More than 90 percent of S&P 500 firms publish environmental, social and governance reports, and more than 600 rating agencies analyze the results. Despite mountains of research, the findings remain mixed on whether ESG goals produce superior returns.
Enter the bots. Increasingly, a specialized set of companies are applying machine learning and artificial intelligence methods to evaluate which firms’ ESG goals are driving improved results and which ones are mere window dressing.
"Ultimately, at the end of the day, it’s about taking ownership," said Emily Huang, chief executive of Idaciti, a company that specializes in machine learning-based ESG analysis, during a session at the inaugural GreenFin event. "This is where technology can help to make processes more robust, more systematic and more reliable over time."
The hope is that this type of rigorous, big-data analysis of everything from sustainability to LinkedIn comments can detect which practices are generating value and which are just getting in the way — and whether AI-driven research beats analyst-driven research.
Large corporations rely upon sustainability research and ESG ratings attract better investors and provide greater transparency to customers and clients. These corporations hire analysis groups, such as Idaciti, to produce and evaluate ESG data with the goal of pointing the company in a more sustainable, equitable direction.
Before the bots, ESG was analyst-driven. Human analysts would collect, process and analyze data related to climate change, safety conditions, human rights, corruption, compliance and more — a process that left room for both intuitive insights and biases in reporting.
"Even just a few years ago, we saw that corporate leaders really didn’t have access to useful risk information on ESG," Susanne Katus said during a panel discussion at GreenFin 21. Katus is vice president of brand and business development at Datamaran, which works with Fortune 500 companies to provide board oversight, monitor risk-management processes and determine the impact of various sustainability factors on financial performance.
Companies such as Idaciti, Datamaran and TruValue Labs are harnessing AI in an effort to optimize ESG analysis. Their emphasis on algorithms and machines are meant to reduce the inherent biases of human analysts which can often tilt corporate reports in unanticipated directions.
In analyst-driven reports, subjectivity can skew results and produce less consistent data than an AI-driven approach. Before AI, judgment calls based on limited anecdotal evidence possibly could lead to slanted or inconsistent data over time. For the reports made by AI, analysts develop algorithms to feed to bots which then apply these to broad-scale analyses.
The best solution may not be one of either-or, but of a middle ground.
Notably, AI has a far superior ability to mine through unstructured data. For example, some companies may provide reports based solely on information provided internally, while others include unstructured data, swaths of information mined by AI from sources outside the company, such as media outlets, advocacy groups and government websites. Mainstreaming the use of unstructured data in corporate sustainability reports presents an opportunity for greater transparency, fewer biases and more consistent standards.
"Artificial intelligence companies like ours use big data and modern computing technologies to analyze massive amounts of this unstructured content," said Hendrik Bartel, co-founder and chief executive of TruValue Labs, during the discussion. Bartel’s company caters to a clientele of asset managers and members of the banking community.
Computer algorithms have advanced to such a point that they can decipher the tone of a social media post, among other qualitative data markers. These machines can work more quickly than human analysts. Qualitative measures — often called "sentimental data" — can help gauge (and bolster) a company’s social accountability and customer impact.
Advancements in sentiment analysis, or opinion mining, allow firms to analyze the very language used by customers to describe the company. These datasets, not long ago, would have been overwhelming if not impossible to sort through for teams of human analysts, but now can be fed into artificial intelligence programs and analyzed in a fraction of the time.
The best solution may not be one of either-or, but of a middle ground, wrote finance and ESG program manager Mark Tulay. The ideal solution, though unproven, may lie in a side-by-side acceptance of analyst data and AI data together.
The data on ESG as a framework remains mixed: Prioritizing ESG doesn’t necessarily promise returns, but ignoring it entirely seems unproductive. Still, the added degree of accountability afforded by ESG data can create a framework that, long-term, could be as important as one based on bottom-line returns.
"To build something that understands ESG on the materiality level and on the widest lens possible... and now having technology help with that is only natural in 2021," Bartel said.