Toward AI-Assisted Civil Quantity Takeoff and Earthwork Estimation: A Human-in-the-Loop Framework for Excavation-Focused Preconstruction Workflows

Abstract

Quantity takeoff remains a foundational but labor-intensive step in civil preconstruction, particularly for excavation, earthwork, and sitework scopes derived from 2D plan sheets and surface-based reports. Current digital workflows already support measurement capture and exportable quantity data through platforms such as Bluebeam Revu, Trimble Business Center, Autodesk Takeoff, and PlanSwift, yet estimators still spend significant effort interpreting legends, reconciling labels, applying assumptions, and converting measurements into estimate-ready line items. This paper proposes a human-in-the-loop framework for AI-assisted civil quantity takeoff focused on excavation and earthwork estimation. Instead of replacing estimator judgment, the framework positions artificial intelligence as a layer for document understanding, quantity normalization, error flagging, and cost-item mapping on top of existing takeoff workflows. The paper defines the technical architecture, data strategy, implementation path, and validation roadmap for an applied civil engineering initiative and implementation direction.