ESG data: Is AI the final frontier?

Where are we now with the ESG data

To date, regulatory bodies have been a significant driver of ESG data demand. However, there is an observed discrepancy between the internal demand driven by regulations and the actual influence of ESG data on decision-making within investment firms. This means the regulatory push has led to a cottage industry of companies which produce ESG reporting, with a significant increase in firms disclosing their ESG metrics, but this is focused more on regulatory compliance than intrinsic corporate interest.

Despite the abundance of ESG data, there are still substantial reporting challenges for companies and asset managers which mainly relate to the inherent complexities of measuring and modelling environmental and social risks.

This situation is starting to change. Stakeholders, including regulators, investors, and companies, now recognise the need to address environmental and social risks systematically and are looking for more than just ‘tick box’ metrics from their reporting. This means there is an increasing focus on objective assessment, measurable outcomes and actions taken based on the reporting.

Despite the abundance of ESG data, there are still substantial reporting challenges for companies and asset managers which mainly relate to the inherent complexities of measuring and modelling environmental and social risks. The scale and scope of these risks make it difficult to achieve perfect data, necessitating continuous engagement and improvement in data collection and analysis methods.

How can qualitative data help?

Both quantitative and qualitative data are necessary for a comprehensive understanding of ESG risks. Quantitative models, which have to date been the preferred method, provide measurable outcomes, while qualitative assessments offer context and narrative scenarios that help illustrate the real-world implications of the data. This dual approach enhances the ability to make informed decisions and develop effective strategies by making the risks and necessary actions more tangible. For example, the Green Finance Institute’s Assessing the Materiality of Nature-Related Financial Risks for the UK report suggests that the deterioration of the natural environment could lead to UK GDP falling by 12% by the 2030s. However, illustrations of the implications of this drop are what can motivate decision-makers to take proactive measures.

Alongside inclusion of qualitative data in analysis, consideration should be given to how to integrate both top-down which starts by taking a broad view of the market and economic situation and bottom-up data which focuses on company specific metrics. Organisations should assess their current capabilities and gradually build the necessary systems to collect, analyse, and act on both these types of data. An iterative approach ensures that data-driven strategies evolve and improve over time. By focusing on integrating both of these approaches it will ensure a comprehensive understanding of ESG risks.

How could ESG data innovation benefit investors?

Focusing on improving data quality across the investment industry will enhance the credibility and effectiveness of ESG strategies. By working towards collecting and using data that can be verified and audited this will increase trust in metrics and help investors to make informed decisions. This will aid ESG to become a core risk management consideration.

Pictured L-R: Emily Forsyth-Davies, Dr Jimena Alvarez and Michael Sher

Focusing on improving data quality across the investment industry will enhance the credibility and effectiveness of ESG strategies.

Accurate measurement of resources, such as water and energy usage, helps to provide actionable targets that make it easier to assess outcomes. By focusing on reliable and standardised measurements, organisations can make more informed decisions to improve efficiency and reduce environmental impacts.

For example, as satellite technology use improves, this can aid the identification and evaluation of environmental risks. This is particularly useful when looking at proximity of company operations to biodiversity hotspots and water sources, and the impact on fragile ecosystems.

Another option is to implement pricing mechanisms on natural resources. By assigning a cost to environmental impacts and negative externalities, such as carbon emissions or water usage, it becomes possible to drive behavioural change through economic incentives. This approach aligns with the principles of capitalism and can encourage more sustainable practices.

Where does AI have the potential to positively impact ESG data?’ There are a number of ways developments in AI can impact ESG data:

1. AI-powered remote sensors and drones can be used for environmental monitoring to provide granular and transparent information at much lower costs than biological field work. For example, drones can be used to measure soil organic matter or water usage, providing detailed and precise data that can inform ESG strategies.
2. ESG data analysis – AI large language models can be used in detecting greenwashing where metric and targets are used inconsistently, and in analysing large datasets. In particular, by identifying discrepancies and inconsistencies in reported data this helps to drive more accurate and reliable insights.
3. ESG data collation, specifically aggregating research as large language models currently perform best in textual analysis. As this technology develops it will be an impressive time saver, significantly reducing analyst time. However, it is worth noting that at the current time this work still needs extensive human checking due to inaccuracies and ‘AI hallucinations’ which can creep in.

Despite the potential benefits of AI, the energy and water usage required to power current AI systems must be considered. For context, large AI models like ChatGPT, are often trained and deployed on power-hungry servers, which can consume a few kilowatts each, the equivalent of the average power consumption of an entire house. In a middle-ground scenario, by 2027 A.I. servers could use between 85 to 134 terawatt hours annually. That’s similar to what Argentina, the Netherlands and Sweden each use in a year, and is about 0.5 percent of the world’s current electricity use^[1]. In addition, Global AI requires significant water usage for both onsite server cooling (scope 1) and offsite electricity generation (scope 2), with scope 1 & 2 water withdrawal in 2027 is estimated to be between 4.2 and 6.6 billion cubic meters, which is roughly 4-6 x the annual water withdrawal of Denmark^[2].

So as this technology continues to develop there is an expectation that there will be a focus on ‘green’ or more efficient algorithms. There is also a definite need for caution in using AI for ESG data. AI-generated insights must be verified and contextualised by human experts to ensure their accuracy and relevance. This careful approach will maximise the benefits of AI while minimising the risks of misinformation.

Participants