Artificial intelligence (AI) can substantially reduce the time required to assess the financial impact of environmental and social risks, according to an article published by the Harvard Business Review (HBR).

The article, authored by Robert G. Eccles of the University of Oxford's Saïd Business School and Shivaram Rajgopal of Columbia Business School, examined whether widely available large language models (LLMs) could connect sustainability-related risks disclosed by companies with their financial performance.

Rather than producing another environmental, social and governance (ESG) score, the researchers tested whether AI could identify the environmental and social issues that companies describe as financially material, map them to specific line items in financial statements and estimate their potential effect on company value.

To assess the approach, the researchers analysed ExxonMobil's publicly available disclosures using four commercially available LLMs. They compared company-reported sustainability risks with financially material issues identified under Sustainability Accounting Standards Board (SASB) guidance and linked those risks to income statements, balance sheets and cash flow statements.

The analysis, which the authors said previously required around 100 hours of manual work, was completed in about an hour using AI, with some tasks taking only a few minutes.

According to the article, the reduction in time and cost could make financially focused sustainability analysis more accessible to investors, analysts and other stakeholders. Instead of relying on general ESG ratings, the method estimates how different sustainability-related scenarios could affect a company's financial position.

In ExxonMobil's case, the analysis estimated sustainability-related financial exposure of between US$5 billion and US$10 billion under a base-case scenario. Under adverse assumptions, the estimate increased to US$12 billion to US$18 billion.

The researchers also found differences between AI models. Gemini Pro 2.5 produced a higher estimate of US$25 billion to US$45 billion, highlighting variations in model outputs.

The authors said these differences show that AI should support, rather than replace, professional judgement. They argued that while AI can perform large-scale analysis more quickly, interpreting the results and assessing their reliability remains the responsibility of analysts and investors.

They added that faster analytical tools could widen access to financial sustainability assessments by reducing the time and resources traditionally needed to complete them. However, they noted that organisations using AI for investment analysis will still need appropriate oversight to evaluate model outputs and apply them in decision-making.