What this automation does
When you receive applications for a role, this automation reads each resume, compares it against your job description and requirements, and generates a structured scorecard. Each candidate gets scored on: years of relevant experience, required skills match, education fit, industry background, and red flags (employment gaps, job hopping, overqualification).
The output is a ranked spreadsheet with all candidates, their scores, and a one-paragraph summary explaining each score. Hiring managers go from reviewing 200 resumes to reviewing a ranked shortlist of the top 20. The AI handles the mechanical screening so humans can focus on the nuanced evaluation of top candidates.
Tools you need
- PDF processing tool: Extracts text from resume PDFs — PyPDF, Adobe Extract, or similar
- OpenAI API (GPT-4): Analyzes resumes against job requirements and generates scorecards ($0.03-0.06 per resume)
- Google Sheets + Apps Script: Stores results, displays rankings, and provides the review interface
How to set it up
Step 1: Create a Google Sheet with the job description, required skills (weighted by importance), minimum experience, and any other criteria. This is your scoring rubric that the AI will use.
Step 2: Build a processing script that reads resume PDFs from a Google Drive folder (or an ATS export). Extract the text from each resume using a PDF parser. For image-based resumes, use OCR first.
Step 3: Send each resume's text plus your job requirements to OpenAI. Ask for a JSON response with: overall match score (1-100), individual criteria scores, matched skills, missing skills, relevant experience summary, potential concerns, and a recommendation (strong yes, yes, maybe, no).
Step 4: Write the results to Google Sheets, sorted by overall score. Add conditional formatting to highlight top candidates. Review the AI's reasoning for the top 20 and bottom 10 to validate accuracy. Adjust scoring weights if needed.
Cost breakdown
| Item | Cost | Notes |
|---|---|---|
| Google Apps Script | $0 | Free with Google Workspace |
| OpenAI API | $10-$25/mo | ~$0.04 per resume at 200-600 resumes/mo |
| PDF processing | $0-$5/mo | PyPDF is free; Adobe Extract for complex layouts |
| Setup time | 30-45 min | One-time per job role |
| Total monthly | $10-$30/mo | Saves 70% of manual screening time |
Frequently asked questions
Yes — AI can inherit biases from training data. Mitigate this by: removing names and photos before AI processing (blind screening), focusing scoring criteria on skills and experience rather than demographics, auditing results for adverse impact across protected groups, and having a human review every rejection. Never use AI as the sole decision-maker for hiring.
Yes. Most ATS platforms (Greenhouse, Lever, Workable) allow resume exports. Some have APIs for direct integration. The AI screening can run as a parallel process — candidates flow through your ATS normally, and the AI rankings appear in a separate dashboard for hiring managers to reference.