The Growth Operator
Research, qualify, draft, stage.
A person decides.
- Pilot
- shadow mode, not yet proven end to end
- 0
- messages it sends on its own, by design
One account through the line
Six workers prepare the account. Nothing passes the gate until a person approves the exact message.
- 01Account researcherVerifies the signal and finds the buyer. A source for every fact.
- 02QualifierScores 0–100, applies the gates, decides.
- 03Person researcherReads what the buyer posts and writes.
- 04Message writerDrafts three touches from verified fields only.
- 05QA evaluatorChecks facts, shape, voice and repetition across the batch.
- 06Reply managerReads the whole thread and drafts one answer.
The line
Designed to do
Research · scoring · drafting · checking · staging
Waits for a person
Sending · inviting · enrolling · publishing · changing a live campaign
Go on, try.
The record what it is, the problem, what changed
- What it is
- An AI agent I designed for Tcules' account-based outbound. The demand system's working half.
- The problem
- Good account-based outbound is research-heavy and easy to let rot: lists go stale, evidence gets lost, a message goes out on a guess.
- What I designed
- One pipeline and one record per account, six worker roles, a check by a separate reviewer, and an approval rule tied to the exact message.
- The rule
- Research, scoring, drafting and internal files can run. Sending, inviting, publishing or changing a live campaign waits for a person, right before the exact action.
- Status
- A shadow-mode pilot. It drafts; it does not send.
- What is not here
- Tcules' accounts, targets, senders and message copy. This page describes the design, not the data.