Due to an incomplete treatment of turbulent convection, 1D stellar evolution codes do not agree with asteroseismic measurements at high frequencies. In order to mend this problem, we replace the outermost layers of 1D models with mean 3D simulations of stellar atmospheres. I present the current stage of our ongoing endeavor to establish a robust and consistent method for such a replacement, discussing several ambiguities in the procedure. Furthermore, I present our first promising results.