For the industrial repair of carbon fiber reinforced plastics (CFRP), scarfing is used to re-establish the load path. In the industry today, the state of the art removal of CFRP layers for a repair process is done mainly manually, leading to a time-consuming process. Therefore, an automation of this process is desirable. Today, vacuum suction blasting is used to pre-treat CFRP surfaces only for the superficial removal of impurities or the removal of the cover layer before bonding. With the common used blasting agents and nozzle geometries, the even removal of larger areas together with the fiber was not possible. This work shows that the technology of vacuum suction blasting was adapted to be used as an automated scarfing method. A combination of blasting parameters, especially the nozzle geometries together with a blasting agent were found to be able to remove CFRP layers precisely, detecting and correcting errors in-line with a line scanner-measuring unit. The presented method allows treating large-scale surfaces, scarfing the area one single layer at a time, increasing the removal rate in comparison to common blasting. With vacuum suction blasting the grinding dust emissions and process forces are low, post cleaning or further surface activation are not necessary and the removal results can directly be controlled. Challenges still exist with inaccurate removal due to interruptions in the blasting program and the generation of sharp edges for a stepped scarf.