment evidence noted.
Centralized updates and monitoring by vendors can reduce operational burden for healthcare providers.
Comparative analyses and deployment reports contrasting vendor-managed services with self-managed open-source deployments; synthesized evidence and stakeholder commentary.
Open-source models enable customization and local retraining that can align models with institutional workflows and patient populations.
Cross-disciplinary literature synthesis and case reports describing local retraining/customization practices; comparative analyses of model adaptability. Evidence is drawn from diverse deployments rather than controlled trials.