How can you use AI to improve your sustainability data quality and accuracy?

The perspective added below is a "wishful" list probably generated by ChatGPT. :) We are getting started with integrating AI into ESG data collection and analysis. The most widely apply use case at the moment is application of NLP for data gathering, cleansing, grouping and pattern recognition. Pattern recognition functionality also allows to set certain triggers to establish controls around data quality.

Automated data collection and processing, reducing errors. Validation and verification, identifying inconsistencies and ensuring accuracy. Predictive analytics, enabling proactive risk management and resource allocation. Natural Language Processing for extracting insights from unstructured data. Automated reporting, improving efficiency and standardization. Continuous monitoring and feedback for real-time sustainability management. AI-powered decision support systems, facilitating data-driven decision-making. However, you still need to use common sense, no computer can replace that...yet...

Based on my experience, here are my key points of AI contributing to data collection and validation: Utilizing AI for automating the collection and validation of data from diverse sources like sensors, invoices, surveys, and databases. Working towards "white box" AI solutions to rely on causal interference algorithms, for example, for data extraction to enhance the speed and accuracy of data capture. While "traditional AI" is great at identifying anomalies, outliers, missing values etc. causal AI based models focus on limiting the analyzed parameters to most impactful ones, which is a great enabler in optimizing computational efficiency of AI model training and data selection / collection.

NLP allows to replace a manual and tedious task of processing huge amounts of ESG data to draw insights. The results can feed into a word cloud, Tableau and Power BI for visualization of findings.

I see a great potential of AI in automating and enhacing the accuracy of LCA calcuations, making them more cost-effective and reliable - AI has potential to streamline the entire process of LCA from data collection to analysis and interpretation. AI allows for more frequent updates of LCAs based on regularly updated and trained data. In acquiring analysis and insights I think the big potential is also in reducing complexity of benchmarking sustainability erformance against industry standards, such GRI, TCFD, CSRD, ISOs, Science-Based Targets etc.

At this stage I wouldn't bet on AI to do your company's ESG reporting! You obviously need a human engagement and close monitoring, as you prepare your report with the help of AI. AI can help you to prepare for the regulatory ESG reporting. It can review and analyze legislation, highlighting and comparing important requirements points.

Data preprocessing in sustainability context is often considered highly complex, so AI can definitely help streamline the processing of data from various structred or unstructured sources as well as personalizing the reporting dashboards to different shareholders. LLM-based AI models can also work like data analysts or scientists, explaining the intricaties and complexities of data in understandable format.

AI can certainly process both numerical and text ESG data collected from multiple organizational units. If a company runs a global operation it might be gathering its environmental (carbon, energy, waste) data from thousands of facilities and sites. AI can cleanse and stabilize these data and draw valuable insight for further decision-making

Acquiring data for sustainability reporting, LCAs or impact assessments is often a very complex process requiring data across digital and physical sources from material and energy use to cloud platforms, sensors data etc. Not forgetting contextual / qualitative business & subject domain data to validate the quality of insights. AI can be the integrating connection point across many sources of data to retrieve contextual insights fast and effortlessly, reducing complexity and time used to retrieve data.

The claims made above are a little premature as many of the AI capabilities around ESG data controls testing and responding to change are currently being tested but not yet implemented at a larger scale.

In context of sustainability or other use cases where ethics & responsibility is considered; To improve data feedback loops related to AI training, we should incorporate white-box AI solutions like causal interference machine learning aiming to explain the causalities and interdepencies of parameters and their relation to outputs, leading to more explainable and reliable use of AI models. The AI industry is full of noise and sustainability agenda should be to cut through it. Especially us in the EU should work towards making this our key point for thought leadership in sustainable use of AI.