AI for Environmental, Social and Governance (ESG) Integration in Strategic Management a PRISMA Systematic Review

Abstract

Background: Companies are using Artificial Intelligence (AI) more and more to include Environmental, Social, and Governance (ESG) factors into their strategic planning. This helps with evaluating what matters most, managing risks, deciding how to invest money, checking supply chains, and sharing information. But right now, there's not enough clear evidence on what works best, when it works, and what risks are involved. This information is spread out across different areas like information systems, finance, operations, and corporate governance.
Objective: We want to bring together high-quality research and expert writings to understand how AI helps with ESG integration at the strategic level. This includes how companies plan, carry out, and monitor their strategies. We'll look for patterns in how effective AI is, what risks are involved, and what rules or controls should be in place. We also want to find where future research should go.

Methods: We followed the PRISMA 2020 guidelines to find studies. We looked through several databases like Scopus, Web of Science, IEEE Xplore, ACM Digital Library, ScienceDirect, SSRN, and selected policy websites from the EU, ISSB, and SEBI. We focused on English-language articles from 2010 to 2025. We included studies that clearly connected AI methods (like natural language processing, large language models, machine learning, computer vision, and knowledge graphs) with ESG practices in strategic management areas such as enterprise risk management, strategy, board oversight, capital budgeting, and performance measurement. We checked the quality of the studies using tools like the Mixed Methods Appraisal Tool for empirical studies, the CASP checklists for qualitative case studies and reviews. We used a narrative synthesis along with vote counting and cross-case analysis to organize the findings.

Results: We started with 3,366 records, and after removing duplicates, 2,275 remained. We reviewed 266 full texts, and 112 met the inclusion criteria. These included 43 empirical studies, 22 qualitative or case studies, 7 mixed-method studies, 19 design or technical studies with strategic evaluation, and 21 reviews or conceptual studies. The main uses of AI in ESG are: (1) using natural language processing and large language models to help with ESG reporting, classify information, and provide assurance; (2) using AI combined with satellite data to check for deforestation and human rights issues in supply chains; (3) using AI to analyze climate and transition risks for strategic decisions and investments; (4) measuring ESG performance and creating scores with explanations; (5) creating governance frameworks that align AI risks with the 'G' part of ESG. There's some evidence that AI improves efficiency and coverage, and there's growing evidence of better decision-making and real-world impact. However, there are risks like measurement errors, biases, lack of transparency, and too much trust in proxy labels.
Conclusions: AI can greatly improve ESG integration when it's part of a governance-focused approach with clear responsibilities, reliable data sources, and ways to manage model risks. It should also align with regulatory standards like the ISSB, ESRS, and SEBI BRSR Core. We suggest a cycle that connects AI, ESG, and strategy, and propose a research plan focusing on data tracking, understanding causes, assessing double materiality at scale, and ensuring human rights due diligence.