Every outreach tool advertises the same list of exit conditions: stop the sequence when they reply, when a meeting is booked, when the deal moves stage. Ours can store those criteria too. It acts on one of them.
The failure mode nobody demos
Reply detection is a heuristic pretending to be a rule. An out-of-office is a reply. A bounce notification is a reply. A one-word "thanks" forwarded by an assistant is a reply. Each of those can silently remove someone from a sequence they were meant to stay in, and you will not notice, because the failure looks identical to working correctly.
The inverse is worse. A real reply that the detector misses means you keep emailing someone who already answered you — the single fastest way to make a warm contact cold.
An automation that is wrong quietly is more expensive than one that does less.
What we enforce instead
Unsubscribe. It is a public link that needs no login, it is unambiguous — the recipient took an explicit action — and it blocks every future send permanently. There is no confidence score involved and nothing to tune.
- The recipient resolves through their linked person record, so an address exists in one place.
- Every send is written to the log with its step and outcome, so you can see what went out.
- Unsubscribes appear in that log too, alongside the sends they prevented.
The honest version of the feature list
We could have shipped reply detection, put it on the pricing page, and let the edge cases land on customers. Plenty of tools do. Naming the limit costs us a checkbox in a comparison table and saves our users the class of bug where the software was confidently wrong on their behalf.
Dana Iyer
Product at RESK
Works on outreach, sequences, and the sending pipeline at RESK.