A mother waits for her child support case to be resolved. A worker awaits a ruling to get his job back. A person under investigation needs to know whether they will be allowed to defend themselves while remaining free. To the courts, these are case files; to those involved, they mean months of lost income, suspended plans, and sleepless nights. Justice should not be measured solely by the judgments it produces, but also by the amount of time it forces people to wait.
Latin America does not have a single judicial system, nor a single cause for its delays. There are differences in budgets, organization, territorial access, and caseloads. Yet throughout the region, a delayed ruling may recognize a right when part of the harm has already become irreversible. The question is how to speed up proceedings without weakening the safeguards that make a decision legitimate. The figures show the scale of the challenge. According to Brazil’s National Council of Justice, the country ended 2025 with 75.5 million pending cases, after receiving 40.9 million new cases during the year. Although the backlog declined, the congestion rate was 62.6%. These figures should not be extrapolated to the rest of Latin America, since they describe the Brazilian reality; they do, however, show that even statistical improvement can coexist with an enormous number of people waiting.
The problem extends beyond the region. For example, the World Justice Project’s 2025 Rule of Law Index recorded a deterioration in civil justice in 68% of the 143 countries and jurisdictions assessed. The indicator covers different dimensions, including delays, the effectiveness of alternative dispute-resolution mechanisms, and improper interference. This does not mean that 68% of trials were delayed; it means that the functioning of civil justice deteriorated in that proportion of countries.
Against this backdrop, artificial intelligence can contribute, but we must be precise about what for. It should not decide whether a person is guilty, whether they deserve to lose their job, or whether they should be separated from their family. It can, however, help judicial teams organize case files, identify duplicate documents, locate precedents, and flag deadlines that are approaching.
Imagine a case involving thousands of pages. A tool could construct a chronology of the proceedings and link each date to the original document. A lawyer could check whether a notification is missing; a judge could locate relevant evidence more quickly. No computer alert would, by itself, establish responsibility. Assessing the facts, hearing the parties, and providing reasons for a judgment would remain human obligations.
Brazil offers a concrete example: Victor, developed by the Federal Supreme Court to support the identification and classification of appeals involving issues of general repercussion. Colombia offers another experience: in its 2024 judgment T-323, the Constitutional Court examined a judge’s use of ChatGPT and addressed the safeguards needed to preserve judicial autonomy and the rights of the parties. Both cases make it possible to discuss the uses and limits of these technologies; neither authorizes us to promise that a program, by itself, will reduce the duration of all trials.
This is where an essential distinction emerges. Classifying a case file in minutes is of little use if the hearing still has no date. Quickly preparing a draft judgment does not give citizens back their time if the ruling remains pending signature or is never enforced. Improvement must be measured from the filing of a case through the decision and, where applicable, through its enforcement. I propose that courts adopting artificial intelligence publish four indicators: first, the total duration of proceedings; second, the number of people waiting beyond the statutory deadlines; third, the proportion of rulings that are actually enforced; and finally, the errors detected and corrected. The results should be compared with those of courts with similar jurisdictions and caseloads, so as not to attribute to technology improvements that actually stem from hiring more staff or reorganizing procedures.
The human dimension can also be quantified. If 5,000 people received a ruling 60 days earlier, the total waiting time avoided would amount to approximately 822 years. This is a hypothetical example, not a result achieved by any court. Nor does it represent additional years of life: it expresses how much less time people would spend waiting for an answer. Such a measure would make sense only if greater speed did not increase errors or limit the right to a defense. An incorrect judgment issued sooner does not constitute progress. For this reason, I propose that every technological project begin with a controlled trial, independent review, and publication of the results. An initial target could be to reduce by 20% the duration of selected administrative procedures over two years, without increasing the number of decisions overturned because of errors attributable to the system. This is a proposed target for evaluation, not demonstrated effectiveness.
Technology also does not replace investment in judges, court staff, public defenders, and infrastructure. A person who must travel for hours to attend a hearing needs effective access to justice, not merely a digital case file. And those without a stable internet connection must retain in-person alternatives for exercising their rights. The ultimate goal is not for institutions to showcase more modern software. It is for a mother to receive, in a timely manner, the resources needed to feed her child; for a worker to know whether they will get their job back; for a person under investigation not to remain indefinitely in a state of uncertainty. A timely judgment cannot restore everything that has been lost, but it can prevent someone from continuing to surrender years of their life to a wait that the system should be working to reduce.
Autor
PhD in Automation and Robotics. He has conducted research at various universities in France, Spain, and Ecuador on energy, technology, and development. His research focuses on the social economy, industrial transformation, and educational development.
