A Systemic Analysis of the Structural Drivers Enabling Its Development in Latin America
A
A
1. INTRODUCTION
Over the last years, artificial intelligence (AI) has moved beyond its original technological boundaries and has become a transversal component of economic systems, public administration and everyday life. The rapid expansion of generative AI since 2022 accelerated the adoption of national and institutional strategies that seek to guide its integration. Although these initiatives are well intentioned, they often treat AI adoption as an independent objective and not as the outcome of a broader ecosystem of pre-existing capacities. When strategies overlook the structural conditions that sustain AI development, they can become technically sophisticated yet operationally fragile and may lead to recommendations that do not align with the real architecture required for a country to achieve digital maturity.
In Latin America this challenge is particularly notable. The region has entered a phase in which the formulation of national AI strategies is no longer sufficient, and governments are now producing more specific plans for education, agriculture, health, justice, security and public management. This transition indicates an evolving policy environment and highlights an important conceptual gap. AI cannot be sustained only through algorithms, pilot projects or technical declarations. It depends on deeper structural foundations that determine whether those initiatives can endure and scale. When these foundations are not aligned, the risk is not an immediate failure but the consolidation of fragile ecosystems that constrain future progress.
Understanding AI as an emergent capability rather than an isolated technology requires recognizing a continuum of underlying activities that enable its development. Among these are research efforts, prototyping, experimentation, talent formation and social appropriation. It also requires stable regulatory environments, inclusive digital infrastructures, technical capacity distributed across institutions and a collective understanding of the realistic potential and limits of AI. When these conditions are not coherently integrated, policies often prioritize the acquisition of external technologies instead of strengthening the pillars that make them sustainable. This can lead to dependence on foreign providers, higher operational costs, limited local adaptation and restricted national capacity to develop or modify AI systems.
This study adopts that systemic perspective. If AI maturity emerges from the interaction of several structural drivers, it must be analysed through indicators that represent those dimensions. The analysis integrates four international sources that capture key aspects of digital development: innovation as measured by the Global Innovation Index (GII), institutional and regulatory readiness as captured by the Global AI Readiness Index (GAIRI), digital infrastructure through the ICT Development Index and public sector digitalisation captured by the E-Government Development Index (EGDI). Using comparable data from seventeen Latin American countries, the study examines the extent to which these structural drivers explain the variability of the Latin American Artificial Intelligence Index (ILIA), currently the most consolidated regional indicator of AI maturity.
The results indicate that AI maturity increases when these four elements converge and reinforce one another. None of them is sufficient on its own and their combined behaviour shows consistent patterns across the region. Countries with higher levels of AI advancement tend to have stronger innovation systems, more robust governance frameworks, broader and more accessible digital infrastructures and public administrations capable of absorbing and scaling emerging technologies.
The purpose of this research is to provide an empirical foundation that supports the design of more coherent and sustainable sector-specific AI strategies. The analysis helps identify the minimum structural conditions required for these strategies to succeed, anticipates fragmentation risks and offers a framework for aligning national capacities with emerging sectoral initiatives. In a context of accelerated technological change, clarity about these factors becomes essential for countries seeking to move from being technology consumers to becoming developers of their own AI solutions, adapted to their realities and sustainable over time.
2. METHODOLOGY
2.1. Study design
This study adopts a quantitative, comparative and correlational design to analyse the structural factors that explain variability in the Latin American Artificial Intelligence Index (ILIA) across 17 countries. The methodological approach integrates four international indices that capture complementary dimensions of the digital ecosystem: innovation capacity, AI governance, digital infrastructure and e-government maturity. The objective is to determine whether these pillars jointly predict the level of AI maturity observed in the region.
2.2. Sample and country selection
The sample includes 17 Latin American countries for which complete and comparable data are available across all indices used in the analysis. Inclusion is determined exclusively by the simultaneous availability of data for ILIA, GII, GAIRI, the ICT Development Index and EGDI, ensuring methodological consistency and avoiding bias due to missing observations.
2.3. Variables
Five main variables were analysed, all normalised by their respective institutions:
ILIA_PROM. Average score of the Latin American Artificial Intelligence Index (ILIA 2025), used as the dependent variable and representing general AI maturity.
GII_PROM. Average score of the Global Innovation Index (GII 2025), capturing innovation capacity, human capital, research, and business sophistication.
GAIRI_Total. Total score of the Global AI Readiness Index (GAIRI 2024), reflecting institutional, regulatory, and strategic preparedness for AI.
ICT Development Index (2025). Composite measure from the ITU that evaluates connectivity, access, affordability and digital infrastructure.
