We will study the formation history of the Milky Way, and the earliest phases of its chemical enrichment, with a sample of more than 1.5 million stars at high galactic latitude. Elemental abundances of up to 20 elements with a precision of better than 0.2 dex will be derived for these stars. The sample will include members of kinematically coherent substructures, which we will associate with their possible birthplaces by means of their abundance signatures and kinematics, allowing us to test models of galaxy formation. Our target catalogue is also expected to contain 30,000 stars at a metallicity of less than one hundredth that of the Sun. This sample will therefore be almost a factor of 100 larger than currently existing samples of metal-poor stars for which precise elemental abundances are available (determined from high-resolution spectroscopy), enabling us to study the early chemical evolution of the Milky Way in unprecedented detail.