Gas turbines are today used in numerous industrial and aeronautical applications. To increase the specific power output and efficiency, a high turbine inlet gas temperature is desired. The high temperature leads to the need of cooling critical components in the hot gas path. Siemens Industrial Turbomachinery AB, SIT AB, in Finspång manufactures gas turbines where the internal cooling of critical components is done through serpentine channels. To utilize the cooling air as efficiently as possible, vortex creating objects are placed inside the channel which result in higher heat transfer. To compute the heat transfer in the channel, correlation based approaches that will give a uniform value for an entire channel are often used. This thesis contains two parts. First, investigating how an automated CAE process can be developed that is able to be incorporated into the SIT AB CAE process of today and with a future vision of a, basically, "one-click-CFD" approach for non-generic geometries. Secondly, how CFD simulations for predicting heat transfer levels inside the cooling channels with high accuracy and that captures local features of heat transfer can be performed. The suggested CAE approach involves the CAD-tool NX for geometry creation and for managing an entire CFD project the ANSYS software Workbench, combined with ANSYS Meshing for generation of computational grid, CFX-pre and CFX for pre-processing and solving and CFD-post for post-processing. This approach is suggested for generic geometries due to the simplicity in incorporating it into existing CAE processes. For the future vision of non-generic geometries, the inhouse developed project manager Concept is suggested. It allows for customized coupling between a broader range of available software tools. To validate the CFX model and to investigate how the CFD calculations should be performed, two cases were set up, one where the CFD model and the inhouse code was compared to experimental data of a generic geometry and one where the CFD model and the inhouse code were compared at engine-like conditions. The results for the experimental case resulted in heat transfer coefficients from the CFD model that were 30% off from experimental data, and the inhouse code maximum deviation was 10%. Compared to previous numerical studies this was considered to be of acceptable accuracy, and the location of data extraction points were considered to cause the deviation in the CFD model. For the engine-like case both CFD and inhouse code predicted the heat transfer level as expected. The simulations were performed in steady state mode on automatically generated meshes with the SST-Reattachment turbulence model. The Reynolds number varied from 10 000 to 80 000 and the meshes were around 4-10M elements in size.