Evidence (8974 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
10085 claims
Filter claims →
Productivity
8974 claims
Filtered →
Governance
8062 claims
Filter claims →
Human-AI Collaboration
7749 claims
Filter claims →
Org Design
5057 claims
Filter claims →
Innovation
4896 claims
Filter claims →
Labor Markets
4088 claims
Filter claims →
Skills & Training
3372 claims
Filter claims →
Inequality
2377 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 | 882 | 244 | 117 | 1097 | 2424 |
| Governance & Regulation | 1010 | 469 | 229 | 135 | 1875 |
| Organizational Efficiency | 977 | 235 | 149 | 90 | 1462 |
| Technology Adoption Rate | 781 | 299 | 143 | 128 | 1362 |
| Research Productivity | 506 | 155 | 74 | 363 | 1110 |
| Output Quality | 555 | 219 | 71 | 70 | 915 |
| Decision Quality | 395 | 200 | 95 | 54 | 751 |
| Firm Productivity | 523 | 67 | 101 | 27 | 724 |
| AI Safety & Ethics | 262 | 309 | 75 | 36 | 688 |
| Market Structure | 195 | 201 | 135 | 30 | 566 |
| Task Allocation | 248 | 77 | 96 | 38 | 464 |
| Innovation Output | 300 | 34 | 55 | 20 | 411 |
| Skill Acquisition | 207 | 75 | 65 | 21 | 368 |
| Employment Level | 138 | 67 | 119 | 24 | 350 |
| Fiscal & Macroeconomic | 156 | 80 | 53 | 33 | 329 |
| Task Completion Time | 211 | 38 | 13 | 16 | 280 |
| Firm Revenue | 183 | 52 | 29 | 5 | 270 |
| Consumer Welfare | 131 | 77 | 48 | 13 | 269 |
| Inequality Measures | 50 | 141 | 54 | 9 | 254 |
| Worker Satisfaction | 104 | 85 | 25 | 13 | 227 |
| Error Rate | 87 | 112 | 11 | 5 | 215 |
| Automation Exposure | 69 | 69 | 37 | 20 | 198 |
| Wages & Compensation | 102 | 49 | 31 | 11 | 193 |
| Team Performance | 115 | 30 | 30 | 11 | 187 |
| Regulatory Compliance | 88 | 74 | 17 | 7 | 186 |
| Training Effectiveness | 109 | 22 | 14 | 21 | 168 |
| Developer Productivity | 116 | 21 | 15 | 8 | 161 |
| Job Displacement | 12 | 92 | 26 | 1 | 131 |
| Hiring & Recruitment | 57 | 12 | 9 | 5 | 83 |
| Skill Obsolescence | 6 | 59 | 10 | 2 | 77 |
| Social Protection | 43 | 17 | 8 | 2 | 70 |
| Creative Output | 35 | 21 | 9 | 4 | 70 |
| Labor Share of Income | 18 | 23 | 17 | 1 | 59 |
| Worker Turnover | 15 | 16 | — | 4 | 35 |
| Industry | — | — | — | 1 | 1 |
Productivity
Remove filter
Systemic alignment between AI algorithms' technological potential, organizational support, and the human factor is necessary to convert AI-driven benefits into sustainable economic growth.
Practical recommendations derived from the theoretical framework, analysis of micro-mechanisms, and observed mismatch between micro gains and macro productivity trends.
Localized micro-level effects of AI implementation include accelerated skill acquisition by employees.
Corporate case analyses and empirical reports indicating faster learning/skill uptake associated with AI-assisted workflows, as synthesized in the paper.
Localized micro-level effects of AI implementation include enhanced quality of generated solutions.
Evidence drawn from corporate cases and empirical reports described in the study showing improvements in output quality when AI tools are used.
Localized micro-level effects of AI implementation include reduced operational task execution time.
Corporate case analyses and empirical reports cited in the paper documenting reductions in task execution time after AI tools are deployed.
The advent of an "artificial intelligence phase" in the US macroeconomic productivity dynamics is justified (identified as a new phase following earlier cycles).
Periodization based on calculated AAPC indices from BLS aggregated productivity series and comparative retrospective analysis of macroeconomic cycles.
Practically, organizations should embed AI technologies in HR systems that foster learning, knowledge utilization, and continuous innovation.
Practical recommendation offered by authors grounded in their empirical results (survey of 750 responses showing mediating roles of AI self-efficacy and digital HRM practices).
The study advances knowledge management theory by highlighting the complementary roles of individual cognitive beliefs (AI self-efficacy) and HR systems (digital HRM practices) in enabling AI-driven learning and capability development.
Author-stated theoretical contribution based on integration of empirical findings and theory; interpretation and framing provided in the paper.
Digital HRM practices function as a significant positive mediator that helps translate AI adoption into enhanced innovation performance.
Mediation analysis reported in the paper using survey data (750 valid responses); paper explicitly states digital HRM practices mediate the AI usage → innovation performance relationship.
