<p>This paper presents a hybrid human–robot workflow for the fabrication of mass-customized sheet-metal panels for discretized double-curved architectural envelopes. The proposed system combines computational design rationalization, vision-based robotic identification and pick-and-place, augmented-reality-assisted manual bending, and a panel-level digital twin for process feedback and traceability. Rather than replacing the operator with a fully automated bending line, the workflow distributes fabrication intelligence across computational models, robotic handling, and human-guided forming. In this way, it supports high geometric variation while reducing infrastructure requirements and preserving human agency in production. A pilot implementation on a discretized façade prototype demonstrates the feasibility of the approach through quantitative assessment of panel recognition, bending accuracy, and fabrication time. The results indicate that the system can support low-volume, high-variation production with lower capital costs than conventional automated bending cells, making it relevant for small and medium-sized enterprises. By rethinking how design, robotic execution, and manual skill are coordinated, this work contributes to current research on augmented human–robot collaboration in architectural fabrication.</p>