EGDI_2024. Overall score of the E-Government Development Index (UN EGDI 2024), assessing the degree of digitalisation in the public sector.
All variables were taken in their published scales, which are already harmonised for cross-country analysis.
2.4. Data sources
Data were obtained from internationally recognised and methodologically robust sources:
Latin American Artificial Intelligence Index (ILIA 2025)
Global Innovation Index (GII 2025), WIPO
Global AI Readiness Index (GAIRI 2024), Oxford Insights
ICT Development Index 2025, ITU
E-Government Development Index (EGDI 2024), United Nations
All sources correspond to the latest available cycle for 2024–2025.
2.5. Hypotheses
The analysis tests six hypotheses:
H1: Higher innovation capacity (GII) is associated with higher AI maturity (ILIA).
H2: Stronger institutional and regulatory AI governance (GAIRI) predicts higher AI maturity.
H3: Digital infrastructure (ICT Index) acts as an enabling condition for AI adoption.
H4: E-government development (EGDI) is positively related to AI maturity.
H5: The combination of GII, GAIRI, ICT and EGDI explains most of the variance in ILIA.
H6: Gaps between ILIA and the other indices reflect structural misalignment in national digital ecosystems.
2.6. Statistical techniques
2.6.1. Pearson correlations
Used to assess the strength and direction of bivariate associations between ILIA and each independent variable. This identifies which pillars move in tandem with AI maturity.
2.6.2. Multiple linear regression
The general model estimated is:
Outputs include coefficient estimates, standard errors, t-values, p-values, R², adjusted R² and ANOVA components (SSR, SSE). The aim is to determine how much variance in ILIA is explained when all predictors act simultaneously.
2.6.3. Multicollinearity diagnostics
Variance Inflation Factor (VIF) and auxiliary R² were computed to evaluate shared variance among predictors. This is necessary because many global indices partially rely on overlapping data sources.
2.6.4. Structural gap analysis
Differences between ILIA and other indices were calculated:
These gaps indicate whether a country’s AI maturity is aligned with its broader digital ecosystem.
International indices may not fully reflect sector-specific or social dimensions relevant to AI.
Multicollinearity among predictors is expected, as global indices often share underlying sources and constructs.
Findings apply specifically to Latin America and may not be generalised to other regions without comparative validation.
3. RESULTS
3.1. General overview
The statistical analysis conducted with data from 17 Latin American countries reveals that artificial intelligence maturity, as measured by the Latin American Artificial Intelligence Index (ILIA), behaves as a systemic rather than a purely technological phenomenon. The bivariate correlations between ILIA and four structural pillars—innovation (GII), AI governance (GAIRI), digital infrastructure (ICT Development Index), and e-government development (EGDI)—are strong and statistically significant in all cases. This indicates that each of these components contributes independently to national AI readiness.
The multiple regression model reinforces this interpretation. When the four pillars are considered jointly, the model explains the majority of the cross-country variation in ILIA. This suggests that AI maturity emerges from the interaction of technological, institutional, infrastructural, and administrative capabilities rather than from isolated investments or pilot initiatives.
3.2. Bivariate correlations
Table 1 summarises the strength of association between ILIA and each structural pillar. All coefficients display strong and statistically significant relationships.
Table 1
Correlations between ILIA and structural pillars
Relationship | n | Pearson r | Interpretation | p-value |
|---|
ILIA – GII_PROM | 17 | 0.9009 | Very strong association | 8.1 × 10⁻⁷ |
ILIA – GAIRI_Total | 17 | 0.9739 | Extremely strong association | 4.5 × 10⁻¹¹ |
ILIA – ICT Development Index | 17 | 0.8106 | High association | 1.4 × 10⁻⁴ |
ILIA – EGDI | 17 | 0.8731 | High association | 2.1 × 10⁻⁵ |
| Correlation coefficients between the Latin American Artificial Intelligence Index ILIA and four structural pillars included in the analysis. Higher values indicate stronger positive associations. All relationships are statistically significant at varying levels. Source: Author’s elaboration based on ILIA 2025, GII 2025, GAIRI 2024, ITU ICT Development Index 2025 and UN EGDI 2024 |
3.3. Multiple regression: a systemic model of AI maturity
The multiple regression model uses ILIA as the dependent variable and the four structural indices as predictors.
