The Automated Candidate Screening & Shortlisting system is a fully automated AI recruitment pipeline that extracts data from resumes, scores candidates 0–100 against the job description, categorizes them (Shortlist / Review / Reject), creates/updates them in Workable with tags and custom fields, and instantly notifies the hiring manager with a one-click summary. Built with Zapier + ChatGPT API + Workable, it reduces manual screening time by 70%, processes each resume in <30 seconds, handles 200+ resumes/day, and was delivered in 9 weeks (15 Sep – 19 Nov 2025).
An event-driven no-code/low-code pipeline:
Intake: Resume arrives via Gmail, Google Drive, or job board webhooks triggering a Zapier workflow.
AI Processing: ChatGPT gpt-4-turbo parses and scores resumes using job-specific prompts, returning structured JSON.
Routing: Zapier Paths use score thresholds to move candidates to the correct stage in Workable.
Outcome: Automatic ATS updates (tags/fields/stages) and instant Slack/Email notification cards to managers. Includes full retry logic and GDPR compliance.
Structured Extraction: Captures contact info, experience years, skills match %, and certifications instantly.
Weighted Scoring: Per-job configurable rubric with confidence scores and red-flag detection.
Auto-Actions: Workable candidate creation, tag application (AI-Shortlist/Review/Reject), and automatic stage moves.
Recruiter Interface: Manager notification cards with scores, summaries, and one-click "View in Workable" links.
Accuracy: 92% alignment with human recruiter decisions after final prompt tuning.
PDF/DOCX documents are converted via Zapier text extraction and sent to ChatGPT with a specific job rubric. The structured JSON response is mapped to Workable fields via threshold branching for ATS updates and notifications. The system is rate-limit safe with delays/queues, and parsing failures are routed to a "Review Needed" folder with Slack alerts. Decisions are logged with prompt versions for audit trails.
Timeline: September 15 – November 19, 2025
Testing: Unit tests per Zap step, end-to-end integration, and UAT with HR (achieving 92% accuracy). Load testing conducted at 250 resumes/day.
Deployment: Duplicate Zaps created for testing before flipping live triggers; 10-day hypercare period post-launch. Rollback strategy allows pausing Zaps in under 5 minutes to revert to manual screening.
Monitoring is managed via Zapier task history and a dedicated Slack channel for error alerts. A Google Sheets dashboard tracks volume, avg scores, rejection rates, and processing times. Monthly prompt and rubric reviews ensure the AI remains aligned with evolving hiring standards. Current production success rate: 99.8%.
Methodology: Agile with daily stand-ups and Trello tracking.
.png)
.png)