Skip to main content
Build a short list you would actually use before scaling to hundreds of prospects. This walkthrough asks a colony to research companies, explain fit, and stop for your review.

​
What you need

Prepare your company website, the customer segment you want to reach, and any exclusions. An ideal customer profile (ICP) describes which companies are a good fit; it is more useful when it includes buying signals and disqualifiers. Use public sources for a first pass. If you choose a workflow that uses LinkedIn or a paid data provider, connect the browser or the required credentials first.
1

Write the brief

Select New Chat, or open a relevant template from View all on the home screen. Replace template variables before deploying.Describe the audience, the evidence you want, and the maximum size of the first batch. The example below is a starting point.
2

Review the target definition

Ask the colony to summarize its inclusion and exclusion criteria before sourcing. Correct misunderstandings in the same chat so later research uses the revised brief.
3

Inspect the first batch

Ask for a saved table. Open Data and check a sample against the source links. Look for duplicate companies, weak evidence, and prospects that match a keyword but not your customer profile.
4

Refine before expanding

Explain why specific rows are good or bad matches. Ask the colony to revise the scoring and replace rejected entries before expanding the list.

​
Review the evidence

A high score should come with a reason you can check. “Works in our industry” is weaker than a specific need supported by a company page or announcement. Keep “not found” distinct from “does not exist.” For example:

​
Use the results

Download a table as CSV from Data when you need to review it elsewhere. To add records to the shared CRM, give the colony a separate instruction specifying which approved rows to add and how to handle duplicates. For outreach, start with drafts:
Review behavior depends on the prompt, skills, and tools used. A template that says “message each prospect” may authorize sending; OpenHive does not impose the same approval step on every workflow. Once a run works, reuse its prompt or add a schedule with an equally explicit scope.