0 cumulative citations
View corpus contextAI is increasingly used to make corporate sustainability decisions—mostly to optimize operations and cut costs and emissions—but progress is hampered by poor ESG data, interoperability and limited attention to social dimensions.
Artificial intelligence in business decision making under ESG criteria a systematic literature review
openalex
review_meta
n/a evidence
7/10 relevance
Full text
usable extracted full text
DOI
Source
PDF
A PRISMA-based review of 119 Scopus articles (2020–2025) finds AI in ESG-focused business decisions is concentrated on operational optimization, predictive models, and automated reporting—delivering cost reductions, emissions cuts, and better forecasts—while adoption is constrained by data quality, governance, explainability, and organizational barriers.
Citation observations
Cumulative provider counts captured on specific dates; providers are never combined.
0 cumulative citations
View corpus contextAbstract This paper examines 119 articles from the SCOPUS database using a systematic search of scientific publications related to the influence of AI on sustainable business decision-making (decisions that incorporate environmental, social and governance—ESG). The results demonstrate that AI is currently applied primarily in process optimization, simulations and predictive models, automated reporting and auditing, stakeholder management, product innovation, and approaches to the circular economy. Furthermore, it is evident that the most documented benefits of these applications are reduced operating costs, lower emissions, improved predictive accuracy, positive engagement, enhanced transparency, and improved energy efficiency. The conclusion is that AI can be considered a powerful catalyst for integrating sustainability principles into corporate decision-making.
Summary
Main Finding
AI is a powerful catalyst for integrating sustainability principles into corporate decision-making, but its documented impact is concentrated on operational and environmental outcomes (costs, emissions, energy efficiency) rather than social outcomes. The systematic review of 119 Scopus open‑access articles (2020–2025) finds that AI is mainly used for operational optimization (supply chains, energy), predictive ESG models, and automation of ESG reporting, while adoption is constrained primarily by data quality, governance and organizational barriers.
Key Points
- Scope: Systematic literature review of 119 peer‑reviewed, open‑access articles indexed in Scopus (published 2020–2025).
- Main practical applications (N articles, % of 119):
- Operational optimization (supply chains, energy): 64 (53.78%)
- ESG predictive models & simulations: 25 (21.01%)
- Automation of ESG reports & audits: 13 (10.92%)
- Stakeholder analysis & marketing: 10 (8.40%)
- Product innovation & circular economy: 7 (5.88%)
- Most reported benefits:
- Reduction in operating costs: 44 (36.97%)
- Decrease in pollutant emissions: 33 (27.73%)
- Improved predictive accuracy: 33 (27.73%)
- Engagement / social transparency: 14 (11.76%)
- Improved energy efficiency: 8 (6.72%)
- Main barriers to adoption:
- Data access & reliability: 72 (60.50%)
- Training & organizational resistance: 34 (28.57%)
- Data governance & interoperability: 33 (27.73%)
- Algorithmic opacity / lack of explainability: 31 (26.05%)
- Infrastructure and tool costs: 30 (25.21%)
- Leading sectors by presence in the literature:
- Energy & utilities: 20 (16.81%)
- Manufacturing / smart factories: 18 (15.13%)
- Finance / fintech: 15 (12.61%)
- Transportation & logistics: 11 (9.24%)
- Agriculture: 10 (8.40%)
- Emerging trends:
- Explainable AI (XAI): 17 (14.29%)
- Human–AI collaborative platforms: 12 (10.08%)
- Blockchain + AI for traceability & finance: 10 (8.40%)
- IoT + edge computing integration: 10 (8.40%)
- Digital twins & simulations: 7 (5.88%)
- Noted imbalance: economic and environmental dimensions dominate; social dimension is underrepresented (~8.4% of application literature).
Data & Methods
- Search strategy: PRISMA-based systematic search in Scopus using keywords across TITLE-ABS-KEY for AI (e.g., "artificial intelligence", "machine learning", "explainable AI"), decision-making terms (e.g., "decision support", "multiple criteria decision analysis"), sustainability/ESG terms, and business/corporate terms.
