Run Semi-Autonomous Sourcing with Gem's AI Agent

1. In Gem's integration settings, connect your ATS (Greenhouse, Workday, SAP SuccessFactors, or Lever, depending on what your company uses)
2. Confirm requisition, candidate, and pipeline-stage data are syncing correctly
3. Flag any sync issues to your Gem account team early. Integration problems are one of the more commonly reported friction points

What you should see: Your open requisitions and existing candidate data appearing inside Gem. Troubleshooting: If data isn't syncing, check whether your ATS requires the connection to be approved on both sides, similar to other recruiting integrations.

Start narrow rather than turning the agent loose across every open req at once.

1. Pick one requisition, ideally one where you already understand what a strong candidate looks like
2. Define the role's must-have skills, experience range, and any deal-breakers inside Gem's sourcing setup
3. Set boundaries on what the agent can do autonomously versus what needs your review. Most teams keep outreach send decisions under human review, at least at first, given the brand and candidate-experience risk of fully automated messaging

What you should see: A configured sourcing campaign tied to your chosen requisition, with clear parameters and review checkpoints.

1. Create a sequence template the agent will personalize per candidate rather than send verbatim
2. Write it in your company's actual voice, referencing the real role and team, not generic filler
3. Set the number of follow-up touches and the spacing between them

What you should see: A multi-step sequence ready to be personalized and applied to sourced candidates.

1. Activate the AI sourcing agent for your chosen requisition
2. Let it search and rank candidates against your parameters
3. Review the first batch closely rather than approving in bulk. This is where you calibrate whether its ranking logic matches your actual judgment

What you should see: A ranked candidate list with the agent's reasoning or match signals visible for each profile.

1. Open the drafted outreach for your reviewed candidates
2. Check for accuracy. AI-drafted messages can reference outdated roles or invent details not actually present in a profile, the same failure mode seen across other AI sourcing tools
3. Approve, edit, or reject each message before it sends, especially in your first few campaigns

What you should see: A queue of personalized messages awaiting your approval, not messages that have already gone out.

Gem can also resurface past applicants who fit a new requisition instead of starting every search from zero.

1. Enable AI Rediscovery for your organization's historical candidate database, if your ATS integration supports it
2. When you open a new req, check what it surfaces from past applicants before sourcing externally
3. Treat these matches as a first pass, not a guarantee. Past applicants may have taken other roles or changed what they're looking for

What you should see: A list of previously sourced or applied candidates who match your new requisition's criteria.

1. Track which agent-sourced candidates actually respond and advance versus which ones stall
2. Adjust your parameters based on what's actually working, not just what the agent theoretically can do
3. Expand to additional requisitions once you trust the calibration on your first one

What you should see: Response and advancement rates that improve as you tune the agent's parameters to your actual hiring bar.