The AI4Justice initiative, led by Connecthink Innovation in partnership with the Centre de Telecomunicacions i Tecnologies de la Informació (CTTI), aims to tackle the task of drafting judicial sentences by introducing a Virtual Assistant. This assistant is designed to reuse information from previous rulings, significantly reducing the time judges spend on repetitive drafting tasks. Discussing the program’s objective, the team explains, “Our goal has always been to enable judges to focus on legal reasoning and decision-making rather than spending countless hours on administrative aspects of their work.”
Key Phases and Milestones
Since its launch, the program has achieved significant milestones despite challenges. Early in the development phase, the team worked closely with judges to tailor the assistant’s features. “The input from judges has been invaluable,” they remark, “particularly in designing a user interface that integrates seamlessly into their workflow.”
A cornerstone of the initiative has been the creation of an anonymization module, which uses advanced natural language processing (NLP) to redact sensitive information such as names and bank details. Reflecting on its completion, the team states, “Ensuring data privacy is non-negotiable, and this module represents a critical step in achieving compliance with regulations while maintaining trust with stakeholders.”
In addition, the development of a vector database and a knowledge graph has enabled the Virtual Assistant to perform semantic searches and provide contextually relevant information. “This infrastructure allows judges to retrieve precedents and insights quickly, making their work not just faster but also more accurate,” they note.
Overcoming Challenges
The development journey was not without obstacles. One of the main challenges was coordinating with judges, whose tight schedules often delayed testing and feedback sessions. “Judges are extremely busy, and finding time for meetings and pilot tests was harder than we anticipated,” the team shares. However, they adapted by extending timelines and adopting a more flexible approach.
Technical hurdles also emerged, particularly around the variability of AI models. Early prototypes struggled to detect certain patterns, such as specific Spanish names and IBAN numbers. “We realized that generic models weren’t enough,” they explain. “Switching to models trained on Spanish-language data and refining our algorithms was a turning point.”
Reflecting on their experiences, they add, “If we could go back, we’d focus on anonymization issues earlier and invest more time in benchmarking AI models before deployment. These steps would have saved us a lot of time and resources.”
Anticipated Impact
The project’s potential impact has generated excitement among stakeholders. Judges, in particular, have been vocal about its promise. “They’ve told us that the Virtual Assistant could transform how they work by automating repetitive tasks and enabling them to tackle complex cases more efficiently,” the team notes.
Beyond individual workflows, the assistant is expected to address systemic issues like case backlogs. “With over 10,000 pending lawsuits in Barcelona alone, this tool could significantly speed up resolutions,” they remark. “Scaling it across Catalonia and even other regions of Spain could create a nationwide impact.”
Lessons for Future Innovation
The AI4Justice experience offers valuable lessons for similar initiatives. The team emphasizes, “Engaging stakeholders from day one is crucial. Their insights helped us refine the program’s objectives and features, ensuring it meets real-world needs.”
They also highlight the importance of modular design, which allows for scalability. “We’ve built a system that can grow—adapting to different regions and use cases—without losing its core functionality,” they explain.
Finally, they stress the critical role of privacy. “From the beginning, we prioritized data security and regulatory compliance. This not only ensured stakeholder trust but also positioned the program for broader adoption,” they conclude.
The AI4Justice project demonstrates how collaboration, innovation, and adaptability can modernize even the most traditional systems, paving the way for a more efficient and accessible judiciary.
