Automation of the Planning Phase of a Construction Project Using AI Agents

The chronic digitalisation deficit within the construction sector induces design anomalies and human errors, leading to a severe erosion of investment profitability. This study aims to implement the automation of resource generation and validation processes, acting as a systemic safety barrier to stabilise analytical workflows. The proposed methodology relies on a Multi-Agent System (MAS) architecture embedded within the n8n environment and powered by Gemini-class language models. The framework integrates a deterministic PostgreSQL database within a Retrieval-Augmented Generation (RAG) architecture, enabling the precise, real-time processing of Construction Law regulations. Applying Chain-of-Thought reasoning alongside structured prompt templates helped eliminate model logic drift, ensuring comprehensive result reproducibility. The deployment of this platform induced a 96% acceleration in the pre-construction phase, reducing the formulation time of Work Breakdown Structure (WBS)/Critical Path Method (CPM) structures from a baseline of 480 min to an average of 20 min. The empirical data demonstrates a radical compression of operational costs (OPEX) concurrent with the marginalisation of the Human Error Probability (HEP) index to a residual level of < 1%. Ultimately, the solution drastically minimised the iterative overhead, confining the design cycle to a single execution while maintaining high level of compliance with the 7R (7 Rights) Logistics Directive.