The problem this solves
A thousand-contact build is a big spend and a long wait, and the worst outcome is discovering at the end that the first hundred were wrong in a way that repeated nine more times. Reviewing in batches as they complete lets a quality problem be caught at row 100 rather than row 1,000, and lets the definition be tightened between batches without losing what has already been built.
The thread is the review surface. Each batch arrives as a compact table under the original request; a reply - 'too many agencies', 'titles are too junior', 'looks good, keep going' - shapes the next batch; and the final list and its CSV close the thread. Nothing is queued blindly, and the record of what was corrected and when sits in the conversation. The Prospecting Agent builds; the thread governs.
Verification runs on every batch, so the count reported is deliverable contacts, and the batches that fall short of 100 verified say why.
How the mission runs
- Confirm the ICP and the batch plan. The Prospecting Agent restates the criteria from <criteria> as a structured definition, proposes the order in which batches will be built - by segment, region or score - and starts on your reply.
- Build and verify a batch of 100. Company search, people search and the email finder produce 100 contacts matching the definition, verification runs on every address, and the batch is completed with deliverable contacts only, with the count of failures noted.
- Post the batch in the thread. The batch is posted as a compact table - company, size, contact, title, verification status - with a one-line summary of what characterizes it, and the agent waits briefly for a reply before starting the next batch.
- Apply corrections between batches. Replies in the thread adjust the definition; the agent restates the change, applies it to the next batch, and offers to re-check earlier batches against the new rule so the whole list ends consistent.
- Save the final list and deliver the CSV. When 1,000 verified contacts are reached, the list is saved as <name> with the final definition and the change history attached, and a CSV is posted in the thread and saved to the workspace.
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
List <name> holding 1,000 verified contacts matching your ICP, built in ten batches each posted in the Slack thread for review, with corrections from the thread applied to later batches and earlier ones re-checked, plus a CSV posted at the end. The thread holds the definition, every batch, every correction and the final counts.
Make it yours
- Change the batch size to 50 for a stricter review or 250 for a trusted definition.
- Push each approved batch to HubSpot or Pipedrive as it completes rather than at the end.
- Run the same loop by email, with each batch arriving as an attachment in the thread.
- Score each batch against the ICP and post the score distribution with the table.
Frequently asked questions
Does the agent wait for approval after every batch?
It waits for a short period you set - long enough to spot-check - and continues if nothing is said, so a build never stalls overnight; a reply at any time still shapes the next batch.
What happens to contacts that fail verification?
They are excluded from the count and listed in the batch summary; the build keeps going until 100 verified contacts fill the batch.
Can I stop early?
Yes. Reply 'stop' and the list is saved with what has been built, the CSV is posted, and no further credits are spent.