3.3.1 Model parameters
A
Table 2
Estimated coefficients for the regression model ILIA = f(GII, GAIRI, ICT, EGDI)
Parameter | Coefficient | Standard Error | t | p | Interpretation |
|---|
Intercept | -14.3363 | 9.3438 | -1.53 | 0.1445 | Not significant |
GII_PROM | -0.0356 | 0.7936 | -0.04 | 0.9650 | Not significant |
GAIRI_Total | 1.3787 | 0.3385 | 4.07 | 0.0015 | Only significant predictor |
ICT Index | 0.0933 | 0.3747 | 0.25 | 0.8075 | Not significant |
EGDI | -17.6274 | 30.7309 | -0.57 | 0.5768 | Not significant |
| Regression coefficients for the multivariate model using ILIA as the dependent variable and four structural pillars as predictors. GAIRI is the only statistically significant predictor, while the remaining coefficients lose significance due to high multicollinearity among indices. Source: Author’s elaboration based on ILIA 2025, GII 2025, GAIRI 2024, ITU ICT Development Index 2025 and UN EGDI 2024 |
3.3.2 Overall model fit
A
Table 3
Global indicators of the regression model
Metric | Value | Interpretation |
|---|
R² | 0.9502 | The model explains 95% of ILIA variance |
Adjusted R² | 0.9336 | Robust fit after penalising model complexity |
SSR | 4180.62 | Explained variation |
SSE | 218.97 | Residual variation |
| Overall model fit statistics for the regression explaining ILIA using four structural pillars. The model demonstrates high explanatory power and robust adjustment. Source: Author’s elaboration based on ILIA 2025, GII 2025, GAIRI 2024, ITU ICT Development Index 2025 and UN EGDI 2024 |
| Although the model as a whole shows an excellent fit, the individual coefficients reveal a distinct pattern: |
GAIRI is the only statistically significant predictor.
The remaining predictors lose significance due to very high mutual collinearity, not because they lack conceptual relevance.
This collinearity is expected because GII, ICT and EGDI share overlapping conceptual and methodological components.
The model therefore supports the idea that AI maturity arises from an integrated environment where governance, innovation, infrastructure and public-sector digital capability evolve together.
3.4 Collinearity among predictors
The collinearity matrix confirms the presence of strong relationships among the predictors, explaining the loss of individual statistical significance despite the high explanatory power of the model.
A
Table 4
Correlation matrix among predictors
| | GII | GAIRI | ICT | EGDI |
|---|
GII | 1.000 | 0.9199 | 0.9194 | 0.8555 |
GAIRI | 0.9199 | 1.000 | 0.8337 | 0.9114 |
ICT | 0.9194 | 0.8337 | 1.000 | 0.8938 |
EGDI | 0.8555 | 0.9114 | 0.8938 | 1.000 |
| Correlation matrix showing the relationships among the four structural predictors included in the regression model. Source: Author’s elaboration based on ILIA 2025, GII 2025, GAIRI 2024, ITU ICT Development Index 2025 and UN EGDI 2024 |
The four predictors exhibit very high correlations, indicating that they form a tightly interrelated structural system. Although each index measures a distinct component of digital development, they move consistently together across the region. This confirms the systemic nature of AI maturity in Latin America, where innovation, governance, infrastructure and public-sector digitalisation tend to evolve jointly.
3.5 Structural sectors associated with AI maturity
Based on the empirical evidence from correlations and regression results, the following sectors demonstrate measurable influence on ILIA values.
A
Table 5
Structural sectors associated with AI maturity
Pillar | Associated sectors | Statistical support |
|---|
Innovation (GII) | Higher education, R&D, scientific output, technological entrepreneurship, business complexity | r = 0.9009 |
AI governance (GAIRI) | Regulatory frameworks, institutional capacity, national AI strategies, public-sector coordination | coef. = 1.3787; p = 0.0015 |
Digital infrastructure (ICT Index) | Connectivity, broadband penetration, affordability, coverage, interoperability | r = 0.8106 |
E-government (EGDI) | Digital services, digital identity, public-sector data management | r = 0.8731 |
| Sectors most closely associated with national AI maturity according to the statistical relationships observed in the correlation and regression analyses. Each pillar reflects a set of system-level capacities that contribute to the conditions enabling AI development. Source: Author’s elaboration based on ILIA 2025, GII 2025, GAIRI 2024, ITU ICT Development Index 2025 and UN EGDI 2024 |
These sectors emerge as the most plausible structural engines conditioning AI maturity in Latin America. The evidence highlights the relevance of designing sector-specific AI strategies within the context of broader systemic capacities.
4. Discussion
The results of this study position artificial intelligence maturity not as a technological variable but as an emergent property of broader developmental systems. The strong and consistent associations between the Latin American Artificial Intelligence Index (ILIA) and the four structural dimensions examined—innovation capacity, institutional governance of AI, digital infrastructure, and e-government development—suggest that AI readiness is embedded within pre-existing national capabilities. These findings align with international research showing that countries capable of generating knowledge, sustaining digital connectivity, and coordinating state action tend to deploy emerging technologies more effectively.
