One LLM agent per trade over construction site photos (trpevski.com)

A new AI system for construction inspections, dubbed Moonraker Inspector, utilizes specialized agents to analyze photos of construction sites, producing tailored reports for each trade—structural, plumbing, electrical, and more. Each inspection generates a verified defect list, a formatted report, and specific findings dispatched to responsible contractors. The significance of this approach lies in its structured accountability: each finding is grounded with a reference to the project’s specifications, ensuring that disputes can be quickly resolved with verifiable evidence. This feature transforms the inspection process into a legal document by assigning a named supervisor to each defect finding, enhancing the reliability and defensibility of the inspections. The underlying technology relies on a rigorous retrieval system that filters project-specific documents, such as building regulations and standards, to ensure that all findings are accurate and current. Notably, this process mitigates the risks associated with outdated specifications by incorporating strict revision control. By employing narrow-scoped AI agents, the system enhances defect detection accuracy, reducing the likelihood of "quiet failures" where issues are overlooked. The efficiency of the AI agents also supports linear scaling of operational costs, as they can be deployed to handle multiple trades simultaneously. This innovative framework not only streamlines the inspection process but also offers reassurance to contractors that their findings are substantiated by clear, actionable references.