A chat demands your attention.
When you send a chat message, it asserts to the receiver: "this is important enough for you to look at it right now."
The receiver doesn't have any process to automatically balance across all of those things vying for their attention.
Agents enter into that same communication channel, saying "Hey, look at me!"
That means you top out at managing like five agents or so.
A background task could proactively do things without keeping you in the loop.
But for that to work, it would have to be structurally aligned with your intentions.
If you could solve this problem then instead of five agents, you could have dozens or even hundreds, autonomously acting on your behalf.
This would unleash the full power of AI for us to accomplish meaningful things.
None of this is possible without having a system that is structurally aligned with your intentions.