How are companies reducing emissions? An LLM-based approach to creating a carbon emissions reduction levers library at scale

Information on carbon emissions reduction levers is scattered across greenhouse gas disclosures and sustainability reports in dense free text, and no systematic, sector- and region-specific reduction-lever library exists for companies to draw on. We propose a multi-agent system that combines large language models with retrieval-augmented generation to extract, classify, and validate carbon-reduction actions from publicly available sustainability reports at scale. The result is a structured, queryable library of reduction levers that can guide corporate decarbonization planning.