What the agent did
Several times a day Claude picks active jobs from the firm's applicant tracking system, reads the company and role success profiles and segment notes, and uses the firm's sourcing tools through Zapier MCP to build a candidate list. A second Claude run checks each list later in the day. A separate process on the firm's own server reads recruiter notes and the Zoom transcript, scores the candidate, researches past employers and writes the submission. A Next-Gen Zap runs at 2 a.m. on weekdays to build prep sheets for screening calls and post a Slack note.
Result, as claimed
Curtis Key says human sourcers' candidates are approved 78 to 83 percent of the time and Claude's about 94 percent. He estimates the sourcing and quality-control automations save the team 240+ hours a week at current volume, and the submission process saves each recruiter 4+ more hours a week. Zapier reports sourcing output up about four times.
This is the result as claimed in the public source linked below. It has not been checked and is not a typical outcome.
Proof
Try it yourself
Score a candidate list against a success profile, then run a separate strict review pass before you look at it.
Prompt
You are my sourcing assistant for the role [job title] at [client]. Success profile: [paste]. Segment notes: [paste]. Candidate list from tools I am allowed to use, contact details removed: [paste]. Step 1: score each candidate 1 to 5 against the profile with one line of evidence from my list. Step 2: drop anyone below 3. Step 3: start a second pass as a strict reviewer who did not build the list and mark each remaining candidate Approve or Reject with a reason. Show a table with Name, Current role, Score, Verdict, Reason and Evidence. Score only on job-related criteria. Do not invent details. Return a draft only. Never send, spend, change settings or write to a live system.
Prompt: starter by Solopreneur Code
Tools
- Claude Source-named tool. Official product URL not independently verified during intake, omitted.
- Zapier MCP Source-named tool. Official product URL not independently verified during intake, omitted.
Steps
- Write a one-page success profile for the role with must-haves and deal-breakers.
- Collect candidates only from sources and tools you are allowed to use, and strip phone numbers and emails.
- Paste the profile and list into the prompt and run it.
- Read the reviewer verdicts and spot-check three approvals against the original profiles.
- Contact candidates yourself and log the outcome so you can compare approval rates over time.
Watch out: AI scoring can carry bias. Keep criteria to job-related skills and experience, never age, race, gender or other personal traits, and let a person make every shortlist decision.
Starter version by Solopreneur Code
Skill: Planning and reviews
Add this skill
Candidate shortlist with second check
Scores a candidate list against your success profile, then a separate strict review pass approves or rejects each one before you see it.
- Works with
- Claude, ChatGPT, Gemini, Any AI chat Any AI chat can work on a pasted list. Pulling jobs from an applicant tracking system needs a connector such as Zapier MCP, as in the story. Not tested.
- Tools in the story
- Claude, Zapier MCP
- Country
- Country unknown
How to use this: Copy the skill, paste it into Claude, ChatGPT or Gemini, and ask it to save it as a skill. If it cannot, it will still follow the steps.
Show skill
If the Copy button does not work, select this text and copy it.