If AI agents could be trusted to generate a better signal/noise ratio by delegation than by working-alongside the AI (where the bottleneck is the human)
They can’t typically (currently) do better on their own than working alongside a human, but a) a human can delegate a lot more tasks than they can collaborate on (and can delegate more cheaply to an AI than to another human), and b) though they’re not as good on their own they’re sometimes good enough.
Consider call centers as a central case here. Companies are finding it a profitable tradeoff to replace human call-center workers with AI even if the AI makes more mistakes, as long as it doesn’t make too many mistakes.
They can’t typically (currently) do better on their own than working alongside a human, but a) a human can delegate a lot more tasks than they can collaborate on (and can delegate more cheaply to an AI than to another human), and b) though they’re not as good on their own they’re sometimes good enough.
Consider call centers as a central case here. Companies are finding it a profitable tradeoff to replace human call-center workers with AI even if the AI makes more mistakes, as long as it doesn’t make too many mistakes.