Since the establishment of the Paris Agreement in 2015, 196 countries have adopted a legally binding treaty addressing climate change. This treaty seeks to limit global temperature increases to well below 2℃ above pre-industrial levels, with efforts focused on capping the rise to 1.5℃.
Tracking emissions and making a fair contribution to environmental efforts has become simpler for businesses, thanks to AI technology that facilitates data collection and actionable insights.
One of the main difficulties in environmental, social, and governance (ESG) reporting is the vast amount and range of data organisations need to handle. A significant challenge AI-powered climate tech platforms faced was figuring out how to build their own small language model (SLM) on top of existing large language models (LLMs), compared to traditional ways to retrieve data and make insights.
Sprih, a cloud-based sustainability software with an AI platform, SustainSense, approached this challenge with a human-centric methodology.
In conversation with AIM, Akash Keshav, CEO and co-founder of Sprih, said, “We aimed to replicate how the human mind processes this data in our SLM.” The company successfully translated around 1,200 different units into a single unified representation. “When I say ‘units’, I mean various ways these units are represented—some reports may use ‘t CO2’ while others might say ‘MT CO2’.”
Predictive Analytics and Interpreting Complex Data
As per a KPMG report, accuracy and reliability are essential for ESG reporting, especially considering changing regulations in different regions. AI demonstrates its strength in this domain by streamlining data validation processes. AI-driven systems can cross-verify information, authenticate ESG data, and detect discrepancies or irregularities.
Fitsol is a SaaS platform designed to tackle the challenges of collecting and integrating ESG data from various inconsistent sources. It supports multi-format data ingestion, including CSV uploads, ERP integrations for systems like SAP, Oracle and Zoho, API connections and email parsing, as well as utility bills, logistics invoices, procurement sheets, and IoT sensor data from industrial sites.
A Fitsol spokesperson told AIM that their platform includes a data confidence scoring layer that flags gaps, duplicates, and inconsistencies before analysis begins. “We also support different partners by providing training and AI customer support to clear their queries in real time and help them with data collection and filing.”
According to Sprih, reports are prioritised automatically in a regulatory context, with discrepancies flagged when two reports present differing numbers. SustainSense’s research team is notified of these conflicts, leading to human intervention to prevent future issues. “Our model keeps getting trained with the intervention of the research team and the AI team, even though we sort of have a human-driven validation process, something called batch processing,” the CEO added.
Risks, Challenges and Sustainability in AI
Predictive analytics enables organisations to foresee risks and opportunities associated with ESG factors, ESG Today reported. AI can forecast how climate change might affect a company’s operations, supply chain, and market demand.
“We pull the information related to the initiatives or actions any particular sectors are taking, which becomes our recommendation starting point for our customers. However, every customer is at a different stage. They have distinct budgets. So we have a planner tool where customers can go and see whatever recommendations come out of the system for that particular industry,” Keshav emphasised.
Ultimately, it is up to the companies to act on these recommendations. They can increase their investment in specific recommendations or activities and assess the impact on the return on investment (ROI) from those actions, which can be incorporated into AI systems.
According to Fitsol, one of its key differentiators is its data integrity engine, which uses AI to identify anomalies in emission reporting.
“In one case, a client reported unusually low packaging emissions. Kyoto flagged this through material-density mismatch detection—it detected that plastic packaging was incorrectly logged as paperboard, underreporting emissions by 70%. The client was able to correct this before submitting their CDP report.”
In another scenario, Kyoto’s platform identified route inefficiencies in logistics data that weren’t visible in spreadsheet-based audits. The fix led to an 8% improvement in carbon efficiency per ton-km.
ESG data presents significant challenges, requiring considerable energy as it is collected from various parts of the organisation. This could create a double standard when evaluating other companies’ ESG targets. Hence, it is important for AI-driven insights to manage energy consumption effectively.
For instance, SustainSense uses a configuration-driven, reusable component approach to avoid duplicating code in codebase. A widget component is utilised throughout the system, saving significant storage space and reducing compute and energy consumption.
Keshav highlighted that in their AI processes, the company efficiently chunks data. “We send only five relevant pages instead of a 500-page report to the LLM. This practice reduces processing data, saving energy and costs.”
Fitsol manages energy consumption by using Cloud-native infrastructure on AWS with auto-scaling and low-latency compute workloads, reducing idle compute and excess energy use. It relies on batch processing for ML model training and inference, minimising high-frequency compute spikes.
“We’re exploring green cloud credits and partnerships with providers offering carbon-neutral server options, especially for expansion into the EU and MENA.”
Decarbonisation Efforts Beyond Just Compliance
Keshav identifies two groups of companies regarding accessibility initiatives. The first group is advanced, having completed assessments and now seeking action. The second group is just beginning its accessibility journey and needs a strong technical or AI foundation to obtain accurate data easily.
“The global economic paradigm is changing as companies are under tremendous pressure from people across the world to account for the social impact of their businesses. Moreover, endless growth with profit as the sole metric is no longer sustainable. The consequences of social and environmental imbalance are mostly seen in the long term. If allowed to go unchecked, the disruption caused may cause a significant dip in growth and corporate valuations,” said Namrata Rana, national head of ESG at KPMG India.
The KPMG report stated that domestic and international investors increasingly integrate ESG factors into their investment evaluations, alongside conventional financial analysis. In this approach, investment firms collect ESG data on various companies to inform their assessments of stock valuation and the associated risks.
Keshav also highlighted how individuals can use the platform to set targets and take action. Users can conduct macro analysis and utilise the marginal abatement curve to visualise their data, allowing them to assess the costs of various activities concerning their available budget. They can evaluate both short-term and long-term ROI.
Many of Fitsol’s enterprise clients are also moving beyond regulatory compliance into proactive carbon reduction using Kyoto. Their AI engine calculates emissions across Scope 1, 2, and 3 and generates real-time decarbonisation recommendations.
“For example, a client in the automotive sector used our insights to shift 40% of their logistics network to EV vendors, reducing emissions by over 1,80,000 kg CO₂e. Another packaging client adopted biodegradable material substitutions based on Kyoto’s lifecycle assessment models, leading to a 12% reduction in packaging emissions.”
According to the company, these decisions are backed by data, modelled for costs, and calculated for ROI, enabling sustainability teams to act with precision rather than merely complying with checklists.
The post How AI is Becoming the Backbone of ESG Compliance appeared first on Analytics India Magazine.
