AI Unveils ESG Priorities of Asian Giants: A Deep Dive into Corporate Disclosures

In a pioneering effort led by the Asian Development Bank (ADB) in partnership with Kyushu University and aiESG, Inc., researchers have applied cutting-edge deep learning tools to analyze how leading companies in East and Southeast Asia communicate their environmental, social, and governance (ESG) priorities. The study, conducted by Chao Li, Alexander Ryota Keeley, Shunsuke Managi, and Satoru Yamadera, analyzed 480 corporate reports from 293 of the region’s largest firms. These reports span eight countries, namely the People’s Republic of China (PRC), Japan, the Republic of Korea, Indonesia, Malaysia, the Philippines, Singapore, and Thailand, and offer a revealing look into how corporations interpret and emphasize ESG concerns in different national contexts. This research is notable not just for its regional coverage but also for its methodology, which leverages a powerful AI model to parse and evaluate corporate language.

TMPT: A New Standard in ESG Text Analysis

At the core of the study lies the Text Match Pre-Trained Transformer (TMPT), a sophisticated AI tool capable of understanding and categorizing ESG-related text with impressive nuance. TMPT was trained on over 159 million multilingual samples drawn from Wikipedia and academic literature in English, Chinese, and Japanese, making it one of the most robust zero-shot learning models developed for ESG analysis to date. With over 519 million trainable parameters, TMPT outperforms earlier models like FinBERT and ESGBERT, achieving nearly 90% accuracy across multilingual test datasets. This transformer-based architecture allows it to understand the contextual relationship between text and 13 predefined ESG topics without prior exposure to those specific corporate reports. It represents a leap in AI's ability to decode meaning from dense, unstructured language and do so across cultural and linguistic boundaries.

Economics and Governance Still Dominate Corporate Narratives

Despite the increasing global emphasis on sustainability, the study finds that companies across the region still devote the most attention to economic and governance issues. Topics such as economic ripple effects, production costs, and governance risk were the most frequently mentioned in the reports analyzed. Social issues like domestic job creation and work environment came next, followed by environmental concerns such as greenhouse gas emissions and mining consumption. These findings suggest that while ESG discourse is maturing, financial performance and regulatory compliance remain top of mind for many corporate communicators. The reasons for this are manifold; profitability is still the clearest metric of success for investors, and discussions around ESG are often framed in ways that align with financial risk mitigation and value creation.

Regional Contrasts: ESG Through a Cultural and Economic Lens

Though general trends prevail, the research also uncovers stark national differences in ESG emphasis. Korean companies emerged as regional leaders in ESG-focused discourse, frequently highlighting environmental and social themes like emissions, safety, and workplace quality. This is consistent with the country’s push for global sustainability leadership and the influence of large conglomerates (chaebols) that are actively reshaping their image. In contrast, Chinese companies heavily emphasized economic and governance topics, a finding likely linked to the economic stress during the final year of the PRC’s COVID-19 restrictions in 2023. Japan showed a balanced approach, with a particular focus on social themes such as job creation and working conditions, reflecting a corporate culture rooted in social responsibility. Meanwhile, companies in Indonesia, Malaysia, and the Philippines prioritized human rights and community engagement, while Singapore’s firms leaned toward economic issues. Thailand presented a notably balanced profile, aligning closely with the overall regional average across all ESG dimensions.

Language Matters: Multilingual Consistency and Communication Gaps

A unique aspect of the study was its analysis of language consistency across English and local-language reports. Among 163 companies that published reports in both languages, those in Japan, Korea, and China showed high consistency, suggesting rigorous translation and standardized reporting protocols. However, companies in Thailand and Indonesia showed more variance, possibly reflecting weaker local ESG vocabularies or inconsistent localization practices. The researchers caution that these discrepancies may also stem from gaps in training data: languages like Thai and Bahasa Indonesia have fewer high-quality ESG-related textual resources compared to Chinese or Japanese. This insight raises an important concern, AI-based tools may unintentionally amplify existing disparities in global ESG literacy, especially in underrepresented languages. As ESG metrics become more influential, ensuring equitable language representation in AI training datasets becomes critical.

A Call for Scalable, Context-Sensitive ESG Monitoring

The study concludes by emphasizing the importance of regional nuance in ESG analysis and calls into question the universal applicability of current ESG rating systems. Corporate strategies and stakeholder expectations vary widely, and a one-size-fits-all approach to ESG measurement may overlook crucial local factors. By employing advanced AI models like TMPT, researchers and policymakers can gain a more granular, culturally sensitive understanding of how companies articulate their ESG commitments. The study also advocates for expanding training datasets and incorporating more underrepresented languages to ensure fairness and accuracy. In doing so, tools like TMPT could become vital instruments in designing smarter ESG frameworks, shaping sustainability agendas, and guiding investment decisions in an increasingly data-driven world.

Altogether, this research not only enriches our understanding of corporate ESG narratives in Asia’s diverse economies but also demonstrates how AI can bridge the analytical gap between policy ambition and corporate storytelling. As ESG becomes a cornerstone of global business strategy, deep learning models like TMPT offer a glimpse into the future of sustainable corporate analysis, multilingual, scalable, and deeply context-aware.