Write · Copy
Written from evidence, not from a merge field
Copy is written the moment a lead enters the queue, personalised from their own profile and the signal that warmed them. There is one writer: the same prompt stack answers a slash command, the scheduled rewrite pass and the queue's own drafting, so the voice cannot fork.
Eleven. Nine came from the rev-ops hiring signal and two reacted to a post about attribution. All eleven are in ICP and none has an outbound row yet.
The Series-A fintech ops campaign has capacity today. Want me to enter them?
What is actually going on
Every claim below is followed by the mechanism under it, because a claim without a mechanism is a slide.
Drafted on entry, not on the way out
A queued row that already carries its words is one quality review can measure and a supervised campaign can read. The launcher's own drafting path stays as the fallback rather than the norm.
The rules you cannot delete sit above the prompt you can edit
A hidden platform layer per role carries no invented claims, no implied prior contact, no personal data in notifications and no arguing with the pre-send gate. It is composed server-side and lives in code, because a key inside a row readable under your own session is unrendered, not hidden.
Quality is a different agent from the one that writes
The review agent measures reply rate per step and names the failing one. The copy agent rewrites it. A reviewer that fixes what it finds leaves nobody able to say what was wrong.
Every prompt is yours to edit, with a reset
Each agent ships with the defaults for its role and a one-click return to them. Editing the prompt changes what the cron does too, because they read the same row.
Step by step
How ai copywriting runs
- 1
Gather
The contact's cached profile, the signal that fired, and your company profile.
- 2
Write
Through the copy role's own prompt stack. The hidden layer, your edited prompt, then the brief.
- 3
Check
The quality agent compares the draft against its evidence, hourly.
- 4
Rewrite
A finding sends the failing step back to copy on the six-hourly pass. It does nothing at all when there is no finding.
The numbers
Facts, not adjectives
Every figure here is a real value from the running product. A cadence from the scheduler, a price from the credit table, a limit from the ramp. A marketing page that invents a limit is a support ticket with a delay on it.
- Review cadence
- Quality hourly, copy every 6 hours
- Findings
- Throttled to one a day each
- Slash commands
- /copy, /leads, /outbound, /intent, /qa
- Hidden layer
- In code, per role, never in a database row
- Reset
- Per prompt, to the role default
AI copywriting, honestly
Can I stop it inventing things?
That rule is not optional and not in the editable prompt. No invented claims and no implied prior contact are in the hidden layer, which is appended to rather than merged into.
Is the slash command a different writer from the cron?
No. Both read the workspace's edited prompt off the agent row, which is the whole reason the row rather than the preset is the source.
Write
What to say, in a voice that is yours
AI copywriting is one part of it. The rest runs too.
Sourcing, enrichment, intent, copy, sending, replies and reporting are one product with one bill and one gate. Connect a channel and the whole motion starts.
Start with sales. $2,000 or $5,000 onboarding, then at least $250 monthly credit funding.