Or train people to do research while in grad school. There are plenty of fields where people don't do post-docs before applying for faculty positions, e.g., all of the fields where there's no money for post-docs, and "hot" fields where people have better job prospects. I know two CS professors at a major research university, who went straight to faculty positions from their PhDs.

The current structure is a holding pattern for absorbing the large population of PhDs who can't get jobs.