3 cumulative citations
View corpus contextStrong data governance, not just smarter algorithms, determines whether AI improves financial decisions; a systematic review of 1,155 studies finds governance maturity mediates the link between AI integration and better financial outcomes, with large estimated effects but based on aggregated observational evidence.
Artificial Intelligence in Data Governance for Financial Decision-Making: A Systematic Review
openalex
review_meta
low evidence
7/10 relevance
Summary only
summary available; pdf_status=paywall
DOI
Source
PDF
A PRISMA-guided review of 1,155 studies finds governance maturity strongly mediates the relationship between AI integration and improved financial outcomes (SEM βs ≈ 0.7), indicating algorithms alone do not guarantee decision quality without robust governance.
Citation observations
Cumulative provider counts captured on specific dates; providers are never combined.
5 cumulative citations
View corpus contextArtificial intelligence (AI) has been increasingly embedded within data-driven financial decision-making; however, its effectiveness was found to remain dependent upon the maturity of data governance frameworks. This systematic review was conducted in accordance with PRISMA 2020 guidelines to synthesise evidence from 1155 Scopus-indexed studies published between 2015 and 2025. A mixed-methods design combining corpus analysis, quantile radar regression, and radar visualisation of structural equation modelling (SEM) was employed. Empirical validation was found to demonstrate a robust model fit (CFI = 0.947; RMSEA = 0.041). Governance maturity was confirmed as a mediating construct (β = 0.73) linking AI integration (β = 0.76) to financial outcomes (β = 0.71). The findings were found to indicate that algorithmic capacity alone does not ensure decision quality without transparent, auditable, and ethically grounded governance systems. A quantile-sensitive radar visualisation is advanced in this review, offering conceptual and methodological novelty for explainable, responsible, and data-centric financial analytics. This study is found to contribute to the ongoing discourse on sustainable digital transformation within AI-enabled financial ecosystems.
Summary
Main Finding
Mature data governance is a crucial mediator that enables AI integration to translate into better financial outcomes. Algorithmic capacity alone is insufficient: transparent, auditable, and ethically grounded governance systems are required to realize the value of AI in financial decision-making.
Key Points
- Systematic review of 1,155 Scopus-indexed studies (2015–2025), conducted per PRISMA 2020 guidelines.
- Mixed-methods synthesis combining corpus analysis, quantile radar regression, and radar visualisation of structural equation modelling (SEM).
- Empirical model shows strong fit (CFI = 0.947; RMSEA = 0.041).
- Governance maturity acts as a mediator with large effect (β = 0.73) in the path from AI integration (β = 0.76) to financial outcomes (β = 0.71).
- Introduces a quantile-sensitive radar visualisation for revealing distributional (quantile) heterogeneity in relationships—supporting explainability and responsible analytics.
- Conceptual contribution: reframes performance as data- and governance-dependent rather than purely algorithm-dependent.
- Practical implication emphasized: audits, transparency, ethical safeguards, and data-centric practices are necessary to capture AI’s financial benefits.
Data & Methods
- Evidence base: 1,155 articles indexed in Scopus, published 2015–2025; selection and reporting aligned with PRISMA 2020.
- Methods:
- Corpus analysis to map themes, terminology, and empirical coverage across the literature.
- Quantile radar regression to estimate how relationships vary across the outcome distribution (i.e., different performance quantiles).
- Structural equation modelling to test hypothesised causal paths (AI integration → governance maturity → financial outcomes).
- Radar visualisations to present SEM results across quantiles for interpretability and explainability.
- Key statistical results: SEM fit indices (CFI = 0.947; RMSEA = 0.041); path coefficients AI→governance β = 0.76, governance→outcomes β = 0.73, AI→outcomes β = 0.71 (reported as part of mediated model).
Implications for AI Economics
- Policy and regulation:
- Prioritise standards for data governance, auditability, and transparency as prerequisites for AI-driven financial products and services.
- Design regulatory regimes that incentivise governance maturity (e.g., certification, disclosure requirements, audit trails).
- Firm strategy and investment:
- Allocate resources to data governance, lineage, quality controls, and ethics/compliance infrastructure, not solely to model development.
- Use quantile-aware monitoring tools to detect when AI yields uneven benefits or amplified risks across different performance segments.