Although ILIA integrates several indicators indirectly drawn from the other indices used in this analysis, the relationships observed cannot be attributed solely to definitional overlap. The magnitude and statistical significance of the correlations, as well as the explanatory power of the multivariate model, indicate the existence of a shared structural architecture. In practical terms, this means that AI adoption tends to succeed where broader innovation and digital governance systems are already functioning, and conversely, it struggles where such systems remain fragmented or underdeveloped.
The interaction among these dimensions also reveals an important dynamic. Innovation systems provide the research, skills, and entrepreneurial capacity that make AI experimentation viable. Governance frameworks contribute strategic direction, regulatory clarity, and institutional coordination. Digital infrastructure enables data circulation and operational deployment. E-government systems create demand for algorithmic tools and anchor the public sector’s capacity to scale them. The convergence of these elements appears to generate the conditions that allow AI to transition from isolated pilots to fully integrated national capabilities.
At the same time, the results underscore a challenge for policymaking. In Latin America, national AI strategies often emerge in parallel to digital transformation, innovation, and connectivity agendas, rather than as part of an integrated policy framework. The evidence presented here suggests that such fragmentation limits policy effectiveness. AI initiatives are more likely to produce sustained impact when they are designed as extensions of existing national capabilities and when they contribute to reinforcing those same systems.
Finally, the analysis raises questions for future research. The current results capture national-level patterns, but they do not yet illuminate how sector-specific capacities—such as those in agriculture, health, education, or public security—mediate AI adoption in practice. Nor do they identify the mechanisms through which innovation, governance, infrastructure, and digital public services interact at operational levels. Understanding these dynamics will be essential for the development of targeted sectoral AI strategies, an area that is rapidly gaining relevance across the region.
5. Conclusions
The empirical results demonstrate that the effective adoption of artificial intelligence does not occur as an isolated technological endeavour but emerges from the interaction of broader structural capacities. AI maturity, as captured by the Latin American Artificial Intelligence Index (ILIA), consistently appears only where national systems of innovation, institutional governance, digital infrastructure, and e-government development are simultaneously present and operational. These dimensions do not act as peripheral enablers; they constitute the foundational conditions that determine whether countries can deploy, adapt, and scale AI systems in a sustainable manner.
Across all statistical analyses conducted, countries exhibiting higher levels of AI maturity are also those with more consolidated innovation ecosystems, functional regulatory and strategic frameworks for AI, stronger connectivity and infrastructure, and more advanced public digital services. The robustness of these associations indicates that national AI strategies must be explicitly integrated within broader innovation, digital transformation, and state modernisation policies. In practical terms, an AI strategy cannot function as a standalone instrument: its effectiveness depends on its coherence with these structural pillars.
The results also reveal a reciprocal implication. In contexts where innovation policies, digital governance frameworks, and e-government systems already exist, the design of an AI strategy must be situated within this institutional architecture. Treating AI as a separate policy domain—detached from ongoing efforts in connectivity, digital public services, and infrastructure development—contradicts the empirical evidence and reduces the likelihood of achieving long-term impact. AI readiness is, in practice, a cumulative outcome of the strengthening of these underlying systems rather than the result of isolated technological initiatives.
Taken together, these findings support a central conclusion. Any national or sectoral AI policy must be conceived as a derivative and integrative component of the wider digital development agenda. Conversely, such policies must also contribute to reinforcing innovation systems, data governance, connectivity, and public-sector digitalisation. Fragmented approaches focused solely on acquiring technologies or deploying isolated pilots are unlikely to produce meaningful or enduring results. The strategic alignment of AI with innovation, infrastructure, and government digitalisation therefore constitutes a necessary condition for the viability and sustainability of AI initiatives in the region.
6. Policy Implications
The findings of this study allow for the identification of clear and evidence-based implications for the design of Artificial Intelligence (AI) policies in Latin America. Each implication derives directly from the statistical relationships observed and does not rely on normative or speculative claims.
First, the results confirm that AI maturity is systematically dependent on national innovation performance. Countries with higher ILIA scores also display stronger research capacity, advanced education systems, technological entrepreneurship, and business sophistication. These elements are not peripheral to an AI strategy; they constitute the enabling conditions that determine whether countries can develop, adapt, and localise AI systems. Strategies that are not articulated with higher education policies, research funding, or support for technological ventures are likely to remain limited in scope. The evidence shows that AI does not advance where the scientific and technological base is weak.
