Unveiling the blackbox within <scp>ESG</scp> ratings' blackbox: Toward a framework for analyzing AI adoption and its impacts

Abstract Artificial intelligence (AI) is transforming entire industries at an unprecedented pace. Yet, established technology adoption theories offer limited tools for characterizing business AI integration and analyzing its effects. These primarily focus on the factors facilitating or hindering adoption, rather than on adoption patterns and impacts. This paper introduces a novel conceptual framework to address this key gap and applies it to the case of the ESG rating industry. ESG raters play a pivotal role in sustainable finance, providing metrics that guide investment decisions globally. However, little is known about the extent and nature of their AI usage and its implications. Through a mixed‐methods approach combining the analysis of job postings, patent filings, research publications, and corporate websites, we examine AI adoption among major ESG raters. Our investigation explores the specific AI technologies employed, their functional applications, the innovations developed, the intensity of AI integration, and the potential impacts of raters' AI adoption. Our results reveal widespread and growing AI adoption across the industry. Our findings show that raters extensively leverage Natural Language Processing to streamline data collection, processing, and analysis. Furthermore, they have pioneered Machine Learning innovations that significantly expand their sustainability assessment capabilities in various domains. These findings mark a considerable departure from prior academic and gray literature that characterized major ESG raters as having minimal AI use, prompting critical questions regarding the implications of this technological transformation for ESG ratings' reliability, transparency, and potential biases.