What should you not hand off to Chloe yet?

Just because Chloe can technically do something doesn't mean it belongs to her yet. Some of this still genuinely needs a person on the other end of the line.

Enthusiasm for a new AI teammate tends to run well ahead of good judgment, especially in the excitement of the first week after setup. Here's a practical list of what to hold back, at least until you've built enough trust in the process through real, observed calls to justify handing more over to her.

None of these are permanent restrictions carved in stone. They're the honest starting boundaries for a team still learning what Chloe does well in their specific context, and they're worth revisiting deliberately as your confidence in her calls grows over the following weeks and months.

Complex objection handling that doesn't fit a script

A lead who pushes back with a real, specific concern — pricing, a bad past experience with a similar product, a detailed competitor comparison — needs a response tailored precisely to what they actually said, not a generic rebuttal pulled from a script. That's still fundamentally a human skill. Don't route this to Chloe expecting a save; you'll likely lose the deal and learn the wrong lesson about what she can do.

High-stakes or emotionally sensitive conversations of any kind

Anything involving a cancellation request, a genuine complaint, or a customer who's clearly upset needs a person who can read tone in real time and adjust their approach on the fly. Chloe isn't built for that kind of emotional nuance, and shouldn't be stretched to cover it just because the setup technically allows it.

Leads you haven't segmented properly yet

If you don't know which list a lead belongs to, or what actually makes them different from your average buyer, don't point Chloe at them and hope it works out. She'll run the exact same script on people who genuinely needed different questions entirely, and the results will look confusing without an obvious explanation why.

Your highest-value or most strategic accounts on the board

A deal that matters disproportionately to your quarter, or to a key relationship you're building, deserves a human's full attention from the very first call, not a standardized script built for volume. Save Chloe for volume work across the rest of the pipeline. Save your best rep specifically for the accounts where the outcome really, genuinely matters to the business.

Anything you haven't listened to yourself yet, personally

If you haven't personally reviewed real call transcripts from a given segment, don't scale Chloe into it just because the setup screen said it was technically ready to go. Trust gets earned call by call, one transcript at a time, not assumed automatically because a configuration screen didn't throw any errors.

What's the underlying pattern across all of these?

Every one of these boundaries comes down to the same underlying question: does this specific situation require judgment that wasn't written down anywhere in the script? If the honest answer is yes, it belongs to a human for now, at least until you've built a script sophisticated enough to actually cover it reliably.