Skip to main content
A good search rarely happens in one go. You ask, look at what comes back, and tell the agent what was off. This guide covers what to say, what the agent can and can’t filter on, and how to keep your results tidy. For the basics, see Search for candidates.

​
Before you search: fix the criteria

The agent builds its searches from your brief, and every result is scored against your criteria. Ten minutes on the criteria saves several rounds of searching. See Write criteria that score well. If there’s context the agent should know but that isn’t a hiring criterion, put it in the brief’s Notes. The agent reads the notes and uses them, including when it searches. For example: “The team is remote-first, so strong async writing matters” or “We’ve had poor results with candidates from large consultancies.” In the job’s chat, say where to look and anything that narrows the field:
Search for candidates in Norway and Sweden.
The agent turns the brief into a set of different search angles, runs up to four rounds, and fills gaps between rounds. It ranks and scores what it finds, then posts a summary in the chat: who scored highest, who stands out on a specific criterion, and what the pool is missing.

​
What you can ask it to filter on

These work reliably when you say them in the chat:
  • Countries and cities.
  • A minimum or maximum number of years of experience.
  • A word in the candidate’s current job title, such as “lead” or “staff”.
  • Companies to include or exclude, up to 20 of each. For example, “exclude our current employer and our two closest competitors”. This applies to that search only.
Two things don’t filter reliably, so put them in words instead:
  • Function or department. Say “backend engineers” rather than asking for the engineering department.
  • Seniority works only partly as a filter. Say “senior” or “lead” in your request, and make seniority a criterion if it matters to the score.

​
Refine with what was wrong

After each search, tell the agent what didn’t fit and why. Be specific:
Too many consultants. Focus on people at product companies, and skip anyone with less than five years of experience.
Each request runs as a new search and appears as a new group on the Scouting tab, titled with what the agent searched for. Earlier groups stay, so you can compare.

​
Keep the Scouting tab tidy

  • Delete a search that went the wrong way: Search actions → Delete search.
  • Export a search or your shortlist for a spreadsheet: Download as CSV.
  • If you’ve changed the criteria since a search ran, click Re-score in that search’s Search actions menu so the scores match the new brief.

​
Know the limits

  • Staffer searches its own candidate database, which covers Europe and North America. It doesn’t search LinkedIn live, and you can’t add someone who isn’t in the database.
  • The database has no email addresses or phone numbers. You get those per person with Request contact info, or you add them yourself. See Get contact details.
  • Maximum sourcing results in Settings → Workspace → Preferences caps how many people one search returns (25, 50, 75 or 100). Higher caps take longer.
  • You can’t stop a search once it starts. If it runs out of time, you get the results it had scored so far, without a warning.
  • Every search uses credits. Broad searches that run all four rounds cost more than narrow ones.

​
From results to pipeline

Shortlist the promising people with the bookmark, then promote them from the Shortlist group. Promoting doesn’t contact anyone: it sets up their application, and the Process tab then suggests your first message. See Shortlist and promote candidates.