The problem this solves
Founders and new sales leaders often start from a blank page: there is a product, a website that explains it, a few early customers, and no written definition of who to sell to next. Defining the ICP by workshop takes weeks and produces a document that nobody tests against real companies. Skipping it produces a list built on someone's instinct, which is tested on real companies at the cost of the first quarter.
A better order is to derive the ICP from the evidence, show the reasoning, and test it immediately with a small list. The website says what the product does and who it is for; the market says which segments exist and how large they are; scoring says which segment fits best and why. Fifty companies from the top segment, each with a verified decision-maker, is enough to learn whether the definition holds.
How the mission runs
- Read the website for the offer and the buyer. The Prospecting Agent crawls <mysite.com> and reads the product pages, pricing, case studies and customer logos to work out what is sold, at what price point, to which function, and which customers already exist. Existing customers are enriched to see what they have in common.
- Define candidate segments. The market is broken into segments by industry, headcount band and technology pattern, drawing on firmographic and technographic data for how many companies exist in each. Each segment gets a short description and a size estimate.
- Score the segments and show the reasoning. Each segment is scored for fit using the evidence from the site and the existing customers - the functions named, the price point, the technologies the customers share - with the reasoning written out. The ranked segments are shown to you before any list is built, so you can adjust before spending on data.
- Build the first 50 and push to HubSpot. From the top segment, company search and people search produce 50 companies with one verified decision-maker each, saved as list <name> and staged into HubSpot as companies and contacts with the segment and score recorded, reviewed before the write.
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 written ICP with six to eight scored segments and the reasoning behind each score, plus list <name> holding 50 companies from the top segment with a verified decision-maker each, created in HubSpot as a reviewed batch with the segment and fit score on every record. The report is short enough to share with a co-founder and specific enough to argue with.
Make it yours
- Provide your ten best customers as well and the scoring weights toward what they share.
- Ask for 20 companies from each of the top three segments to test them side by side.
- Deliver to Pipedrive, Attio or a Google Sheet, or skip the CRM write until the segment proves out.
Frequently asked questions
How does the agent know who my buyer is from the website?
From the functions the copy addresses, the roles named in case studies and the problems the pricing page solves. It states its inference so you can correct it before the segments are scored.
What if my website is thin?
The agent asks for what is missing - a short description of the product and a few customers - and proceeds from that, noting which conclusions rest on your answers rather than the site.
Can I re-run the segment scoring later?
Yes. The ICP is saved with the list, and as customers accumulate, the scoring can be repeated against the newer evidence.