The problem this solves
Every outbound program rests on an answer to one question - who exactly should we sell to - and at most companies that answer is folklore. The ICP was written in a workshop two years ago, describes the customers the founders happened to know, and has never been tested against what the product page actually promises today. Lists get built to match the folklore, outreach underperforms, and the diagnosis lands anywhere except the targeting.
Deriving the ICP properly is an analysis job with real inputs: what your site says you do, which segments your competitors court, which kinds of companies exhibit the problem your product removes, and how those segments differ in reachability and deal shape. Done honestly it produces something more useful than a persona document - a ranked list of segments with a fit score and the reasoning attached, so the first list you build starts from the best odds rather than from habit.
This mission runs that whole chain from one input: your URL. It reads your site the way a stranger would, studies the market around it, scores the segments, and then does the work most analyses stop short of - building the verified contact list from the top segment and drafting the first touch for every name on it.
How the mission runs
- Read your site like a buyer. The crawler reads your pages - product, pricing, cases, positioning - and extracts what you actually sell, to whom, at what altitude. The ICP work starts from the site you have, which is also how every prospect will judge the claim your outreach makes.
- Derive and score the segments. The market around your offer is broken into candidate segments, each scored for fit with the reasoning shown: problem intensity, budget shape, reachability, and how well your current proof matches what that segment needs to hear. Scores come with evidence, never as bare numbers.
- Build the list from the winner. The top segment becomes a real list: companies matching its firmographic shape, then the people who would own the purchase, with titles and seniority matched to how your product is bought.
- Verify before writing. Every contact's email is verified for deliverability before any drafting happens, so the writing effort lands only on messages that can arrive and your sender reputation is never spent on guesses.
- Draft the first touches. Each contact gets a drafted first touch grounded in the segment reasoning: why this company plausibly has the problem, in language that mirrors your site's own claim. Drafts land for your review; nothing sends itself.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A pipeline starter kit from one URL: your ICP as ranked, fit-scored segments with the reasoning attached, a verified 30-contact list from the strongest segment, and a reviewed-ready first-touch draft per contact. The targeting logic is written down, so the next list starts from evidence instead of folklore.
Make it yours
- Run it quarterly: the scores shift as your site, proof, and market move, and the reranking is the strategy conversation worth having.
- Point it at a product launch page instead of the homepage to derive the ICP for one offer rather than the whole company.
- Feed the second-ranked segment to a parallel run, so the two lists become an honest A/B on targeting rather than on copy.
Frequently asked questions
What makes the fit scores trustworthy?
Each score carries its reasoning: the evidence from your site, the market signals for the segment, and the named factors that raised or lowered it. You can disagree with a factor and rerun - the score is an argument you can inspect, never an oracle.
How is this different from just buying a list?
A bought list answers "who matches these filters." This mission first derives which filters deserve to win, shows its work, and then builds the list - so the expensive part of outbound, aiming it, is done before a single contact is pulled.
Does anything send automatically?
No. The mission ends at verified contacts and reviewed-ready drafts. Sending stays with your team and your sequencing tools, with approval gates available for teams that want them enforced.