Second, institutional governance emerges as a decisive factor. GAIRI scores indicate that the presence of regulatory frameworks, dedicated organisational units, inter-ministerial coordination mechanisms, and public-sector technical capability is associated with higher AI maturity. This implies that an AI strategy must be supported by a stable administrative structure with clear mandates and continuity in institutional leadership. Weak governance does not simply slow AI adoption; it increases the risk of fragmented decision-making, duplication of efforts, and inefficient allocation of public resources.
Third, digital infrastructure is confirmed as a cross-cutting enabler. AI maturity is significantly higher in countries with stronger urban and rural connectivity, broader broadband penetration, greater affordability of data, and more robust cloud and interoperability capabilities. This means that AI programmes cannot be designed independently of national connectivity conditions. The evidence indicates that sectoral AI initiatives—whether in education, agriculture, health or public administration—become constrained when infrastructure is insufficient, limiting scalability and long-term sustainability.
Fourth, the development of e-government acts as an institutional bridge between AI and public value creation. EGDI scores show that countries with consolidated digital public services, functional digital identity systems, and strategic data management exhibit higher levels of AI maturity. This implies that AI must be embedded within broader state modernisation efforts, including digital process redesign, interoperable information systems, and strategic use of public data.
Taken together, these results converge in a central implication. AI policies must be conceived as components of a broader systemic architecture. They must be aligned with—and reinforced by—innovation strategies, institutional governance, digital infrastructure, and e-government development. Fragmented approaches focused solely on acquiring technologies or deploying isolated pilots show limited potential for durable impact. By contrast, strategies that are integrated into the structural systems of the country are better positioned to produce sustained results, reduce gaps, and advance toward a more sovereign and effective adoption of AI.
7. Limitations
This study presents several limitations that should be considered when interpreting the findings.
First, the analysis relies exclusively on internationally aggregated indices. Although ILIA, GII, GAIRI, the ICT Development Index, and the EGDI are methodologically robust and comparable across countries, none of them captures the sector-specific or social dynamics that influence AI adoption within countries. The results describe structural patterns at the national level but do not explain how these conditions operate in concrete domains such as agriculture, education, health, or public security.
Second, multicollinearity among predictors is an inherent feature of the study design. These global indices share conceptual foundations, data sources, and measurement assumptions, which naturally produces strong correlations among them. This is reflected in the high VIF values and the loss of statistical significance of some individual coefficients, even though the overall model maintains a high level of explanatory power. The regression results must therefore be interpreted systemically rather than as independent causal contributions of each variable.
Third, the sample includes only the 17 Latin American countries for which complete and comparable data are simultaneously available for the 2024–2025 cycle. This ensures methodological consistency but reduces statistical power and limits the ability to test more complex or non-linear models. In addition, the findings are specific to Latin America and cannot be directly generalised to other regions without comparative validation.
Finally, the study provides a short-term snapshot. Although the indices used correspond to the most recent cycle, they do not allow the analysis of temporal dynamics or long-term trajectories. As a result, the findings identify robust structural associations but do not establish causality or capture lagged effects of recent policy interventions.
8. Future Research Directions
Based on the limitations identified and the empirical patterns observed, several avenues for future research can expand and operationalise the results of this study.
First, future work should move from country-level aggregates to sector-level analysis. Integrating ILIA with indicators specific to agriculture, health, education, or public administration would make it possible to examine how structural capacities translate into concrete AI deployment in each domain. This would help determine whether countries with stronger structural foundations actually succeed in implementing AI solutions in strategic sectors.
Second, incorporating a temporal dimension would substantially strengthen the evidence base. Longitudinal analyses using historical series of ILIA, GII, GAIRI, ICT, and EGDI would allow researchers to identify trajectories, lagged effects, and turning points, and to assess whether improvements in innovation, governance, or infrastructure consistently precede gains in AI maturity. Such designs would also support stronger causal inference.
Third, future research could address multicollinearity explicitly through advanced statistical techniques, such as principal component analysis, structural equation modelling, or regularised regression methods (e.g., ridge, LASSO). These approaches may help isolate latent constructs that are shared across global indices and differentiate between broad contextual factors and more proximate determinants of AI readiness.
Fourth, complementing global indices with national indicators and micro-level data would provide a richer empirical foundation. Potential sources include administrative records on AI projects, surveys of public institutions, universities and firms, and detailed sectoral case studies. These data would help validate whether the structural relationships identified at the macro level align with practices and decision-making processes within institutions.
Finally, comparative research across different world regions would help determine whether the structural patterns identified in Latin America are context-specific or reflect more universal mechanisms. Cross-regional comparisons could clarify which systemic drivers are globally consistent and which depend on particular institutional, economic, or technological configurations.