Business case for leveraging AI for carbon accounting
Category: Latest News
Upload Date: 10/7/2024
Alongside Permutable.ai, we explore the advantages of employing AI techniques for carbon reporting and accounting.
At Carbon Responsible, we’re always discussing new technologies and tools that have potential to provide efficiencies and savings for our clients– whether it be cost or process based.
Generative AI (GenAI) technology, for example, could support efficient data automation involved in some aspects of carbon accounting.
Our Carbon Data Lead, Octavia, sees GenAI as a powerful aide but not a turnkey solution and explains some of the associated benefits.
The benefits of AI in Carbon Accounting
Carbon Accounting is a framework for measuring, recording, and reporting the amount of greenhouse gas emissions associated with an organisation’s activities.
Data is king here – there are hundreds of data points involved in carbon accounting so leveraging AI to transform data collection and optimise the entire process is a powerful prospect.
One company that has been pushing the boundaries in this space is Permutable AI. They offer real-time AI-driven sustainability data intelligence including news sentiment analysis around company emissions tracking, net zero targets and initiatives. They have also developed an innovative framework GreenProof to analyse the feasibility of net zero targets and identify greenwashing.
They have also been working towards improving the accuracy of reported company carbon emission estimates using advanced machine learning techniques. They achieved an accuracy of carbon emissions modelling for different emission scopes at a level of 68 – 99% better than using country averages.
Permutable AI’s vision is to develop of a carbon emissions prediction calculator using their advanced technologies and securing necessary funding to realise this vision to help more accurately plug the gaps in corporate emissions reporting.
The primary objective is to refine their predictive model using similar machine learning techniques that have proven successful in increasing the accuracy of carbon emission estimates. This involves expanding the dataset upon which it is trained, enhancing their algorithms, and continuously testing the model to ensure its reliability and accuracy.
The potential of AI-driven sustainability data intelligence, as demonstrated by leading companies like Permutable AI, highlights the transformative impact such tools can have on enhancing transparency and accountability in carbon emissions reporting.
CMO Talya Stone of Permutable AI highlights the importance of innovation in this area, stating, “By leveraging real-time data, sophisticated analytics, and machine learning models, we have the potential to deliver a robust solution that predicts corporate carbon footprints with unprecedented accuracy. This initiative has the ability to empower businesses to meet their sustainability goals more effectively and align more closely with global environmental standards.”
How does this benefit our clients in practice?
We identify a main benefit being improved supply chain emissions reporting, thanks to machine learning.
Reporting supply chain emissions is challenging for several reasons. It’s difficult collecting data from suppliers due to low response rates and suppliers not calculating their own emissions. This is compounded by the fact supply chains can be very long spanning into 100s of companies.
For companies that are just starting to calculate supply chain emissions, we can use machine learning to heatmap their entire supply chain very quickly. This can inform/pinpoint where further data collection efforts might be taken.
For our mature clients, these ML estimates can be used to fill in data gaps to create a more complete picture of their supply chain which is useful to year-on-year comparisons.
What’s the implication for Heads of Sustainability / Data Teams etc?
ML methods increase the complexity of the data we are using to calculate corporate emissions. Heads of Sustainability/Data Teams will have to navigate the intricacies of using more complex data.
The explainability and transparency of the models used to estimate corporate emissions will have to be understood by data teams. Different providers will use different models and to select the best provider sustainability teams might have to understand the credibility and quality of the datasets used to train the model, or the performance of the model against other estimation methods.
Understanding how these estimates are generated is also key for transparency and the ability to communicate confidently to stakeholders and other members of the organization. This is particularly important if these estimates are used to inform strategic decision making or assess risk.
The need for human-backed experience – why AI can’t compete
ML methods have the potential to support carbon measurement, however, to obtain the highest quality data this still requires data collection and measurement.
Looking to the future. Maintaining awareness of future regulatory scene & how ML might fit into this.
Human backed expertise to navigate scenarios where ML methods may not have been explicitly considered. For instance, it becomes necessary to determine how ML-estimated data fits into existing data quality scales, such as the PCAF (Partnership for Carbon Accounting Financials) scale, and to anticipate how this data might be perceived during audits.
Given the regulatory scene is rapidly changing (e.g. FCA ESG Ratings) this is something that we can guide clients on.