A adoção da inteligência artificial nas auditorias de obras públicas e de infraestrutura do TCU: uma análise da complementariedade com as abordagens e técnicas tradicionais de auditoria

Abstract

This study analyzes the effectiveness of artificial intelligence (AI) tools, specifically Alice and the Budget Analysis System (SAO), compared to traditional audit methods applied by the Federal Court of Auditors (TCU) in public infrastructure works between 2015 and 2024. The research adopted a qualitative approach, based on documentary analysis of Annual Activity Reports, judgments and institutional publications, triangulated with academic literature. The main objective was to evaluate the capacity of these tools to identify irregularities, such as overpricing, restrictive clauses and errors in budgets, and to analyze how their speed and scalability complement traditional manual techniques, such as physical inspections and document analysis. The results show that AI offers greater agility and breadth in data screening, with emphasis on Alice's performance in the analysis of public notices and SAO in the verification of road works budgets, demonstrating its potential to optimize audit processes. However, traditional approaches demonstrate greater accuracy in on-site audits and in the verification of physical execution, being indispensable in contexts with non-standardized data and complementing the performance of AI in aspects that require more in-depth or on-site analysis. The conclusion points to the complementarity between the two models, recommending the adoption of hybrid audits, investments in technical training and data standardization as strategies for improving external control.