If it worked for most shit and escalated to a human when it actually needed to, reliably, I'd be fine with it.
I don't believe there's a realistic chance that there's a lot of overlap between the people willing to invest to actually do it properly and the people paying for AI instead of people though.