The objective of this study was to outline a fully automated, X-ray-based, mass-customization pipeline for knee replacement surgery, thoroughly evaluate its robustness across a range of demographics, and quantify necessary input requirements. The pipeline developed uses various machine learning-based methods to enable the automated workflow. Convolutional neural networks initially extract information from inputted bi-planar X-rays, point depth and statistical shape models are used to reconstruct three-dimensional models of the subjects' anatomy, and finally computer-aided design scripts are employed to generate customized implant designs. The pipeline was tested on a range of subjects using three different fit metrics to evaluate performance. A digitally reconstructed radiograph method was adopted to enable a sensitivity analysis of input X-ray alignment and calibration. Subject sex, height, age, and knee side were concluded not to significantly impact performance. The pipeline was found to be sensitive to subject ethnicity, but this was likely due to limited diversity in the training data. Arthritis severity was also found to impact performance, suggesting further work is required to confirm suitability for use with more severe cases. X-ray alignment and dimensional calibration were highlighted as paramount to achieve accurate results. Consequentially, an alignment accuracy of ±5–10 deg and dimensional calibration accuracy of ±2–5%, are stipulated. In summary, the study demonstrated the pipeline's robustness and suitability for a broad range of subjects. The tool could afford substantial advantages over off-the-shelf and other customization solutions, but practical implications such as regulatory requirements need to be further considered.