S Skillsheet Docs

Agentic Sourcing

Agentic sourcing is Skillsheet’s most powerful feature. Instead of manually searching and filtering candidates, you describe the role and let an AI agent do the heavy lifting. The agent autonomously searches across candidate data, evaluates fit, and delivers a ranked list of matches.

What Is Agentic Sourcing?

Traditional sourcing requires you to write queries, adjust filters, review results, and repeat. Agentic sourcing flips this workflow. You provide a description of what you need, and an AI agent:

  1. Interprets your requirements (skills, experience, industry, location, and more).
  2. Searches across Skillsheet’s entire candidate database.
  3. Evaluates each candidate against your criteria using multiple signals.
  4. Ranks candidates by match quality.
  5. Delivers a curated shortlist for your review.

The agent works in the background, so you can continue with other tasks while it sources candidates for you.

How It Works

Agentic sourcing uses a multi-step AI pipeline:

  • Requirement analysis — The agent breaks down your description into structured criteria (required skills, preferred experience, location constraints, and so on).
  • Broad search — The agent casts a wide net across the candidate database, using semantic understanding to find relevant profiles.
  • Deep evaluation — Each candidate is scored against your criteria. The agent considers skills, experience, career trajectory, and other relevance signals.
  • Ranking and filtering — Candidates are ranked by overall match score. The agent filters out candidates who do not meet hard requirements.
  • Delivery — The final shortlist is presented to you with match scores and explanations for each ranking.

From a Job Position

  1. Navigate to Positions in the sidebar.
  2. Open the position you want to source for.
  3. Click the Start Agentic Search button.
  4. The agent uses the position’s title, description, requirements, and location to begin sourcing.
  5. Optionally, add extra instructions in the Additional Notes field (e.g., “Prioritize candidates with startup experience” or “Exclude candidates currently at competitor companies”).
  6. Click Launch.

From a Custom Query

  1. Go to the Search page.
  2. Click the Agentic Search tab (next to the standard search bar).
  3. Describe what you are looking for in plain language. Be as detailed as possible.
  4. Click Launch Agentic Search.

The more detail you provide, the better the agent performs. Include information about required and preferred skills, years of experience, industry background, team size context, and any dealbreakers.

Monitoring Search Progress

Once launched, the agentic search runs in the background. You can monitor its progress:

  • A status indicator appears on the Positions page or in the Search tab showing the current phase (Analyzing, Searching, Evaluating, Ranking).
  • Real-time updates show how many candidates have been evaluated so far.
  • You will receive a notification when the search is complete.
  • Typical searches take between 2 and 10 minutes depending on the complexity of your requirements and the size of the candidate pool.

You do not need to keep the page open. The agent continues working even if you navigate away.

Reviewing AI-Found Candidates

When the search finishes, open the results to see your ranked candidate list:

  • Each candidate has a match score (0-100) indicating overall fit.
  • An AI summary explains why the candidate was ranked at that position, highlighting strengths and potential gaps.
  • Candidates are ordered from highest to lowest match score by default.
  • Click on any candidate to view their full profile, including their skillsheet, interview results, and resume.

Understanding Match Scores

  • 90-100 — Excellent match. The candidate meets or exceeds all stated requirements.
  • 75-89 — Strong match. The candidate meets most requirements with minor gaps.
  • 60-74 — Moderate match. The candidate has relevant experience but is missing some key criteria.
  • Below 60 — Weak match. The candidate may have transferable skills but does not align closely with the role.

Accepting or Rejecting AI Suggestions

For each candidate in the results, you can:

  • Accept — Adds the candidate to your pipeline for the associated position. They are placed in the New stage by default.
  • Reject — Removes the candidate from the results. The agent uses your rejections to improve future searches.
  • Save for Later — Bookmarks the candidate without adding them to the pipeline.

You can also accept or reject candidates in bulk by selecting multiple candidates and using the Bulk Actions menu.

Credits and Cost

Agentic searches consume more credits than manual searches because of the additional AI processing involved:

ActionCredit Cost
Manual search1 credit
Agentic search5 credits

Credits are deducted when the search is launched, not when it completes. Your current credit balance is always visible in the top-right corner of the dashboard.

Use Agentic Sourcing When

  • You have a well-defined role and want a ranked shortlist without manual effort.
  • You are sourcing for a hard-to-fill position and want the AI to explore broadly.
  • You are running multiple searches in parallel and need to save time.
  • You want the AI to consider nuanced criteria that are hard to express with filters alone.

Use Manual Search When

  • You want quick, exploratory results for a loosely defined need.
  • You already have a specific candidate or company in mind.
  • You want to conserve credits for high-priority roles.
  • You prefer hands-on control over the search process.

Both approaches are complementary. Many recruiters use manual search for initial exploration and agentic sourcing for deep, comprehensive sourcing on priority roles.