What's the one support task you'd hand to AI tomorrow — and the one you'd never let it touch?

Building DeskCrew, the thing I keep coming back to is the trust line. Most "AI support" either over-promises (auto-sends wrong answers) or under-delivers (a glorified FAQ).

I landed on AI-drafts-the-reply, human-approves-before-send — but I'm genuinely curious where other people draw the line.

If you run support, or you're the founder answering DMs for your own product: what's the one repetitive task you'd happily hand to AI today? And what's the one you'd never trust it with?

Trying to build the human-in-the-loop model around how people actually think about this — so real takes welcome.