Making an antibody-enzyme conjugate isn't trivial. By using a primary/secondary set-up you can use the same well-characterised conjugate in combination with many different primary antibodies (as long as these primaries are all raised in the same species). There is also the possibility of some amplification: for example, if the secondary is an anti-Fab then two secondary Igs will bind to each primary.
**Response to OP comment**
Most primary antibodies in common use are derived from rabbit or mouse, and most are IgG. So, for example, the secondary antibodies goat anti-rabbit IgG and goat anti-mouse IgG will cover most experiments.