Reliable estimation of manufacturing process times is a key challenge in sheet metal production, particularly for small and medium-sized enterprises that rely on heterogeneous documentation originating from multiple subcontractors. Engineering drawings often contain dense dimensioning, annotations, and layout variations that complicate robust geometry extraction, while native CAD models provide accurate geometric representations but may lack complete manufacturing and process-related metadata. This work presents an agent-based, multi-source AI framework that integrates information from engineering drawings and CAD models to support data-driven manufacturing time estimation, with a focus on laser cutting processes. Textual metadata, such as material specifications and thickness, are extracted from drawings using OCR-based document analysis combined with layout-aware document understanding methods. In parallel, geometric features relevant to laser cutting are deterministically derived from CAD models by generating flat patterns and computing manufacturing-oriented descriptors, including total cut length, hole-related features, and geometry complexity indicators. This feature extraction strategy follows established principles of feature-based manufacturing analysis for cost and time estimation. The extraction pipeline is organized into specialized agents responsible for geometry extraction, metadata extraction, and validation, enabling confidence-aware cross-checking between data sources and mitigating failure modes commonly associated with single-source extraction pipelines. Extracted features are combined with historical routing data to train and evaluate data-driven regression models for laser cutting time estimation, with the objective of assessing the feasibility of predicting process time from automatically extracted manufacturing features. Beyond laser cutting, the proposed framework is designed to be extensible to additional sheet metal processes such as bending and welding, providing a scalable foundation for AI-assisted industrial decision-making in production planning, time estimation, and cost assessment under uncertainty.