Agentes verticais de inteligência artificial: suporte para gestão universitária

Abstract

Brazilian public administration faces a continuous tension between the traditional bureaucratic model and the pursuit of efficiency-focused managerial governance. Public universities are situated within this context, and the University of the State of Mato Grosso (UNEMAT) recognizes the need for technological innovation in its Institutional Development Plan (PDI). However, a significant gap persists between this strategic vision and operational practice, constituting the central problem of this research. This problem materializes in the manual management of a massive volume of administrative acts, totaling 7,155 ordinances published by the Rectory and Pro-Rectories between January 2023 and mid-2025. Given this scenario, this research sought to answer how the validation of an AI Vertical Agent (AVIA) prototype by UNEMAT managers demonstrates its value as a support tool for optimizing operational efficiency and qualifying strategic decision-making. To this end, the study proposed the development and validation of the aforementioned prototype. The methodology is characterized as applied research, utilizing a quantitative-descriptive approach centered on the prototype's validation by a census (N=7) of the strategic stakeholders directly responsible for the process. Data collection employed a structured, anonymous questionnaire, administered individually following a live demonstration of the tool, with analysis based on descriptive statistics. The results were conclusive, revealing a unanimous consensus (100% positive evaluation) regarding operational efficiency gains, the high usability of the interface, and the tool's capacity to fully meet the identified needs. There was a strong consensus on the impact on quality, with 85.7% of managers agreeing that the tool reduces errors and 100% attesting to its capability to ensure standardization. This perception translated into a full endorsement for implementation, with 100% of managers recommending its adoption and 57.1% advocating for its immediate prioritization. It is concluded that the research answers its central question by demonstrating that the prototype's validation confirms its value on three distinct levels: operational, by optimizing efficiency and mitigating errors; tactical, by strengthening governance and compliance; and strategic, by enabling data-driven management. The dissertation resulted in a functional technological product, the GestorDoc IA software, which has been duly filed for a Computer Program Registration with the INPI.