31 - USO DE IA COMO FERRAMENTA DE SUPORTE À DECISÃO EM SUPRIMENTOS EM EMPRESA DE FUNDAÇÕES

Doi: 10.47094/COBRAMSEG2026.2/31

PALAVRAS CHAVE: Inteligência artificial, suprimentos, almoxarifado, obras de fundações, estoque, nível de serviço.

KEYWORDS: Artificial intelligence, supplies, warehouse, foundation works, inventory, service level.

ABSTRACT:
This paper presents an Artificial Intelligence (AI) agent used as a decision-support tool for supply management in a foundation contractor, aimed at keeping construction workfronts supplied and reducing stoppages from missing materials. The agent runs on commercial large language models, in a hybrid setup with Anthropic Claude and Azure OpenAI: for each item it reads the movement history and the parameters registered in the ERP and returns a diagnosis (average consumption, days of coverage, risk of stockout or excess) and a recommended action with its justification, reviewed by the supply team before acting. The replenishment logic follows classic inventory theory. The agent was validated on real warehouse data from Geofix Engenharia: 3,879 outbound transactions in January 2026 and twelve-month item histories (April 2025 to March 2026). The analysis found three concrete problems affecting the work: a maximum stock level nearly three times higher than needed (MIG wire 1.2, about R$ 9,870 tied up); a minimum parameter above real consumption (electrode 557, about R$ 2,018 freed); and an invisible material-loss cycle from the lack of return traceability (copper grounding cable, slurry sector). The approach does not require complex ERP integration and can be adopted gradually.

Autores

• Giuliano Silva Castro de Oliveira

• Marcio Abreu de Freitas

• Willians Nascimento de Almeida