The problem this solves
Agency lists are easy to build and hard to trust. Every agency claims every vertical on its services page, because the services page is written to lose no enquiry. The claim that actually predicts a good partnership is narrower: has this agency published a case study in the vertical, with a named client and a result. That is a fact about a specific page on a specific site, and no firmographic filter can express it.
Teams that need this today read agency sites by hand. It is genuinely useful work and it does not scale past an afternoon, so the list ends up being the twenty agencies somebody had time to open. The agencies that would have qualified on page three of the search results never get looked at.
There is a second cost to doing it by hand, which is that the proof gets lost. Somebody reads a case study, decides the agency qualifies, and writes the name into a sheet. The page that justified the decision is not recorded, so the first outreach message has to go back and find it again before it can reference anything specific.
How the mission runs
- Turn published proof into a testable condition. The condition is written as a sentence: the agency publishes at least one client case study in <vertical> on its own site. The Prospecting Agent treats that as something to verify per candidate, so a services-page claim does not pass and a real case study does.
- Generate and size the candidate agencies. Candidates are found across the open web and checked against the headcount band in the prompt, ten to two hundred employees. Size matters for this motion because it sets who answers a partnership approach, so it is applied as a condition rather than filtered afterwards.
- Verify the case study and keep the page. Each candidate is read and given a verdict with the finding behind it, including the page where the proof sits. That address is the artefact worth keeping: it is what makes the first message specific, and it is what lets a colleague sanity-check the list without repeating the research.
- Add the decision maker and verify the email. For every agency that qualified, the mission finds the founder or managing director, then runs the address through verification so what lands in the CRM is deliverable. Contact details come from the finders that verify them, which is deliberate: a researched address would be a pattern guess.
- Write companies and people into Attio. Agencies land as companies and the contacts as people linked to them, with the case-study link written into a note on the company record. The note is what makes the CRM entry self-explanatory three weeks later when someone picks the account up.
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
Twenty-five agencies in Attio as companies with linked people, each carrying a verified email, the headcount band, and a note holding the case-study address and the sentence that qualified it. The near-misses stay on the run with the condition they failed, so a vertical that returns thin results can be re-run with the wording adjusted.
Make it yours
- Change the proof condition to a published integration, a partner badge or a conference talk when a case study is not the signal that matters in your market.
- Add a technology condition so the list only keeps agencies whose own site shows the stack you build for.
- Run it per vertical and keep each result as its own list, so a partner team can work one market at a time with the evidence attached.
- Swap the destination for HubSpot, Pipedrive or a Google Sheet when the partner motion lives outside the main CRM.
Frequently asked questions
How is this different from filtering an agency directory?
A directory records what an agency says about itself. This records what an agency has published, which is a higher bar and a much better predictor. The two lists overlap less than you would expect.
Does the mission read the case study or just find the link?
It reads the page and keeps the wording that satisfied the condition, so the row carries a finding rather than a bare link. That is what makes the first outreach line writable straight from the list.
Why are the emails found separately rather than researched?
Researching an address from the open web tends to produce a plausible pattern rather than a confirmed mailbox. The contact step uses the finders that verify deliverability, so the CRM receives an address that has been checked.
What if an agency qualifies but the contact is missing?
The agency still lands in Attio with its evidence, and the contact field stays empty rather than being filled with a guess. An honest gap is easy to work, and an invented address costs a domain reputation.