- Measurement and evaluation:
- Incorporate governance-maturity metrics in ROI and productivity assessments of AI projects.
- Employ distribution-sensitive (quantile) analyses to surface heterogeneity in AI impacts—important for risk management and pricing.
- Research agenda:
- Further empirical work to operationalise governance maturity metrics, test causal mechanisms in longitudinal settings, and validate quantile-visualisation tools in field deployments.
- Broader economic perspective:
- Sustainable digital transformation in finance requires aligning algorithmic improvements with institution-level governance; otherwise, AI can produce uneven or unstable economic benefits.
Assessment
Paper Typereview_meta
Evidence Strengthlow — The paper synthesises a large literature but relies on aggregated, observational study measures and SEM to infer mediation; SEM fit and large sample do not establish causal identification because underlying studies are heterogeneous, often non-experimental, and subject to publication and measurement bias.
Methods Rigormedium — The review follows PRISMA and applies novel mixed methods (corpus analysis, quantile radar regression, SEM) with reported fit statistics, which demonstrates methodological care and innovation; however, limitations include potential selection bias (Scopus-only), unclear quality weighting of included studies, heterogeneous designs and measures across sources, and limited description of sensitivity analyses or risk-of-bias assessment.
Sample1155 studies indexed in Scopus published between 2015 and 2025 covering AI applications in data-driven financial decision-making; data are study-level (literature corpus) measures used in corpus analysis, quantile radar regression, and aggregated input into SEM rather than primary microdata on firms or workers.
Themesgovernance productivity
IdentificationSystematic review and meta-synthesis using corpus analysis and structural equation modelling (SEM) on aggregated measures extracted from 1,155 Scopus-indexed studies; mediation inferred from SEM parameter estimates and quantile radar regression rather than from randomized or quasi-experimental variation.
GeneralizabilityRestricted to studies indexed in Scopus (may omit non-indexed, sector reports, or grey literature), Heterogeneous study designs and measures across included papers limit comparability, Findings pertain to financial decision-making contexts and may not generalise to non-financial sectors, Cross-jurisdictional differences in regulation and governance not fully accounted for, Temporal cutoff (2015–2025) may miss rapid post-2025 developments or lagged effects, Aggregation at the study level obscures firm-/worker-level heterogeneity
Claims (7)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| This systematic review synthesised evidence from 1,155 Scopus-indexed studies published between 2015 and 2025 in accordance with PRISMA 2020 guidelines. Other | positive | volume and scope of literature reviewed |
Reading fidelity
high
Study strength
high
|
n=1155
|
| A mixed-methods design combining corpus analysis, quantile radar regression, and radar visualisation of structural equation modelling (SEM) was employed. Other | positive | methodological approach (mixed-methods integration) |
Reading fidelity
high
Study strength
high
|
not reported
|
| Empirical validation demonstrated a robust SEM model fit (CFI = 0.947; RMSEA = 0.041). Other | positive | model fit (goodness-of-fit indices) |
Reading fidelity
high
Study strength
medium
|
CFI = 0.947; RMSEA = 0.041
|
| Governance maturity was confirmed as a mediating construct (β = 0.73) linking AI integration (β = 0.76) to financial outcomes (β = 0.71). Firm Revenue | positive | financial outcomes |
Reading fidelity
high
Study strength
medium
|
Governance β = 0.73; AI integration β = 0.76; Financial outcomes β = 0.71
|
| AI effectiveness in financial decision-making remains dependent upon the maturity of data governance frameworks (i.e., algorithmic capacity alone does not ensure decision quality without transparent, auditable, and ethically grounded governance systems). Decision Quality | positive | decision quality |
Reading fidelity
high
Study strength
medium
|
not reported
|
| A quantile-sensitive radar visualisation is advanced in this review, offering conceptual and methodological novelty for explainable, responsible, and data-centric financial analytics. Innovation Output | positive | methodological innovation (visualisation technique) |
Reading fidelity
high
Study strength
low
|
not reported
|
| This study contributes to the ongoing discourse on sustainable digital transformation within AI-enabled financial ecosystems. Governance And Regulation | positive | scholarly contribution to discourse |
Reading fidelity
high
Study strength
low
|
not reported
|