Artificial intelligence can help institutional investors and asset managers align their emerging market investment strategies with the United Nations’ sustainable development goals. AI can leapfrog gaps in environmental, social and governance data disclosures and reshape business models to drive economic growth and societal progress.
EMs offer attractive investment opportunities to achieve financial returns and impact. ESG-integrated assets under management are expected to increase from $18.4tn in 2021 to $33.9tn by 2026 in EMs.
Although they still lag developed markets, ESG disclosure requirements have grown exponentially in EMs. For example, of the 62 EMs with sustainable bond issuances in 2023, 51% had ESG and climate disclosure requirements. The International Financial Reporting Standards’ framework on sustainability and climate-related disclosures will help to drive an increase in reporting.
Many EM regulators, like Brazil’s Comissão de Valores Mobiliários, are already signalling their intent to require companies to disclose information according to these requirements. Regulators are also mandating domestic ESG disclosure requirements such as the Securities and Exchange Board of India’s Business Responsibility and Sustainability Reporting framework for the top 1,000 Indian-listed companies to disclose quantitative and standardised ESG information.
Turning to AI
As investment opportunities and disclosures increase, so do the demands of asset managers and industry for solutions that efficiently structure information to measure, track and validate climate and other ESG risks. Unsurprisingly, there is growing interest in the role AI-powered solutions can play to increase the speed and efficiency of the ESG disclosure process, especially in the post-ChatGPT world.
Use cases range from corporate reporting to investor analysis to regulatory oversight. AI-enabled providers span companies that provide ESG data, analytical services and software-as-a-service solutions. Such solutions help both issuers and investors comply with and track regulatory requirements. Investor-focused providers also offer solutions for company analysis, risk management, portfolio management and issuer engagement. For regulators, solutions centre on enabling oversight with a focus on understanding assurance and the type of disclosure framework in use.
While nearly all providers integrate AI into operations to provide services, for some companies like Risk Insights, inhouse AI, machine learning and big data analysis are central to their approach to developing ESG ratings and analysis. With a sub-Saharan Africa focus, Risk Insights provides analysis on all companies listed on major exchanges, such as in South Africa, Nigeria and Kenya.
RepRisk is another example of a company that has strategically combined AI, advanced machine learning and data science capabilities with human intelligence to identify material ESG risks for companies, real assets and countries in developed economies and EMs. The company uses data science and advanced machine learning alongside highly trained analysts to develop insights for users. RepRisk has made its methodology and data science models available for use by others.
Experiments with machine learning and data science
In December 2023 the World Bank’s first AI as a service, MALENA, was released for use publicly. The result of the International Finance Corporation’s inhouse experimentation with machine learning and advanced data science, MALENA is a suite of AI-powered solutions designed for EMs. The solution has been trained using IFC’s many decades of sustainability and impact data and ESG expertise.
Using natural language processing, MALENA can read text, identify over 1,000 climate, gender and ESG terms, and generate insights from sentiment analysis. While general-purpose AI models have been criticised for failing to understand ESG context, MALENA’s training on domain-specific vocabulary and data has led to high performance and accuracy of results. Users like asset managers are screening target issuers and portfolio companies with MALENA.
However, it has been the recent advances in large language models that have taken the world by storm. Open AI’s GPT suite, Meta’s Llama family and Mistral AI have really ignited the imagination. These solutions provide better quality text summarisation and content interpretation and analysis. For instance, with model prompt refinements and retrieval-augmented generation techniques, these models are very effective at finding answers to questions.
Inhouse experiments with finding answers from company disclosures to ESG-domain questionnaires have seen steady improvements in AI model accuracy and better-quality answers. Not only can users quickly receive answers to tailored questionnaires, but importantly such applications can highlight what information companies have not disclosed (that could be material to their sector) as well as errors or anomalies.
For regulators and investors such applications speed up the approach to scoring companies, developing engagement strategies and peer-to-peer benchmarking. This in turn leads to faster and more efficient reviews, quicker project approvals and better resource management.
Atiyah Curmally is Principal Environmental Scientist, International Finance Corporation.
This article featured in the Sustainable Policy Institute Journal, Q3 2024 edition.
