Companies use ESG metrics to assess and track their accountability and transparency in environmental, social and governance practices. Investor focus on ESG metrics has surged recently as social inclusion and sustainability become increasingly important. However, simply including these metrics in your annual reports isn't enough. A true commitment to ESG requires innovative solutions for effective strategising and measurement.
Ever since ChatGPT’s launch in November 2022, artificial intelligence (AI) has remained at the apex of innovation. While numerous reports explore AI's potential impact on ESG reporting, few offer practical guidance for implementation. What the Institute of Chartered Accountants in England and Wales (ICAEW) intriguingly describes as a ‘nascent marriage’ has broad applicability across ESG-focused industries.
In general, deploying AI for ESG purposes involves using AI for advanced analytics and insights (including real-time reporting). AI’s speed and efficiency mean it can deliver ESG-orientated insights faster (and more affordably) than an analyst.
For example, we’ve written about how AI can foster socially responsible lending practices. AI can support a mutually beneficial relationship with underrepresented borrower groups by screening candidates using alternative lending criteria. However, specific applications of generative AI, like AI-driven lending, are not applicable use cases for all firms seeking to improve operational efficiency through intelligent algorithms.
One of AI’s most immediate practical applications is reviewing portfolios and statistics in a human-like manner. The conversational chat interfaces of large language models (LLMs) – such as ChatGPT, Gemini and Claude – make it easy for analysts to enter questions like ‘Can you compare these ESG metrics between companies?’ Or, ’Are they using the same definitions and reporting standards?’ and receive near-immediate responses.
From an AI perspective, using publically available LLMs contradicts ESG’s core objectives. Let’s explore why. Through extensive testing, we’ve gauged that LLMs generate hallucinations (or fictitious information) at a rate of approximately one per page. The visibility of the hallucinations varies.
For instance, a missing data point would be fairly obvious, whereas a slipped decimal place might not be. Publicly available LLMs are unsuitable for critical data, including most ESG data. ESG reporting’s function as a ‘transparent means for companies to communicate their progress and performance’ becomes impossible when its data is run through an insecure public platform, compromising its accuracy and integrity.
In contrast, one of the more rewarding methods of harnessing AI for ESG data is enhancing the capabilities of a (pre-existing) data platform. Let’s say your company has a platform that collates materials for ESG analysis (e.g. social media posts, news articles and financial statements). Manually sifting through those materials to extract relevant parts is time-consuming and expensive. Considering the salaries of ESG analysts, plus their years of expertise, manual extraction quickly becomes inefficient. In contrast, a well-trained AI mechanism could sample all relevant information in seconds.
Of course, we must always consider the benefits of AI from an ethical perspective. Responsible AI encourages due consideration of the ethics of deploying AI. Note that most sources covering the responsible AI movement – including Microsoft, IBM and Google – often focus on preventing bias through governance rather than promoting sustainability. Of course, failing to mention AI’s environmental impact may be in their best interests, considering the exorbitant energy costs of training AI.
The environmental impact of AI training is a growing concern. However, there are two potential solutions on the horizon. Firstly, energy providers can mitigate the environmental impact of AI training by supplying clean energy. Moreover, AI users concerned about the environmental footprint of AI training should prioritise energy-efficient AI-powered solutions.
Overall, the relationship between AI and ESG is delicate. Though AI’s ability to unearth important insights and automate key tasks for more strategic (human) efforts may benefit ESG initiatives and reporting, its misuse may undermine these advantages.
Collating ESG data demands transparency and accountability, two attributes that may cease to exist after deploying a publicly available LLM. However, the evolving relationship between ESG and developing intelligent technologies makes using AI inevitable. EY describes their relationship as ‘symbiotic’. Of course, you can only achieve symbiosis (in this case) if AI can provide real value despite its environmental impact and tendencies towards bias and inaccuracies.
AI’s acuity and speed make it useful and beneficial when you deploy it conscientiously. As a starting point, leveraging AI to automate routine processes frees ESG specialists to intervene efficiently whenever an AI system identifies potential issues.
Evolution AI has developed a tool to facilitate access to ESG data by unlocking insights from financial statements. Financial Statements AI extracts data from financial reports, leveraging it to calculate key financial ratios. You can try Financial Statements AI for free - book a demo or email hello@evolution.ai.
Interested in getting insights about how developing AI will affect ESG? Follow us on LinkedIn and X (formerly Twitter).
