AI · Outreach Agent

Prospector

Personalised outreach that learns which messages actually get replies — without spamming anyone.

NodeAI rewriteReply trackingGmail draftsCLI

The idea

Most teams sit on a prospect list they only ever work a fraction of. Prospector takes the whole list, works out who each contact is and where they are in the sequence, and matches them to the right message — then tracks which templates earn replies so the winners rise and the duds get cut. It does the useful half of an AI SDR, without the part that gets you blocked: nothing goes out without a human.

How it works

01
Segment
Classifies each contact by who they are — technical, revenue, ops — from their role and company.
02
Stage
Works out where they are in the sequence: cold, follow-up, engaged or exhausted — derived from the send history, never re-touching someone who replied.
03
Match & personalise
Picks the template built for that segment and stage, fills it with the contact's details, and optionally rewrites it to sound human and specific.
04
Send — gated
Dry-run by default; drafts to Gmail for review; live send only with an explicit per-run flag. Human in the loop, always.
05
Learn
Logs which template went to whom, then computes reply-rate per template and per segment — so you keep what works and kill the slop.

Not spam-ware

Auto-blasting cold email at strangers is a fast way to get your domain burned. Prospector inverts the default: the human-review path is the easy one, and the fully-autonomous path is deliberately awkward to trigger. Messages with an unfilled placeholder or no address on file are never sent.

Built from a reel, made responsible
Inspired by the "rebuild the funded AI SDR yourself" genre — kept to the parts worth keeping: segmentation, templating, and honest reply-rate data.