HR Recruitment Screening Pipeline

1. The Problem

A typical open role generates 80–250 applications. Reading 200 CVs, shortlisting 20 candidates, scheduling first-round interviews, and conducting initial screening calls consumes 15–20 hours of HR or hiring manager time — before a single meaningful conversation has taken place. Multiply across multiple simultaneous open roles and recruitment becomes a full-time job in itself. More critically, manual screening is biased toward familiarity and pattern-matching, not potential — meaning the best candidates are regularly screened out by the least effective part of the process.

2. Integration & Webhook Setup

Follow the exact steps below to configure and deploy this automation inside your OpenClaw workspace.

1. 1Create a new agent named `recruitment-pipeline`. Connect your Applicant Tracking System (Greenhouse, Lever, Workable, or SmartRecruiters all supported) and your async video platform (HireVue or Spark Hire via API).
2. 2Build your `SCORING_CRITERIA` for each role type in your organisation: define the must-have qualifications, preferred qualifications, and behavioural signals that predict success in your specific environment. Weight each criterion — must-haves carry 3× the weight of preferred criteria.
3. 3Configure your `SCREENING_QUESTIONS` bank: build 10–15 role-specific async video questions per job family. The agent selects the 4 most relevant questions for each candidate based on their CV profile — personalised screening at scale.
4. 4Set your `BIAS_MITIGATION` parameters: configure the agent to assess based on structured criteria only, with name and photo fields excluded from the initial scoring pass. The scored shortlist is presented with candidate identifiers only — personal details visible only after the hiring manager confirms they want to proceed with each candidate.
5. 5Configure your rejection communication template: the automated rejection should be warm, specific, and prompt. Candidates who receive a personalised rejection within 48 hours report significantly better employer brand perception than those who wait weeks for a generic response.

3. The Context Payload (context.json)

Save this file as: .openclaw/agents/hr-recruitment-screening-pipeline/context.json

{
  "automation_id": "24",
  "title": "HR Recruitment Screening Pipeline",
  "level": 3,
  "tier": "Enterprise Playbook",
  "setup_time": "1 week",
  "estimated_api_cost": "~$40–$90/mo",
  "client_price_range": "$2,000–$10,000",
  "agents": [
    {
      "role": "orchestrator",
      "model": "claude-3-5-sonnet",
      "temperature": 0.2,
      "max_tokens": 4096
    }
  ],
  "memory": "session",
  "output_format": "structured_json",
  "human_review_gate": true,
  "documentation_standard": "required"
}

4. Execution Commands

Run these commands from your openclaw-workshop/ directory to validate, test, and schedule this automation. Commands are taken directly from The OpenClaw Income Engine, Appendix G.

# ── STEP 1: Register application webhook ──
$ openclaw webhook create recruitment-pipeline --trigger application.created
$ 
# ── STEP 2: Load scoring criteria for all role families ──
$ openclaw config load-criteria recruitment-pipeline --file ./scoring_criteria.json
$ 
# ── STEP 3: Load screening question bank ──
$ openclaw config load-questions recruitment-pipeline --file ./screening_questions.json
$ 
# ── STEP 4: Run calibration on 10 past applications (test mode) ──
$ openclaw run recruitment-pipeline --calibrate --input ./past_applications/ --limit 10
# Review scoring accuracy vs actual hiring outcomes before going live.
$ 
# ── STEP 5: Activate webhook and go live ──
$ openclaw webhook activate recruitment-pipeline
$ 
# ── Score a single application on demand ──
$ openclaw run recruitment-pipeline --application-id <ats_application_id>
$ 
# ── View shortlist for an active role ──
$ openclaw run recruitment-pipeline --role-id <ats_job_id> --shortlist
$ 
# ── Release personal details for a shortlisted candidate ──
$ openclaw run recruitment-pipeline --application-id <id> --reveal-identity
$ 
# ── Generate bias audit report for last 30 days ──
$ openclaw run recruitment-pipeline --bias-audit --days 30