- Filters applied: document type = research articles; language = English or Spanish; publication stage = completed articles; year range = 2020–2025; access = open access.
- Initial hits: 130 papers; after abstract/full‑text screening and exclusion of off‑topic/withdrawn items, final sample = 119 articles.
- Quality control: PRISMA screening combined with full‑text manual review for internal coherence (no standardized checklist used). Two databases exported (query metadata and Zotero collection) with figshare DOIs provided by the author.
- Limitations of method highlighted by the author: reliance on Scopus and open‑access filter (may omit paywalled work), manual coherence check but no formal quality scoring, and scope restricted to 2020–2025.
Implications for AI Economics
- Productivity & cost effects
- Strong evidence that AI yields measurable operational gains (cost reductions, efficiency) — relevant for firm‑level productivity estimates and TFP studies. Economists can quantify these effects across sectors (energy, manufacturing, logistics) where adoption is concentrated.
- Environmental externalities & decarbonization
- Documented emission reductions and energy efficiency gains suggest AI can be an instrument for climate mitigation. But rigorous causal estimates and lifecycle accounting (including compute carbon footprint and rebound effects) are needed to assess net environmental impact.
- Distributional and labor market impacts
- Operational focus implies potential for task automation and labor reallocation. The literature's limited attention to social outcomes signals a gap: economic research should estimate employment, wage, and skill-upgrading effects of AI adoption within sustainable transformation.
- Data as an economic input and potential market failure
- Data access/quality is the principal barrier—this raises economics questions about data provision as a public good, incentives for ESG disclosure, pricing of high‑quality ESG data, and the role of standards/regulation to correct coordination failures.
- Governance, transparency, and regulation
- Algorithmic opacity and governance/interoperability challenges point to demand for explainability, auditability, and standardized metrics. From a policy perspective, regulators may need to mandate XAI, reporting standards, and interoperable data formats to unlock social value from AI.
- Sectoral policy targeting & investment
- Evidence concentration in energy, manufacturing, and finance suggests targeted R&D subsidies, infrastructure investment (especially for SMEs), and workforce training in those sectors will yield higher ESG returns from AI.
- Research gaps and priorities for economists
- Need for causal and quantitative studies: firm‑level panel studies linking AI adoption to productivity, emissions, ESG ratings and profitability.
- Measurement: standardized metrics to compare AI‑driven sustainability outcomes across firms/countries.
- Macro modeling: aggregate effects of widespread AI adoption on emissions, employment, and welfare (including rebound effects).
- Valuation of social outcomes: develop methods to integrate underexplored social dimensions into cost‑benefit analysis of AI deployment for sustainability.
- Policy experiments: evaluate subsidies, data‑sharing mandates, or disclosure rules to address the dominant barrier of data quality/access.
Suggested short research questions for AI economics inspired by the review: - What is the causal effect of firm‑level AI adoption on energy consumption and emissions intensity? - How do data‑sharing mandates or standardized ESG reporting affect firms’ AI adoption and sustainability performance? - What are the distributional labor impacts (occupation, wages) associated with AI-driven operational optimization in manufacturing and logistics? - How large are the net environmental gains once the carbon costs of compute and potential rebound effects are accounted for?
Reference: Vásquez‑Correa, J. L. (2026). Artificial intelligence in business decision making under ESG criteria: a systematic literature review. Discover Environment, 4:395. DOI: 10.1007/s44274-026-00907-w (open access).
Assessment
Paper Typereview_meta
Evidence Strengthn/a — This is a narrative/systematic review synthesizing findings from 119 heterogeneous studies rather than producing primary causal estimates; it does not itself identify causal effects.
Methods Rigormedium — Uses an explicit search query and PRISMA flow, reads full texts and documents databases/exports, but restricts sources to Scopus open‑access items (2020–2025), uses no standardized quality appraisal (e.g., risk-of-bias checklist), and provides limited handling of heterogeneity/publication bias.
