Translational research often involves the analysis of large numbers of plasma samples which are inherently complex and require significant sample preparation for metabolomics analysis. Manual sample preparation for metabolomics analysis involves quenching (protein precipitation) and lipid removal (liquid-liquid extraction) which is both time-consuming and inherently error-prone. Biological variation in plasma samples impacts the levels and composition of both proteins and lipids which can further complicate sample preparation and analysis. We have developed a semi-automated sample preparation workflow for metabolites in plasma to facilitate translational studies. As part of the development, different approaches to protein precipitation were tested to assess both quenching effectiveness and sample stability. For lipid removal, classic liquid-liquid extraction is known to be ineffective in completely removing some classes of lipid which can cause ionization suppression in subsequent LC/MS analysis. In this study, a sorbent-based approach was found to be more effective in removing the majority of the lipids thus improving the analysis results. The sorbent combines size exclusion and hydrophobic interactions between the sorbent and the long aliphatic chain of lipids to efficiently remove lipids without unwanted loss of metabolites. After optimization of quenching and lipid removal, the protocol was automated on the Bravo workstation and reproducibility was assessed using a targeted metabolomics approach on a triple quadrupole LC/MS system.