It is important to understand how the fascinating diversity of modern enzymes has manifested itself since the onset of life. A powerful approach to address this question is the computational and experimental retracing of evolutionary processes by protein engineering techniques such as rational design, directed evolution, and ancestral sequence reconstruction. In this review, we show that these methods were used to elucidate how first enzymes have emerged from ligand-binding protein scaffolds, how the enormous variety of enzymes with similar catalytic mechanisms but different substrate specificities have emerged, and how natural selection has driven the formation of oligomeric enzyme complexes from their monomeric ancestors.