2026-10-192026-10-192026-07-10PEREIRA, Maria Eduarda Alves. Avaliação do impacto da inteligência artificial na acurácia de levantamentos de quantitativos: um estudo comparativo voltado à tomada de decisão em engenharia de custos. Orientador: Marcelo Rassy Teixeira. 2026. 56 f. Trabalho de Curso (Bacharelado em Engenharia Civil) – Faculdade de Engenharia Civil, Instituto de Tecnologia, Universidade Federal do Pará, Belém, 2026. Disponível em: https://bdm.ufpa.br/handle/prefix/10388. Acesso em:.https://bdm.ufpa.br/handle/prefix/10388The integration of Artificial Intelligence (AI) tools into the construction industry has led to significant changes in engineering cost processes, requiring the technical validation of new technologies against traditional methods. The present study aims to evaluate the accuracy of Large Multimodal Language Models (GPT, Claude, and Gemini) in the extraction of quantity takeoffs for foundations, aiming to support decision-making in executive budgeting. A comparative case study was conducted on a large residential building, confronting the concrete volumes, formwork areas, and excavation volumes of 14 structural blocks extracted by the conventional method (spreadsheet) and the automated method (Single-Pass Inference). The methodology was based on prompt engineering and the analysis of the absolute Percentage Relative Error, adopting an acceptable limit margin of 5%. The results demonstrated a polarized technological landscape: while the GPT and Gemini architectures achieved high technical precision, with accumulated errors below 1.2% in all input categories, the Claude model presented mathematical and geometric instability, with deviations reaching 16.15%. The study mapped severe limitations in the interpretation of complex polygonal geometries and identified stochastic inconsistencies generated by textual noise in element nomenclature. It is concluded that, for the analyzed sample of foundations, AI demonstrated potential to optimize the workflow, suggesting a transition of the engineer's role towards auditing activities, validating itself as a decision-support tool rather than an autonomous mechanism for replacing human reasoning.Acesso AbertoOrçamentosEngenharia de custosInteligência artificialChatGPTBudgetingCost EngineeringArtificial IntelligenceCNPQ::ENGENHARIAS::ENGENHARIA CIVILTrabalho de Curso - Graduação - MonografiaAttribution-NonCommercial-NoDerivs 3.0 Brazil