Sample119 peer‑reviewed, open‑access research articles indexed in Scopus, published in English or Spanish between 2020 and 2025, selected via a keyword query targeting AI/ML and sustainability/ESG in business decision-making; excluded non‑research items, in‑press/withdrawn works and non‑open access outputs.
Themesadoption productivity
IdentificationSystematic literature review using a PRISMA-based screening of the Scopus database with a defined keyword query and sequential filters (document type=research articles, language=English/Spanish, publication stage=completed, year range=2020–2025, open access). No causal identification strategy (descriptive synthesis only).
GeneralizabilityRestricted to Scopus-indexed, open-access journals which may omit relevant subscription or grey literature, Limited to publications from 2020–2025, so earlier/longitudinal literature is excluded, Language-limited to English and Spanish, excluding other languages, Heterogeneous set of study types (case studies, simulations, models, empirical analyses) with no standardized quality weighting, Findings are descriptive and cannot establish causal impacts of AI on economic outcomes
Claims (11)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| Operational optimization of supply chains and energy systems is the most frequently documented application of AI in sustainable business decision-making. Adoption Rate | positive | Frequency of literature documenting AI applications in operational optimization |
Reading fidelity
high
Study strength
medium
|
n=119
64 articles (53.78%)
|
| AI-based ESG predictive models and simulations are a major documented application in sustainable business decision-making, particularly for calculating environmental and social indicators. Decision Quality | positive | Frequency of literature documenting ESG predictive models and simulations |
Reading fidelity
high
Study strength
medium
|
n=119
25 articles (21.01%)
|
| The most frequently reported benefit of applying AI in sustainable business decision-making is a reduction in operating costs. Organizational Efficiency | positive | Reduction in operating costs |
Reading fidelity
high
Study strength
medium
|
n=119
44 articles (36.97%)
|
| Reduced pollutant emissions are among the most frequently documented benefits of AI use in sustainable business decision-making. Organizational Efficiency | positive | Reduction in pollutant emissions |
Reading fidelity
high
Study strength
medium
|
n=119
33 articles (27.73%)
|
| Improved predictive accuracy is a frequently reported benefit of AI in sustainable business decision-making. Decision Quality | positive | Predictive accuracy for anticipating scenarios and managing sustainability risks |
Reading fidelity
high
Study strength
medium
|
n=119
33 articles (27.73%)
|
| Data access and reliability is the most frequently documented barrier to adopting AI for sustainable business decisions. Governance And Regulation | negative | Barrier to AI adoption caused by difficulty obtaining, integrating, and maintaining high-quality ESG data |
Reading fidelity
high
Study strength
medium
|
n=119
72 articles (60.50%)
|
| Training needs and organizational resistance are substantial barriers to AI adoption for sustainable decision-making. Training Effectiveness | negative | AI adoption constraints related to digital skills and internal organizational resistance |
Reading fidelity
high
Study strength
medium
|
n=119
34 articles (28.57%)
|
| Algorithmic opacity and lack of explainability are recurring barriers to AI-supported sustainable decision-making. Ai Safety And Ethics | negative | Transparency and explainability of automated decision-making |
Reading fidelity
high
Study strength
medium
|
n=119
31 articles (26.05%)
|
| Energy and utilities is the sector most frequently represented in the literature on AI adoption for sustainable business decision-making. Adoption Rate | positive | Frequency of sectoral coverage of AI adoption |
Reading fidelity
high
Study strength
medium
|
n=119
20 articles (16.81%)
|
| Explainable AI and algorithmic transparency are the most frequently identified emerging trends in AI use for sustainable business decision-making. Ai Safety And Ethics | positive | Frequency of literature identifying explainable AI and transparency as an emerging trend |
Reading fidelity
high
Study strength
medium
|
n=119
17 articles (14.29%)
|
| The reviewed literature gives greater attention to economic and environmental ESG dimensions than to the social dimension. Inequality | negative | Relative representation of ESG dimensions in the reviewed AI applications |
Reading fidelity
high
Study strength
low
|
n=119
social dimension: 8.40%
|