Why this exists.
Pulse shows both ads to 120 simulated customers who match your buyer and tells you which one wins — and why.
A patchwork that breaks.
- A shared spreadsheet nobody opens on time
- Prompts copy-pasted into a chat window
- A contractor who disappears for two weeks
- Output that lands in a different shape every run
A workflow that ships.
- One brief, one cadence, one place to read it
- Every claim cited, every step reviewable
- A finished artifact in your team’s format
- You stop splitting ad budget between a winner and a dud.
Four moves.
Pulse runs each move with a preview attached — so you know what lands before you ever hit send. Skip freely once you know which parts carry the weight.
- 01Kickoff
Upload the two versions.
Pulse starts with the brief and asks only for what's missing. No boilerplate intake form, no setup meeting.
Chat · marieteYUpload the two versions.PPulse is working - 02Gather
Describe who you are selling to.
Sources are pulled, cleaned, and cross-checked against prior runs — every claim carries a citation you can trace.
checklist · step-02Source connected · describe who youContext loadedFirst pass completeSource connected - 03Reason
Run the test. Four minutes.
The agent thinks out loud where it matters — trade-offs named, assumptions surfaced, judgments explained.
ranked results01Pick · ad creative0.9402The · ad creative0.8203Ad · ad creative0.67 - 04Draft
Put the whole budget behind the winner.
A first draft lands in the format your team already uses. You edit the last 10%, not the first 90%.
delivery · inboxPnewPulse → your teamjust now · scheduled weeklyPick the winning adYou stop splitting ad budget between a winner and a dud.Open briefing
Configure Pulse.
Pulse runs on structured setup, not freeform prompts. Fill the fields once and the run is reproducible every time — same agents, same sources, same output shape.
Inputs in, outputs out.
Pulse runs on the inputs on the left and hands back the artifacts on the right. Skip any input — the agent will ask for it the first time it needs it.
- One source of truth (CSV, CRM, or warehouse)
- A one-paragraph brief on the goal
- The KPI you want to move
- A scored, cited brief you can forward
- A structured file for downstream automation
- An alert when anything material changes
A finished artifact, not a todo list.
Every run ends the same way — a packaged brief in the channel your team already reads. Here's a preview of what shows up.
Here's the brief for this week. I ran the playbook end-to-end, flagged anything that shifted against last run, and packaged the output for Slack and the shared drive.
- Upload the two versions.
- Describe who you are selling to.
- Run the test. Four minutes.
- Put the whole budget behind the winner.
Where teams stall.
Three ways we see this go sideways — and how to avoid each one.
Pointing the agent at stale or half-connected data. Clean the source once, compound every run after.
Running it once and forgetting. Put it on a weekly cadence so the numbers actually move.
Skipping the first review. Check the first run by hand — trust compounds from there.
Before you start.
Usually one source is enough to see value. Pulse can run on a CSV paste for the first pass; connect the CRM, the data warehouse, or the tool of record once you want it to run on its own.
Most teams put this on a weekly cadence. That's the sweet spot between "too noisy to read" and "too stale to act on". Adjust once you see how the numbers behave.
Whoever owns the downstream action. Pulse hands back a finished result — the value is in somebody actually reading it and shipping the decision the same day.
It usually isn't. The first pass is calibration — tell Pulse what was off, rerun, and the second is close. By the fourth it reads like a teammate.