I will also point out that automation is currently starting to eat fields that were traditionally un-automatable due to either skill, labor price (or both), including everything from complex farm work done for cheap to surgical robots.
What advantage do we will then have? The only possible advantage I can see is that raising a human for 20 years + housing + food + other expenses and then employing him would be cheaper than replicating another AGI, but I cannot imagine this to be true.
I'd expect supercapable AGI to command higher wages than less-capable humans, so they wouldn't be pushing out service jobs.
This is more of a bridge solution to full automation. "We need doctors to control dumb surgical robots until we're advanced enough that surgical robots can operate without a human controlling them" etc
We're doing the "assist" thing already in all the places that we can't quite do the "machine only" thing.
Even self-driving cars mirror this: the goal is a vehicle that doesn't require an operator. The bridged reality, currently, is that the operator must remain alert and able to take over any time.