Every incoming campaign reply gets classified automatically:

Incoming reply
    |
    |-- Out-of-office pattern check (no AI needed) --> OOO
    |
    |-- Hard-no pattern check ("unsubscribe",
    |   "stop emailing") --------------------------> NOT_INTERESTED
    |
    +-- AI classification --------------------------> HOT / WARM / COLD
                                                      / NOT_INTERESTED / SPAM

Label	Meaning	Example
HOT	Direct intent to meet or buy	”Can we schedule a call this week?”
WARM	Interested, wants more info	”Send me some case studies”
COLD	Not now, but door open	”Not right now, maybe later”
NOT_INTERESTED	Explicit no	”Not interested” / “Remove me”
OOO	Out-of-office auto-reply	”I’m on vacation until…”
SPAM	Irrelevant / promotional	-

Every classification comes with a confidence score:

0.95-1.00: crystal clear
0.80-0.94: clear with minor doubt
0.60-0.79: ambiguous, judgment call
below 0.60: very unclear - worth a human look

Corty learns from your corrections. When you reclassify a reply, that correction is stored in your organization’s classification log. Future classifications check the log first - so Corty gets more accurate for your market every week. Subtle language distinctions (like Dutch “op dit moment geen interesse” = COLD vs. “geen interesse” = NOT_INTERESTED) are handled explicitly.

How reply classification works (HOT, WARM, COLD...)