Digital HRM practices function as a significant positive mediator that helps translate AI adoption into enhanced employee job performance.
Mediation analysis reported in the paper using survey data (750 valid responses); paper explicitly states digital HRM practices are a significant mediator between AI usage and work performance.
AI self-efficacy functions as a significant positive mediator that helps translate AI adoption into enhanced innovation performance.
Mediation analysis reported in the paper using survey data (750 valid responses); paper explicitly states AI self-efficacy mediates the AI usage → innovation performance relationship.
AI self-efficacy functions as a significant positive mediator that helps translate AI adoption into enhanced employee job performance.
Mediation analysis reported in the paper using survey data (750 valid responses); paper explicitly states AI self-efficacy is a significant positive mediator between AI usage and work performance.
AI usage contributes to enhanced innovation performance.
Cross-sectional survey analysis of 750 responses; paper reports positive association between AI adoption/usage and organizational/employee-level innovation outcomes.
AI usage contributes to improved employee job performance.
Cross-sectional survey analysis of 750 responses; paper reports positive association between AI adoption/usage and employee job performance (analysis described as testing mediation via cognitive and HR mechanisms).
Dynamic retrieval architectures are structurally insulated from the combinatorial collapse in task-success probability that afflicts manual attachment approaches.
Comparative theoretical analysis within the paper contrasting model behavior under manual attachment versus dynamic retrieval architectures (described analytically, based on the probabilistic model).
Using agriculture as a revealing template, the paper shows how embodied AI reshapes business models in traditional industries, moving them from product performance toward continuous workflow optimization, lifecycle-based orchestration, and recurring, trust-based monetization.
Conceptual application/example stated in the abstract (sectoral illustrative case in agriculture; theoretical case analysis rather than empirical study).
The paper derives nine propositions specifying how embodied AI transforms business models.
Count of theoretical propositions stated in the abstract (conceptual contribution).
The authors develop two complementary models: (1) a transition model showing how business models shift from asset-based and episodic logics toward adaptive, data-driven systems; and (2) an integrative embodied AI business model grid explaining how changes are generated through reconfiguration of value activities, interdependencies, and governance across actors and technologies.
Paper's stated theoretical contribution in the abstract (development of two conceptual models).
Firms must redesign what activities are performed, how they are linked, and who controls them when adopting embodied AI.
Conceptual proposition in the abstract arguing organizational redesign is required (theoretical reasoning).
Embodied AI participates in operations and moves beyond supporting decision-making to become a constitutive element of value creation.
Theoretical/analytical claim presented in the abstract as part of the paper's argument (conceptual analysis).
Embodied Artificial Intelligence (AI) refers to AI systems in which intelligence is embedded in physical systems and emerges through interaction with its environment, acting through continuous cycles of sensing, decision-making, actuation, and learning.
Conceptual/definitional statement made in the paper's abstract (theoretical definition).
The paper provides novel causal evidence on how discretion design in human-supervised AI affects decision quality.
Randomized field experiment with causal identification via random assignment to override regimes; analyses of primary outcomes and LATE for selection effects.
Gains are largest for experienced workers, high-incentive SKUs, and growth-stage SKUs.
Heterogeneous treatment effect analysis in the randomized experiment showing larger effects in subgroups defined by worker experience, SKU incentive level, and SKU growth stage.
Under constrained overrides, workers select better SKUs to override, confirmed via local average treatment effects.
Analysis using local average treatment effects (LATE) reported in paper showing that constrained policy leads to selection of higher-value SKUs for overrides.
Constrained overrides reduce inventory by 1.28% without harming sales.
Randomized field experiment comparing constrained-overrides arm (two-per-machine downward limit) to control and/or free overrides; reported 1.28% inventory reduction and no statistically significant negative effect on sales.
We propose a constrained override policy that limits overrides per decision episode to enable selective filtering that prioritizes high-value overrides.
Conceptual/methodological proposal in the paper; motivates the experimental intervention (no direct quantitative evidence presented for the proposal itself aside from subsequent experimental test).
A design oriented prompt raised visual quality, 4.5 versus 3.0 on a 5 point scale, without lifting function, and a one paragraph paraphrase of its directive reproduced the entire lift.
Comparison of visual quality scores across prompt conditions reporting mean visual quality 4.5 vs 3.0 (5-point scale) with no change in functional score; paraphrase reproduced effect.
Raising reasoning effort from High to xHigh lifted first try perfect runs from 28 percent to 89 percent and cut corrective prompts about five fold, for 9 to 29 percent more cost.
Experimental comparison of two reasoning effort levels reporting first-try perfect run rates (28% → 89%), reduction in corrective prompts (~5x reduction), and cost increase (9–29%).
Capability tier dominated: frontier models clustered near the ceiling while a low cost local model fell to 24 to 37 points.
Comparison of aggregate scores across model capability tiers reported in the study; reported score range for the low-cost local model (24–37 points out of 42).
The paper recommends that the regulatory authority in Nigeria design guidelines for AI adoption in banks.
