dry mode
Sign in
No name picked on this device yet: decisions will be logged as rep. Say who you are so your approvals and uploads carry your name.
Overview
HomePipelineReportsFree offers
Products
AtlasAtlas MediaContentPiGenPiVerseAMC
Admin
People & roles
Library
Who we targetHow it worksThe rulesSuppressionLinkedIn content
The rule
The engine drafts and organizes. It never sends. A person approves every list and every message.

How it works

I write the first draft. You decide what goes out.

Every week I read our own data, find the brands worth a conversation, and write each one a full cadence built on their real numbers. Then I stop and hand it to you. Nothing reaches a prospect unless you approve it, and I never send anything myself.

The whole pipeline, end to end

Data teams feed it, the engine writes, and you decide what goes out. Your part is steps four and five; everything before is done by the time you open a run.

  1. 1Data teams
    Upload your data

    Each team drops its CSV on the Prepare page. That is the whole ask.

  2. 2Weekly, by Lina
    The engine runs

    It finds the brands, writes each a full cadence from real numbers, and checks every line.

  3. 3This week
    Cadences appear

    One per brand, ready to read. Nothing has been sent.

  4. 4Reps
    You review each touchyou are here

    Approve, edit, or reject every message. Decisions save as you go.

  5. 5Reps
    Send from HubSpot

    Download the approved messages as a CSV and send on your own schedule.

Your part, in detail

Steps four and five above are the job. Here is exactly what each looks like.

  1. Open a run and read the cadence

    This is the job

    Open a run and read its cadence top to bottom. Each message is built on that brand's real numbers.

  2. Approve, edit, or reject each touch

    Your call, every message

    Approve what is good, Edit what needs your voice, Reject what does not fit. Your decisions save the moment you click, so you can stop and come back.

  3. Download the HubSpot CSV

    One click, per run

    When you have been through a run, download its CSV. It contains only what you approved: rejected and unreviewed messages are left out.

  4. You send, from HubSpot

    Always a person

    Import the CSV and send on your own schedule. This is the line the engine does not cross: it drafts and queues, you send.

Why the messages say what they say

Every message leads on one real number about that brand. If you are wondering where a figure came from, this is the answer, and it is worth knowing before you get asked on a call.

Observable

Atlas, Content, PiGen

We can see these numbers ourselves: a marketplace download, a listing audit we ran, an AI visibility scan. So the message cites the prospect’s own real figure.

“Sony is cited in 8 of 10 AI answers, but its GEO score sits at 60/100, and Samsung takes the recommendation ahead of it.”

Benchmark

PiVerse, AMC

These live in the client’s own systems, so we cannot know them cold. The message opens on a segment benchmark and offers a free pilot that reveals their real number.

“Advertisers running search and DSP together typically see about 38% overlap. A free AMC pilot finds your exact number.”

Worth remembering: a benchmark product can never claim a private figure it does not have. If you ever see one do that, reject it and tell Lina. That is a bug, not a style choice.

The rules I enforce on myself

These are checked on every message before you ever see it. You should not have to police them, but you should know they exist, because they are why some things are missing.

  • A client is never prospected
    L'Oreal, Kenvue, Clorox and Swiss Arabian, plus every brand they own, are removed with a logged reason. A wider list of existing clients is held back too and routed to the account owner as expansion instead.
  • Nobody is touched twice
    If another play contacted someone in the last 21 days, they are held back.
  • Turkey never gets an email
    Turkish cadences are calls, LinkedIn and WhatsApp only.
  • No data, no draft
    If a brand's numbers are missing or stale, I raise a task for the data team instead of writing something generic.

What to look for when you review

I check the hard rules. You are checking the things only a person can judge.

Does it sound like you?

You are the one sending it. If a line is not how you would say it, Edit it. Your edit is what exports.

Is the number still true?

The data is from this week, but you may know something I do not. If a stat is stale or wrong, reject it.

Is this the right person?

I match on role and category. If the contact is clearly wrong for the pitch, reject and it drops out of the CSV.

Quick answers

Can I break something by clicking around?

No. Nothing here sends. The worst case is approving something you did not mean to, and you can hit Reset on any touch to put it back.

Do I have to finish a run in one sitting?

No. Every decision saves as you make it. Close the tab and come back; your work is where you left it.

What happens if I edit a message?

Your version replaces mine in the CSV. The original is kept underneath, so Reset always restores my draft if you change your mind.

Why does a contact have no email?

If we could not verify an address the contact is flagged for LinkedIn, but the cadence may still contain email touches. If you see an email touch for someone with no verified address, reject it and tell Lina.

The numbers changed since last week. Why?

Each week is a fresh read of the data. If a brand's position moved, the message moves with it.

How do I get my team's data in?

Upload it on the Prepare data page: pick your product, choose the market, drop the CSV. It is checked the moment it arrives, and it joins the next weekly run.

That is the whole job

Read, decide, export, send. Everything before that is mine; everything after it is yours.

Go to this week’s runs →