AI’s Role in Making ESG Data More Reliable
Investors need quality ESG data and automation at scale to enhance critical decision-making
Small and large asset managers alike are figuring out how they can harness AI to enhance their investment decision-making while managing and measuring ESG performance. AI could be the vital link in connecting the dots on the path to a more sustainable future. It can solve data challenges that investors and corporations wouldn’t be able to without technology.
“When we think about ESG data, we actually use that data in a variety of ways, everything from our bottom-up risk and impact assessment to our regulatory and other client reporting, to enforcing our compliance screens,” said David Klausner, ESG Specialist at PGIM Fixed Income, on a Responsible Investor webinar sponsored by Clarity AI. “I think this is why AI is so important in this process. We don’t have the capacity to handle all of this data. We simply couldn’t do our ESG processes at scale as a large institutional investor.”
The ability for AI to help automate ESG data functions and scale that across the portfolios of investors’ clients means that asset managers can spend more time focusing on their core roles instead of scrubbing through data. AI also enables smaller asset managers to access data they might not have internal structure to research.
For instance, two companies based in Florida may not be exposed to the same level of physical climate risk (such as from increased possibility of floods) when one has taken active measures to adapt by moving electrical systems up to a higher floor, and the second has not. This context, when built into investment models, can enhance decision-making. “I think AI plays a very important role in helping bring that information together,” David said.
Value of reliable ESG data
AI’s ability to focus on data quality provides a solution for the challenge of reliable data. Even data reported by companies can be error-prone either because of human mishaps or inaccurate interpretations. “We’ve been able to identify out of all the reported data, what data is reliable, and thus can be trusted, and what data might be subject to errors,” said Borja Cadenato, Director of ESG Products at Clarity AI. “Using statistical machine learning models to compare data of a company with data from previous years, or with data from other companies in the same industry, helps us identify what of those data points might be wrong.”
Beyond reported data, he pointed to a “huge gap” that needs to be filled by combing through all the data that is available. “AI can also make a difference there.” For example, Clarity AI leverages machine learning models to estimate emissions data for more than 30,000 companies, to provide investors with a more comparable and comprehensive view of emissions.
What makes a trustworthy ESG data model?
Building trust between the data provider and the data user is a first step for any technology in any industry, Nick Pelosi, Associate Director for Engagement at Federated Hermes, pointed out on the webinar. With AI, that trust starts with the workforce and ensuring that the data is free of bias.
“It’s very important to be able to explain the models for people to trust the numbers and therefore integrate them into their investment decisions,” said Robert Smith, Director of Machine Learning Engineering at Clarity AI. “For me, machine learning models are a way of systematically comparing an infinite number of different approaches, what we call features, that will ultimately drive the estimated value, with an objective function that chooses between them, and selects the best one,” said Robert. There is a statistical result that provides a confidence interval (a certain percentage level of knowing that the data is as accurate as it could be) offering an additional layer of transparency to strengthen our users’ trust.
“The idea is not to get rid of humans but to have them validate the models and a much smaller number of results. These are key aspects of building trust in AI-powered models. Part of it might also require a change in mindset in some sectors.”
Many ways to leverage AI for ESG
There isn’t a singular approach to AI. While ChatGPT and generative AI are most familiar to people nowadays, Robert noted that they only skim the surface of all AI approaches that are available to investors and corporations for solving ESG data challenges.
As more data becomes available, AI will allow for much more aggregation, and a greater ability to do analysis and surface insights to inform decisions. “We have Natural Language Processing models which read hundreds of thousands of news articles on a daily basis and are able to categorize them into one of 40 different categories and assign severity to them. We were able to show that companies that had a certain level of controversy led to a 2 to 5% underperformance after six months.”
From NLP (Natural Language Processing) techniques to satellite imaging for tracking emissions to biodiversity AI for calculating endangered species numbers, AI can widen or open the door for improved ESG data that informs business decisions.
Contact us to learn more about how Clarity AI uses artificial intelligence to support any sustainability-related need.