Companies are usually overloaded with data that they may not know how to take advantage of. On the other hand, artificial intelligence (AI) techniques are known to “keep learning” as the data increase. In this context, our research question emerges: what AI-based methods, in the literature, could be used to automatize business processes and support the decision-making processes of companies? To fill this gap, in this paper, we performed a review of the literature to identify these techniques. We ensured the usage of methods since they allowed reproducibility and extensions. We applied our search string in the Scopus and Web of Science databases and discovered 21 relevant papers pertaining to our question. In these papers, we identified methods that automated tasks and helped analysts make assertive decisions when designing, extending, or reengineering business processes. The authors applied diverse AI techniques, such as K-means, Bayesian networks, and swarm intelligence. Our analysis provides statistics about the techniques and problems being tackled and point to possible future directions.

Delays in legal proceedings significantly impact both corporate finances and individual livelihoods. Traditional methods for managing these delays typically rely on subjective assessments of what constitutes a reasonable process duration. This study explores a more precise approach by integrating machine learning and process mining techniques to enhance prediction of legal proceedings’ overall time. Diverging from previous works that either utilized machine learning analysis or process mining in isolation, this research synergizes these approaches. We applied process mining clustering techniques to over 60,000 cases from Brazilian labor courts to segment cases based on their procedural movements, creating clusters. These clusters, along with other procedural characteristics, such as case subject, class, and digital status, were then incorporated into a feature set for regression modelling. We employed linear regression, support vector regressor, and gradient boosting based methods to develop models that predicted case duration. The gradient boosting model demonstrated the best performance with an $$R^2$$R 2 -score of 0.87. Furthermore, our analysis identifies time bands where the model performs better and employs explainable AI techniques to elucidate key features influencing case durations. The clustering features emerged among the most significant for the task. The proposed combined approach offers a comprehensive method for analyzing and forecasting legal case timelines and also shows the potential of process mining clustering techniques to improve the analysis.

Delays in legal proceedings can significantly harm both corporate finances and individual lives. While many jurisdictions rely on subjective criteria such as reasonable process duration to manage case processing times, machine learning offers an alternative approach by estimating lawsuit durations and assessing the impact of multiple variables on these timelines. This study explores the use of machine learning to estimate lawsuit duration, considering multiple variables, including control-flow characteristics. We observed that the sequences of procedural movements exhibit complex patterns, making them suitable for analysis using clustering techniques employed in process mining. We framed the problem as a supervised learning regression task, incorporating clustering results as additional features. Employing linear regression, support vector regression, and gradient boosting methods, we analyzed over 60,000 cases from Brazilian labor courts. The gradient boosting model achieved an R2-score of 0.87, and feature importance analysis revealed that clustering techniques features emerged among the most significant in time estimation.

O tempo exigido para resolução de processos judiciais impacta a economia, a confiança dos investidores em um país e a vida da população. No Brasil, princípios de boas práticas fornecem diretrizes referentes à duração razoável do processo. Entretanto, modelos de aprendizagem de máquina podem prever com precisão o tempo total processual e prover insights referentes a possíveis causas de ineficiência. Trabalhos da literatura fornecem soluções nesse sentido, porém não consideram as movimentações processuais realizadas no processo durante seu trâmite. Este trabalho utiliza modelos de aprendizagem de máquina treinados com características pautadas na sequência de movimentações processuais e características do tribunal responsável pelo processo. Investiga assim a habilidade dos modelos para a tarefa, além de considerar a importância das features envolvidas. Foram utilizados processos eleitorais de diferentes tribunais regionais, TREs, de estados do nordeste brasileiro. Os modelos foram capazes de prever o tempo total processual com alta acurácia em que o Light Gradient Boosting Machine (LGBM) obteve R² médio de 0.9. Os resultados sugerem a eficiência da abordagem e a exploração de técnicas de mineração de processos como diferencial para a tarefa.

A morosidade na tramitação de processos judiciais representa um desafio crítico para a eficiência do sistema de justiça. Este artigo apresenta uma metodologia que integra Mineração de Processos e Machine Learning para analisar o tempo de processamento de casos na Justiça Comum brasileira. A abordagem inicia com a aplicação do algoritmo K-means, utilizando a representação bag of activities, para agrupar processos com fluxos de trabalho similares. Posteriormente, esses aglomerados são subdivididos com base na duração dos traços, classificando-os em categorias de execução rápida, média e lenta. A etapa final emprega o algoritmo PrefixSpan para identificar padrões sequenciais de atividades recorrentes dentro de cada perfil temporal. Os resultados demonstram comportamentos processuais característicos e sequências de atividades predominantes para cada faixa de duração, fornecendo subsídios essenciais para a otimização da gestão judiciária e o aprimoramento dos fluxos de trabalho nos tribunais.