Overview
AI Hiring Agent — Automated Candidate Sourcing & Screening
Owner: Alhasan Farajallah · Type: Playbook (training-grade)
1. Objective
Given a role and a pool of applications, automatically source, summarise, score, and compare candidates against the job's real requirements — replacing hours of manual CV triage with structured, consistent, explainable shortlists.
2. Prerequisites
- An LLM (Claude) with document-reading capability.
- The job description with clearly stated must-have and nice-to-have criteria.
- The candidate documents (CVs / applications).
- A hiring manager who reviews and owns the final shortlist (human-in-the-loop).
3. Step-by-step workflow
- Define the rubric. Convert the job description into a weighted scoring rubric (must-haves gate; nice-to-haves add points). Confirm the rubric with the hiring manager.
- Ingest candidates. Load each CV; the agent produces a structured summary (experience, skills, education, notable gaps).
- Score against the rubric. Each candidate is scored per criterion with a short justification citing the CV — no opaque numbers.
- Rank & compare. Produce a ranked table and a side-by-side comparison of the top N.
- Flag risks. Surface unverified claims, gaps, or inconsistencies for human review.
- Hiring-manager review. The manager reviews justifications, adjusts, and approves the shortlist. The agent never auto-rejects a person.
- Generate interview aids. For shortlisted candidates, draft tailored interview questions targeting gaps and must-haves.
4. Key prompts
- "Convert this job description into a weighted scoring rubric separating must-haves from nice-to-haves."
- "Summarise this CV into experience, skills, education, and gaps; then score it against the rubric with a one-line justification citing the CV for each criterion."
- "Rank these candidates and give a side-by-side comparison of the top 5, flagging any unverified or inconsistent claims."
5. Validation / Q&A checkpoints (for training)
- Q: What governs scoring? A: A weighted rubric derived from the job description, confirmed by the hiring manager.
- Q: Can the agent reject a candidate? A: No — it scores and flags; the human owns shortlist and rejection decisions.
- Q: How is each score explainable? A: Every criterion score carries a one-line justification citing the CV.
- Q: What does the agent produce for interviews? A: Tailored questions targeting each candidate's gaps and the role's must-haves.
6. Measured impact
CV triage time cut from hours to minutes per role; consistent, explainable scoring; reduced bias from applying one rubric uniformly with human oversight.