Technology Adoption Rate

Updated Apr 06, 2026

Papers 279 (87 full-text)

Claims 564

Evidence strength: Mixed — RCTs show near-term causal levers; most sectoral and macro evidence is observational and context-specific.

Bottom Line

Brief, hands-on training increases voluntary use of generative AI; communicating uncertain privacy risks and adding AI labels reduce uptake and engagement Chen and Bao (2026); Erlei et al. (2026); Seeger et al. (2026). Adoption is uneven across sectors and regions; gaps in infrastructure, institutional capacity, and organizational resources are repeatedly associated with slower uptake in public services, healthcare, and smallholder contexts Vallejo Manzur and Álvarez-Aros; Nyamawe and Shao; Axmedov (2026); Axmedov (2026).

Pair access with short, hands-on training; it increases voluntary use, mainly by bringing in new users Chen and Bao (2026).
Avoid wide privacy-risk ranges; ambiguity lowers adoption more than a clear, even high, point estimate Erlei et al. (2026).
Do not rely on transparency alone; build trust first and target segments where trust is already higher if you need real sharing, not just stated intent Rosenberger et al. (2026).
Expect labels to reduce engagement; late disclosure helps only for AI-enhanced (not fully AI-generated) content—set labeling and timing with this in mind Seeger et al. (2026).
To scale beyond pilots, fund the enablers: infrastructure, change management, skills, and governance where evidence links these gaps to stalled adoption Vallejo Manzur and Álvarez-Aros; Liu; Nyamawe and Shao.

A brief, targeted training increased voluntary use of large language models (LLMs) from 26% to 41% in a randomized trial with law students, mainly by bringing in new users Chen and Bao (2026).
Ambiguous data-leak ranges (e.g., “10–50%”) reduced adoption of AI personalization versus neutral framing, while a clear 30% leak probability did not; shown in a randomized experiment across sensitive and anonymized data labels Erlei et al. (2026).
Labeling content as AI-generated or AI-enhanced reduced emotional and behavioral engagement in two online experiments; late disclosure improved emotional engagement only for AI-enhanced content Seeger et al. (2026).
In a pre-registered online experiment, transparency alone did not increase actual data sharing; it raised stated willingness only among users with higher pre-existing trust, and immediate sharing seemed driven by fast, intuitive choices (stated vs. actual behavior diverged) Rosenberger et al. (2026).

Reviews of HR analytics cite privacy, compliance risk, bias and opacity, and weak change management as adoption barriers Taduvana et al. (2026).
In procurement, insufficient organizational resources were associated with lower AI adoption in survey data (β = -0.19, p < 0.05) Liu.
Public-sector and low- and middle-income country (LMIC) settings face infrastructure gaps, weak data ecosystems, and talent-retention issues that are associated with slower AI adoption in governance Vallejo Manzur and Álvarez-Aros.
Healthcare AI in Africa faces limited infrastructure, scarce local compute, weak synthetic-data regulation, and clinician skepticism about validity, based on descriptive and policy evidence Nyamawe and Shao.
Smallholder agriculture adoption of digital and precision tools is constrained by capital, infrastructure, skills, credit, tenure security, and sociocultural factors, across multiple synthesized studies Axmedov (2026); Axmedov (2026).
Case evidence from Indonesian oil refining points to infrastructure gaps, weak data governance, skills shortages, high upfront costs, and organizational inertia as drivers of a “pilot trap” that blocks scaling Muljono et al..

Within countries, higher AI readiness is positively associated with e-government development; models tracking within-country changes show a coefficient of 0.17 (p < 0.001), with large country heterogeneity Spivakovskyy et al. (2026).
In China, green innovation is concentrated in coastal provinces with limited inland diffusion, constraining its impact on regional carbon inequality Fan (2026).
Across 277 Chinese cities, evidence is consistent with a stage-dependent path: mature robotics sectors are linked to greater robot adoption, which is associated with higher energy efficiency and then lower emissions; this channel appears inactive at early stages Lin et al. (2026).
Cross-country and industry comparisons suggest unequal gains from adopting new technological capital, with early adopters and capital-rich settings benefiting more Mici et al..
For European firms, AI exposure correlates with exporting to most destinations except China; big data analytics and blockchain adoption show no significant correlation with exporting in statistical models Fikirli and Şahin (2026).

A randomized field experiment that helped startups map AI into production increased discovered AI use cases by 44% and raised the likelihood of acquiring paying customers by 11 percentage points Kim et al. (2026).
A case study in automotive software development reported that a graph-based, LLM-powered workflow approach supported incremental adoption without disrupting existing practices Wang et al. (2026).
Observational evidence shows users create and consume AI-generated content differently from human-generated content in online ecosystems Shi et al. (2026).
In the near term, theory points to “bounded autonomy”: AI as supervised copilots and constrained executors in financial markets Gong (2026).

Long-run, real-world effects of disclosure and labeling policies on customer behavior; current evidence is from online experiments, not at-scale deployments Seeger et al. (2026).
Whether transparency and trust-building produce durable changes in actual sharing and system use over weeks or months, given immediate choices were intuitive in a one-shot task Rosenberger et al. (2026).
Which policy bundles best diffuse AI from leading to lagging regions; current diffusion evidence is descriptive and location-specific Fan (2026).
In public services and LMIC healthcare, large-scale deployment studies that quantify adoption pathways and outcomes under real governance constraints are scarce Nyamawe and Shao; Vallejo Manzur and Álvarez-Aros.
How heterogeneous transition dynamics play out at economy-wide scale—who adopts when, who gains, and how spillovers evolve—needs causal identification beyond current comparative and model-based analyses Mici et al..