Agentic AI Aims to Cut Brazil’s Legal Bottlenecks Over 13 Years

Brazil’s 80.6 million-case judicial backlog shows how procedural inefficiency has become an economic risk, delaying legal certainty and investment. For Mexico, where court congestion and fiscal limits coincide with nearshoring-driven disputes, Agentic AI offers a scalable path to accelerate resolutions without expanding public payrolls. The shift directly affects the judiciary, legal services, cloud providers, and investors, while increasing the relevance of AI governance, transparency, and regulatory oversight.
 

The Brazilian Judiciary resolved 44.8 million cases in 2024, notable but insufficient against a backlog of 80.6 million pending records. The implementation of agentic AI positions as a technology capable of reducing the estimated resolution time from fifteen years to only two. This transition transforms legal operations from simple digitalization to autonomous agency.

The efficiency crisis in the Brazilian legal system does not result from a lack of human effort. Instead, it stems from the obsolescence of traditional operating models facing a volume of demand that exceeds manual processing capacity. "The question that the state and private initiative must answer is not how to work more, but how to work differently," says Gustavo de Paula, Country Manager, Xertica.ai.

This approach emphasizes that the human productivity ceiling has been reached. Expanding the legal machinery through hiring is no longer fiscally sustainable, as the current public budget allocates approximately BRL 689 (US$125) per inhabitant for system maintenance.

The current situation of the Brazilian Judiciary is defined by a technical contradiction. The system operates with a positive annual balance, resolving 44.8 million processes in 2024 compared to the 39.4 million that entered.

However, this surplus of 5.4 million processes per year is marginal compared to the accumulated stock of 80.6 million records. At the current pace, the total elimination of the inventory would only be reached in 2040.

This operational inefficiency transcends bureaucracy to become a macroeconomic risk factor. Legal proceedings in Brazil have an average completion time of seven years, which creates an environment of legal uncertainty. This climate inhibits foreign direct investment (FDI) and stops capital flow in various sectors of the economy.

The "Justice in Numbers" report from the National Council of Justice (CNJ) confirms that the diagnosis is critical. If the exclusive reliance on manual effort continues, citizens and companies starting processes today will receive a final judgment in fifteen years. Pressure on the public budget prevents the expansion of the payroll of magistrates, forcing a search for technology to alter the temporal resolution curve.

From GenAI to Agentic AI

While Generative AI (Gen AI) focuses on creating content based on statistical probabilities, which can lead to "hallucinations" or inaccurate data, Agentic AI operates as an autonomous agent within a closed and controlled technical ecosystem. 

“(Agentic AI) is capable of analyzing stock, identifying patterns in millions of repetitive processes, and preparing decision minutes that strictly comply with case law, allowing the judge to act as a strategic reviewer rather than a data operator,” says de Paula. The new paradigm, proposed by the Agentic Justice research from Xertica.ai, focuses on fair decision-making. 

Technical Differentiation of Agentic AI

The main characteristics of agentic AI include:

• Massive data analysis: The ability to process the total inventory and identify patterns in millions of repetitive processes.

• Respect for jurisprudence: The generation of draft decisions that strictly align with established legal precedents.

• The role of the magistrate: Technology does not replace the judge. Instead, it positions the judge as a strategic reviewer, freeing the individual from data processing tasks to focus on equity and legal complexity.

The viability of this technology has been demonstrated in specific implementations, such as in the Ministry of Environment of Minas Gerais (SEMAD). After integrating AI systems, the response time in institutional processes decreased by 83% within a few weeks, showing that automating non-strategic processes allows government organizations to recover their operational and strategic capacity.

The architecture behind these solutions relies on high-reliability models. Xertica.ai, founded in 2016 and recognized as a Google Cloud partner in Latin America, developed the X-Factor model with support from the MIT G-Lab. This model facilitates the adoption of cloud technology to accelerate value creation and productivity in critical sectors such as justice, health, and education.

With operations in eight countries and estimated revenue of US$120 million for 2024, the organization applies solutions ranging from monitoring risk zones to advanced legal data management. The corporate mission focuses on the responsible use of technology to generate concrete positive impacts on the lives of citizens, transforming organizational structures through the cloud and specialized consulting.

The massive implementation of technological agency is the only identified path to meet the goal of anticipating 2040 to 2026. By automating repetitive tasks, the Brazilian Judiciary can process the accumulated inventory of 80.6 million cases exponentially faster than manual methods. Agentic technology is no longer an aesthetic modernization option; it is a structural necessity for the sustainability of the rule of law and economic stability in Brazil.

“AI offers potential to address human limitations, particularly in eliminating cognitive biases and emotional influences that often affect human judgment,” reads a National Library of Medicine (NIH) article on the topic. “These advantages suggest that its use in courtrooms is not only inevitable but also essential for achieving fairer and more efficient trials.”

However, the NIH highlights that these benefits must be balanced against the unique AI challenges in judicial contexts, particularly regarding algorithmic bias and the lack of transparency. 

Since AI relies on historical training data, it may internalize and replicate existing prejudices, which directly threatens the impartiality of legal rulings. Furthermore, the black-box problem poses a major obstacle to institutional accountability. The specific logic an algorithm uses to reach a judgment is often opaque, as the internal decision criteria and learning processes remain inaccessible. This lack of clarity prevents technology from meeting the rigorous transparency standards required in a courtroom, regardless of arguments comparing algorithmic complexity to human reasoning.