Leveraging LLM agents and knowledge models for automated legal reasoning

Abstract

Delivering judicial decisions demands interpreting complex texts, analyzing evidence, and reasoning through jurisprudence and law. Recent advances in generative artificial intelligence, particularly large language models (LLM), have shown potential to automate parts of this process, but practical benefits in real-world courts remain limited. This paper introduces SARA, an LLM-powered platform deployed in a Brazilian court, which demonstrates significant efficiency and quality gains through the integration of LLM agents with knowledge models on jurisprudential and basic legal concepts. SARA extracts key elements from documents, including claims, requests, and evidence, and generates reasoning grounded in retrieved precedents. The Jur-KG, modeled through an ontology of concepts such as LegalRelation, LegalGrounds, and LegalClaims, enables semantic matching and retrieval of relevant case law. By representing cases according to the Legal Case Ontology for the Brazilian Judicial System, SARA supports traceable reasoning and addresses competence questions to assess coverage, coherence, and justification of AI-generated outputs. Deployment results show measurable improvements in processing time, consistency, and explainability while ensuring compliance with Brazil's National Council of Justice guidelines. Combining LLM-based agents with domain-specific knowledge graphs yields innovative capabilities and proven impact in judicial decision-making.