Advancements in AI are making it possible for businesses to assess the financial impact of ESG investments in a fraction of the time previously required, according to a new article published by Harvard Business Review.
In the article, AI Can Measure How ESG Really Impacts the Bottom Line, authors Robert G. Eccles of Oxford University’s Saïd Business School and Shivaram Rajgopal of Columbia Business highlight how they used four widely available large language models (LLMs) to analyse ExxonMobil’s public disclosures.
“The goal was not to produce another ESG score, and it was not to single out ExxonMobil,” they state. “It was to test whether AI could do something sustainability analysis has long struggled to do at scale: take the environmental and social issues a company itself discloses as financially relevant, map them to specific income-statement, balance-sheet, and cash-flow line items, and estimate how strong or weak performance on each would affect the company’s value.
“One of us had done this kind of analysis by hand before. It took about 100 hours. With AI, the same core work took roughly an hour; parts of it were done in minutes.”
More accessible
As the authors noted, this reduction in time – and associated cost – can make financially-grounded sustainability analysis accessible to a much wider range of investors and stakeholders.
The methodology used by the authors began with identifying the environmental and social risks a company already discloses in its annual reports, and comparing these with financially material issues identified in industry standards (such as those developed by the SASB). These risks are then linked to financial statement line items, allowing users to estimate how different sustainability performance scenarios could influence company value.
“The output is not a slogan or a score. It is the kind of analysis an investor can actually use,” they state.
Test case
In the case of ExxonMobil, for example, the analysis estimated sustainability-related financial exposure of between $5 billion and $10 billion under a base case, rising to between $12 billion and $18 billion under adverse assumptions.
However, as the authors noted, some LLMs arrived at a different figure – Gemini Pro 2.5 highlighted a figure of $25–45 billion, reinforcing the importance of human judgement in the process.
“Better tools raise the premium on the expertise needed to use them well; they do not remove it,” they note. “Still, the unlock is real, and it changes who can do the work. An analysis that was artisanal is becoming industrial – company by company, sector by sector.
“This has profound implications for the ESG ecosystem. We do not have the inside scoop on the deliberations underway inside the ratings firms, the standard setters, or the fund regulators, but the capability we are describing is being put directly into the hands of individuals at a pace much faster than these institutions can move.” Read more here.
