Integrating sustainability into strategy: LLM-Based decision support from automotive sustainability reports

Companies struggle to keep pace with emerging, industry-specific sustainability trends, making the integration of sustainability into strategy increasingly demanding. As new European sustainability reports become publicly available, they create opportunities for within-industry benchmarking and ESG decision-making. Given their length and scale, AI-enabled decision support may make comparative analysis more feasible than manual review alone. This study compares alternative Large Language Model (LLM) configurations for sustainability-report analysis and evaluates their strategic decision-support value. Using all publicly available CSRD-compliant sustainability reports from European automobile manufacturers, the study tests retrieval-augmented generation (RAG)-enabled and non-RAG LLM approaches on strategic disclosures. Outputs are assessed manually using a literature-grounded rubric of decision-support requirements, with an emphasis on sense-making in complex, ill-structured tasks. The findings suggest that current LLM-based approaches are better suited to screening, disclosure detection, and initial thematic synthesis than to stand-alone strategic assessment of sustainability disclosures, for which expert verification remains necessary.