The Outbound Prompt Set | Sovereign Pivot

How to use this

Fill the brackets once. Keep the file.

Each prompt has bracketed variables. Fill them in for your company and your first trigger, then save the filled-in versions somewhere your whole team can reach — a shared doc is enough. That document is your harness: the place where prompts get refined as you learn what works, instead of six people running six different versions and nobody able to say which one sent the bad email.

Run these in whichever assistant you already have. They assume no CRM, no orchestration platform, and no budget approval. The first version of this system is a spreadsheet with four columns: trigger, account, date sent, reply.

Start with one trigger and one segment. Add a second of anything only when the first is running.

Define the signal

Signal Hunter

Start from the trigger, not from the data you can get. Reversing that order is why most signal projects produce noise.

Our buying trigger is: [describe the event that means a company might buy — a funding round, a new head of department, a published tender, a stack change].

Find every public venue where this trigger becomes visible. For each venue, tell me: how the trigger appears there, whether access is free or paid, and how quickly it shows up after the event.

Then propose a monitoring cadence matched to the trigger's half-life, and flag which venues I can check manually this week without any tooling.

Score and threshold

Lead Scorer

Trigger strength times ICP fit. The threshold matters more than the formula — decide what gets dropped before a human ever sees it.

Here is our ICP: [industry, company size, buyer role, the problem we solve].
Here are accounts that surfaced on the trigger [trigger name]: [paste list].

Score each on two axes, 1–5: trigger strength (how recent and how strongly it predicts buying) and ICP fit. Show both scores and the product.

Recommend a cut-off below which we should not reach out at all, and explain what we lose by setting it there.

Translate the proof

Cold market entry

Home-market credibility does not cross the border by itself. A US buyer checks for their own regulator, not the equivalent, and unrecognized client names carry no signal.

Our credentials in our home market are: [clearances, certifications, marquee clients, deployment numbers, revenue].

We are selling into [target market] to [buyer role].

For each credential, tell me: does it carry weight with this buyer as stated, does it need restating in a local equivalent, or does it carry no signal at all? For the ones that need restating, write the sentence I should use instead.

Then list what proof this buyer would expect that we do not currently have.

Write the opener

Personalization Writer

The test is whether a human would recognise it as researched rather than as a merge field with a company name in it.

Account: [company]. Trigger that surfaced them: [specific event, with date].
What we do: [one sentence, plain language].
The claim I want to make: [the single specific thing we can prove].

Write a four-sentence opener that references the trigger specifically enough that it could not be sent to any other company. No pleasantries, no "I hope this finds you well," no summary of their own business back to them.

Then tell me which sentence is doing the most work, and what would be lost if I cut the other three.

Ask for a design partner

Feedback motion

Not a softer sales sequence. The ask is asymmetric on purpose: twenty minutes of their problem, and your roadmap moves.

We are looking for design partners, not buyers. We want [number] practitioners who have the problem [problem] and will tell us candidly where our product is wrong.

Person: [name, role, company]. Signal that they have this problem: [what they published, hired for, or built].

Write the ask. Make explicit what they get and what it costs them in time. Do not imply a purchase, a demo, or a pilot. If the message reads like a sales sequence wearing a different label, rewrite it.

Classify the reply

Reply Classifier

This is the stage teams skip, and skipping it is what leaves the loop open.

Reply received: [paste]
Trigger ID that caused the send: [trigger name or ID]
Claim used in the message: [the claim]

Classify as one of: interested / not now with a date / wrong person / no.
Record the classification against the trigger ID and the claim.
If "wrong person," name the role we should have written to instead.
Then draft the next touch.

Tag the loss

Closing the loop

Stall reasons and churn reasons are the same reasons, eighteen months apart. Clustering them is how outbound starts paying downstream.

Here are our closed-lost and no-reply notes from the last [period]: [paste].

Cluster the reasons. Separate them into: missing capability, wrong timing, wrong account, and message failure.

Count each cluster against the trigger that surfaced the account. Tell me which trigger produces the most wrong-account losses, and whether the problem is the trigger itself or our scoring threshold.

Read your own answer

Day one

Buyers ask an assistant before they ask a vendor. A company the model cannot resolve is not ranked low — it is absent from the consideration set.

What does [company name] do? Who are the leading vendors for [your category]?

Run both questions in three different assistants. Then bring the answers back here and tell me: what is factually wrong, what is missing, and whether the description is specific enough that a buyer could tell us apart from a competitor.

Then write the one-sentence description we should use everywhere, identically.