a human can screw up in a non scalable manner an agent can do it at scale and worse it would attract bad actors who will certainly exploit that
This is what everybody gets wrong: Agents aren't bad or evil. Nor crazy. They don't steal. They pretty much work (after some setup and periods of supervision) as expected. And yes, sometimes they might be wrong and spend more. Same as the human. From a business standpoint it is just risk/reward structure.
loading story #50480670