Artificial Intelligence in Brazilian Courts: Historical Evidence, a Proposed Framework and Effectiveness Index

Background: This theoretical-empirical study aimed to evaluate the effectiveness of artificial intelligence (AI). In the public sector, AI-based systems are expanding, yet empirical evidence on courts remains scarce. This study proposes and applies a multidimensional framework and an effectiveness index (i – Eff) to assess AI in Brazilian courts of justice, documenting an early phase of adoption.

Methods: Sixteen semi-structured interviews with judges and IT managers were conducted across federal, state, labor, and superior courts (12 h 28 min; 87,794 words). Content analysis with lexical statistics was performed using IRaMuTeQ software and operationalized across five dimensions: environment, reliability, autonomy, technics, and results. The equally weighted score (max 100) is based on the presence of variables mapped to this multidimensional model.

Results: AI use concentrated on initial case screening and tax-enforcement workflows; fully autonomous systems were not observed – AI provided suggestions while judges retained final decision-making authority. Measurement of efficiency gains was uneven because many tools remained in the pilot/implementation phase. In the i – Eff synthesis, three courts reached medium, ten showed lower, and three presented no effectiveness.

Conclusion: Early AI in Brazilian courts plays a supportive, not disruptive, role. Strategic alignment and result-measurement remain challenging, while transparency, data governance, and training are pivotal to scale effectiveness: the framework and i – Eff metric offer a replicable approach for cross-court benchmarking and longitudinal tracking.