Current state-of-the-art in automotive computational aerodynamics relies on either multi-block structured grids or homogeneous unstructured tetra or hexa meshes. This paper presents a novel approach of unstructured solution-adaptive grids and a generalized tree-based Adaptive Cartesian/Prismatic (ACP) grid concept for automotive aerodynamic applications. The proposed concept resolves several problems which plague tetrahedral grids such as boundary layer grid cell aspect ratio or large grid count. ACP employs unstructured adaptive layer of prisms (or quads in 2D) near solid walls intersected with an adaptive tree-based Cartesian mesh in the rest of the computational space. The prismatic layer resolves the viscous wall layer with high aspect ratio mesh whereas the Cartesian tree mesh provides the smooth grid transition, allows grid coarsening and offers the best support for accurate numerical schemes. Moreover, the ACP approach has a great potential for fully automated meshing of complex automotive geometries. The paper also presents a new CFD solution methodology for Navier-Stokes equations on mixed element type unstructured grids including polyhedra. A robust pressure-based flow solver for all flow speeds and solution-based grid adapter are described. The grid generation and solution methodologies are demonstrated on several flow problems directly related to automotive aerodynamics. Comparison between computational results and available benchmark experimental data is presented. Finally, implications and projections of the proposed methodology for automotive aerodynamic design is discussed.