Given the challenging environment in which corporations make disclosures around their operations, including their ESG initiatives, it's unclear if the impact of artificial intelligence on this process will help or hurt
The impact that environmental, social & governance (ESG) issues will have on corporate reporting is beginning to gain a lot of attention. Ongoing challenges around the lack of standardization in disclosure standards, metrics used to measure progress, and the units of measure for specific ESG issues top the list of concerns — and these challenges exist for both investors and companies.
And adding generative artificial intelligence (AI) into the mix, though it comes with significant opportunities as well, is likely to complicate things further.
Investors are increasingly demanding more transparency and data about how companies are utilizing and investing in generative AI. It’s impossible to base assessments on assumptions or on information that may not be accurate or available. Essential information about how companies are implementing AI needs to come directly from the companies themselves.
The good news is that the emergence of machine learning and generative AI offers substantial remedies to the challenges around data and reporting. ESG strategy expert John Friedman says these challenges around AI data collection is similar to those faced when companies are tasked with collecting greenhouse gas emission data.
“If you think about your greenhouse gas emissions associated with your electricity use of your facilities, getting the data requires 12-months’ worth of data, multiplied by the number of facilities that you have,” Friedman explains. “Then, you have to map where each facility is located and the electricity mix because there are different emissions profiles across the U.S. based on how electricity is generated in that region.”
The emergence of machine learning and generative AI offers substantial remedies to the challenges around data and reporting.
Based on this example, it is easy to see how the data for this one metric gets unwieldy fast. However, the advantages are also immense. AI and machine learning have this tremendous promise of being able to pull this data in real time, identify anomalies in the data through pattern detection, and pinpoint findings that need to be reviewed and audited. The combination of savings in time and effort across many functions and individuals is noteworthy.
And for those who think AI will completely take over ESG strategy execution and data management, Friedman notes that this is unlikely. “There will still be a place for the people with ESG expertise,” he adds. “The inputs into the analysis, insights, and findings need to be reviewed and audited.” Simply put, AI produces efficiency by making manual work easier, and people with specific expertise are still necessary to ensure the data set is complete.
Use of AI in corporate ESG reporting
Corporate reporting continues to be a headache for many companies and their investors who may want to gain insight from corporate disclosures. For example, there remains a myriad of disclosure standards and frameworks with no consensus on the metrics that are used to demonstrate progress nor a standard approach to measurement. Indeed, some frameworks require a quarterly metric while others require measurement annually.
Jonathan Ha, CEO of Seneca ESG, describes what he hears from customers. “ESG by nature is so subjective and what ESG data disclosers really want is agreement on one set of questions,” Ha explains. “It’s cumbersome to be asked similar questions at different points in time throughout the year.”
While consensus on one ESG disclosure framework is still far off into the future, Ha shared how Seneca ESG is addressing the challenge using large language models (LLMs) to build an AI-powered ESG assistant by incorporating a similarity score for questions that are alike across disclosure frameworks.
AI and machine learning have this tremendous promise of being able to pull this data in real time, identify anomalies in the data through pattern detection, and pinpoint findings that need to be reviewed and audited.
For example, as a data discloser is working on ESG reporting within a company, the ESG assistant proactively advises the discloser as to which are the similar questions from other selected frameworks, to improve efficiency and accuracy for those involved in corporate ESG reporting across the organization. The ESG assistant also provides recommendations on how to improve the responses to ensure accuracy and quality of answers.
To ensure that the AI-powered ESG assistant is producing the best guidance as possible, Seneca is testing generative AI approaches to understand the best prompts that produce the most optimal results based on experienced practitioners and thought leaders in the area of corporate ESG reporting.
How questions are phrased — known as prompt engineering — is a method of using natural-language processing in which the output of the models, such as ChatGPT, is guided and influenced by the design and structure of the input prompt. The goal of prompt engineering is to optimize how users ask the model a question to achieve the most accurate output because the way questions are phrased matters to ensure accuracy and relevancy in results.
Generative AI offers both powerful opportunities in solving for ESG challenges around corporate reporting. In the short-term, the lack of clarity and consistency on AI reporting and ESG disclosure standards make progress slower than most companies and disclosure experts want. However, once the standards are clarified, generative AI will provide a strong pathway for ESG reporting to be truly integrated with financial reporting, enabling a future in which, paraphrasing Friedman’s words, ESG will one day be called just business.
