The problem this solves
Some of the best qualifying questions are about how a business is run, and none of them are fields in a database. Does this clinic run its own laboratory or send samples out. Does this manufacturer keep a design team in house or brief an agency. Does this retailer employ its own delivery drivers. The answer changes who you should talk to and what you should say, and it is not in any firmographic record.
The signal does exist in public, just not in a structured form. It shows up in the way a company describes its own service, in the roles it advertises, in the equipment it mentions on an about page. A person reading two pages can usually answer with confidence. The problem has always been that reading two pages for four hundred companies is a week of work nobody has.
Because the answer is a judgement, it also needs to come with its reasoning. A list that asserts a company runs something in house without showing where that came from cannot be checked, and a list that cannot be checked tends to get quietly ignored by the people meant to sell from it.
How the mission runs
- Write the operating question as a condition. The prompt states it plainly: the company runs <capability> in house rather than sending it to a third party. Written this way it becomes a test applied to each candidate, which is what allows an operating characteristic to work as a filter at all.
- Scope the search by type and region. The business type and region in the prompt bound where candidates come from. Keeping the scope tight matters more here than on a firmographic search, because every candidate costs a read and a verdict.
- Read the evidence and record the wording. Each candidate is checked against its own site and careers pages, and the sentence that settled the question is kept with the row. A company advertising for a role that only exists when the work is done internally is often stronger evidence than a services page.
- Keep the near-misses visible. Candidates that fail come back with the condition they failed and what was found instead, which frequently turns out to be the useful part. Learning that most companies in a segment outsource the capability is itself a finding about the market.
- Save the list and brief the team. Matches are written to list <name> with the evidence column intact, and a summary goes to #sales in Slack with the CSV attached, so the people who will act on it see the reasoning at the same time as the names.
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 list of 25 qualified companies with domain, firmographics and an evidence column carrying the wording that proved the capability is run internally, plus a Slack summary with the CSV attached and the near-misses recorded against the run.
Make it yours
- Invert the condition to find companies that outsource the capability, which is the better list when you sell the service rather than the tooling.
- Add a size condition so the result only keeps companies large enough to have made the decision deliberately.
- Run the same condition across three regions as separate lists to see where the operating model differs by market.
- Follow up with a contact pass for the function owner once the operating model is confirmed, so outreach reaches the person who owns the decision.
Frequently asked questions
Is a judgement like this reliable enough to sell from?
Each row carries the finding and the page behind it, so the judgement is auditable rather than asserted. Rows where the evidence is thin are visible as such, and a reader can overturn one in seconds.
Why do job ads count as evidence?
A company hiring for a role that only exists when work is done internally has told you about its operating model more reliably than its marketing pages have. Careers pages are written for a different audience and tend to be more literal.
What if almost nothing matches?
That is a real answer about the market, and the near-misses will show whether the condition was too strict or the segment simply works another way. Rewording the condition and running a small sample again is the cheap next step.
Can this run on companies we already have?
Yes. When you already hold the accounts, the per-row research path is the better fit: it answers the same question for every row of an existing list instead of searching for new companies.