Policy/recommendation stated in the paper's conclusions, offered in response to the observed positive associations between AI investments/automation and bank ROA.
The paper recommends that banks in Nigeria invest more in AI infrastructure and invest in staff capacity in AI technologies.
Policy/recommendation stated in the paper's conclusions, based on the reported empirical associations between AI-related investments/activities and ROA.
AI-structured investments have a significant impact on asset utilization in Nigeria.
Synthesis/conclusion drawn from the empirical results (positive and significant associations between AI investment/disclosure/automation measures and ROA) using secondary data from ten purposively selected DMBs listed on the Nigerian Exchange Group for 2015–2024.
Bank size has a positive impact on the return on assets (ROA) of banks in Nigeria.
Empirical analysis of secondary data (ex post facto design) covering ten purposively selected DMBs listed on the Nigerian Exchange Group from 2015 to 2024; bank size was included as an explanatory/control variable and found to have a positive association with ROA in the paper's analysis.
AI operational automation using chatbots has a positive and significant impact on the return on assets (ROA) of banks in Nigeria.
Empirical analysis using an ex post facto research design on secondary data from ten purposively selected DMBs listed on the Nigerian Exchange Group for the period 2015–2024. The paper reports that measures of operational automation (chatbot use) are positively and significantly associated with ROA.
AI software expenses disclosure has a positive and significant impact on the return on assets (ROA) of banks in Nigeria.
Empirical analysis using an ex post facto research design on secondary data from ten purposively selected DMBs listed on the Nigerian Exchange Group for the period 2015–2024. The paper reports a positive and statistically significant association between disclosed AI software expenses and ROA.
AI software book value has a positive and significant impact on the return on assets (ROA) of banks in Nigeria.
Empirical analysis using an ex post facto research design on secondary data from ten purposively selected deposit money banks (DMBs) listed on the Nigerian Exchange Group for the period 2015–2024. The paper reports that AI software book value is positively and significantly associated with ROA (regression results reported in the paper).
Adoption restructured code review around automation: per-reviewer load roughly doubled.
Counts of review actions per reviewer over time reported in paper; authors state per-reviewer load approximately doubled after adoption/mandate.
A staggered difference-in-differences design links the within-developer share of this gain to AI adoption and to a further gain that grows with accumulated use.
Quasi-experimental analysis using a staggered difference-in-differences design on the developer-level panel; within-developer variation in AI adoption and cumulative usage examined.
Per-capita throughput eventually doubled, reaching 2.09x the pre-mandate baseline in April 2026.
Outcome measurement of merged pull requests per engineer over time (panel analysis); reported multiplier (2.09x) at April 2026 relative to pre-mandate baseline.
We study a documented enterprise "2x" mandate at a mid-sized, AI-forward company that has been committed to doubling merged pull requests per engineer since mid-2025.
Paper describes a single documented case study (one mid-sized firm) with an internal mandate beginning mid-2025; qualitative documentation of the mandate reported in methods.
AI adoption is likely to have a positive effect on labour productivity in the United States, but the magnitude will depend on broad diffusion, responsible governance, reskilling, and effective integration into real production processes.
Paper's concluding synthesis of secondary macro/micro evidence and recent experimental studies.
Recent experiments (published from 2020 onward) show strong task-level productivity gains, including faster writing, improved customer support performance, and quicker software development.
Paper cites experimental research from 2020+ reporting task-level improvements in writing speed, customer support metrics, and software development tasks.
AI enables faster knowledge processing.
Conceptual assertion in the paper supported by secondary evidence and experimental studies about information/knowledge tasks.
AI supports software development, enabling quicker software development.
Paper cites recent experiments (from 2020 onward) showing faster software development when using AI tools.
AI adoption can automate routine cognitive tasks.
Conceptual claim in paper, supported by secondary literature on task automation and cited experimental work.
AI adoption can improve worker decision-making.
Paper's conceptual synthesis and references to experimental research indicating decision support benefits.
AI adoption can raise labour productivity by reducing task completion time.
Conceptual argument in paper supported by recent experimental research (studies from 2020 onward) showing faster task completion in specific tasks.
AI is increasingly used in software development, customer service, professional writing, data analytics, health services, logistics, finance, and other knowledge-intensive activities.
Reported in paper based on secondary evidence from multiple sources (listed above).
Artificial intelligence (AI) adoption has become one of the most important economic changes in the United States.
Statement in paper supported by secondary literature synthesis (U.S. Census Bureau, BLS, OECD, IMF, Stanford AI Index, McKinsey Global Institute, NBER).
Wages of labor that is essential for building AI increase faster than overall GDP.
Analytical economic model / comparative statics showing relative wage growth for AI-building labor. No empirical sample reported.
The article contributes to organisation studies by theorising agentic AI as an emerging object of organising and by specifying the interface conditions under which human and agentic organisational behaviour can jointly support collective intelligence.
Author's stated contribution in the paper (conceptual/theoretical contribution). Method: synthesis/theorisation; no empirical quantification provided in excerpt.