n8n Resume Screening That Books the Interview, Not Just Scores It

Scoring Resumes Is the Easy Part. Most Workflows Stop There.

An n8n resume screening workflow that scores candidates and dumps them in a spreadsheet has done maybe half the job. The candidates you actually want are the ones you lose to a slow follow-up. They get a faster reply from another company while your shortlist sits in a sheet waiting for someone to notice it.

The templates ranking for this all stop at score-and-log. Score the resume, rank the candidates, write to Google Sheets, done. None of them close the loop into scheduling: the auto-invite, the calendar slot, the Slack ping to the hiring team. That's the part that wins candidates, and it's the gap this post fills.

n8n is well suited here because the whole funnel is connected nodes. The resume comes in, gets scored, and the strong ones flow straight into interview booking without a human copying anything between tools.

What You Can Automate in Screening

A complete screening workflow runs from inbox to booked interview:

• Resume intake: a Gmail or Drive trigger catches every incoming application
• Text extraction: pull readable text from PDF and DOCX attachments
• Rubric scoring: an AI node scores each candidate 0-100 against fixed criteria with reasons
• Consistent ranking: a per-criterion breakdown so scores stay comparable across candidates
• Shortlist routing: candidates above the threshold move on; the rest get logged and a polite decline
• Interview invites: auto-send a booking link and create the calendar event
• Team notification: Slack the hiring team when a strong candidate clears screening

The Resume Screening Pipeline

Gmail / Drive Trigger (new application)
  → Extract (PDF/DOCX → text)
  → OpenAI (score 0-100 against rubric → per-criterion JSON + reason)
  → Code (parse JSON, compute weighted total)
  → Airtable / Google Sheets (log candidate + score + reasons)
  → IF (score >= threshold?)
      → shortlist:
          → Gmail (interview invite + booking link)
          → Google Calendar (create slot)
          → Slack #hiring (strong candidate: name, score, summary)
      → below:
          → Gmail (courteous decline, on a delay)

1. Catch the application

A Gmail trigger filtered to a careers label, or a Drive trigger on an applications folder, picks up each resume as it lands. Filtering at the trigger keeps unrelated mail out of the pipeline. For high volume, point applicants at a single intake address and let the label do the routing.

2. Extract the text

Resumes arrive as PDF and DOCX. Extract the text before scoring; the model needs words, not a binary blob. n8n's extract-from-file capability handles common formats. Watch for image-only PDFs (a scanned resume), which need an OCR step or they'll score as empty.

3. Score against a fixed rubric

This is where consistency lives or dies. Use one rubric, ask for a structured breakdown, and never change it mid-batch:

You are screening candidates for [ROLE]. Score each criterion 0-20:
- Relevant experience
- Required skills match
- Seniority fit
- Communication (from the writing)
- Red flags (gaps, job hopping) — score inversely

Return ONLY JSON:
{ "criteria": { "experience": n, "skills": n, ... },
  "total": sum, "reason": "two sentences", "flags": ["..."] }

Requiring a per-criterion score does two things. It keeps candidates comparable, because everyone is judged on the same axes. And it makes the score auditable, so a human can see why someone got an 82 instead of trusting a bare number. Scores drift the moment the rubric is vague, so version it per role and keep it identical across the batch.

A note on responsible use: an AI score is a triage signal, not a hiring decision. Keep a human in the loop on every shortlist, watch for criteria that proxy for protected characteristics, and log the reasons so decisions stay reviewable. The workflow ranks; people decide.

4. Always parse the model output

The OpenAI node returns text. A Code node parses the JSON into fields before anything branches on the score. Skip this and the IF node compares undefined to your threshold and routes everyone the same way. Add the parse step every time, even when it feels redundant. The execution log makes a missing parse obvious; the silent mis-routing it causes is much harder to spot.

5. Close the loop into scheduling

Here's the part the ranking templates miss. When a candidate clears the threshold, don't just log them. Send the interview invite with a booking link, create the Google Calendar event, and post to the hiring team's Slack with the name, score, and two-sentence summary. The below-threshold branch sends a courteous decline, ideally on a short delay so it doesn't arrive sixty seconds after they hit submit. Closing this loop is the whole point: a strong candidate hears back in minutes, not days.

Implementation Patterns

Pattern 1 — Rubric as a versioned constant. Store the rubric once and reference it, rather than retyping it per workflow. Same role, same rubric, every candidate. The VA Hiring Workflow template ships a scoring rubric plus the shortlist-to-scheduling branch already wired, so the loop closes out of the box.

Set (rubric constant) → OpenAI (score) → Code (parse + weighted total)

Pattern 2 — Threshold routing, both branches handled. Don't only act on the winners. Route the below-threshold candidates to a polite, delayed decline. A funnel that ghosts rejected applicants damages the employer brand the role is trying to build.

IF (score >= 75) → invite + calendar + Slack ; else → delayed decline

Pattern 3 — Human-in-the-loop on the shortlist. Auto-book the slot, but flag the shortlist for a human to confirm before the offer stage. Automation handles the speed; people handle the judgment.

n8n Nodes You'll Use Most

Getting Started

1. Add a Gmail or Drive trigger filtered to your applications inbox or folder.
2. Add an Extract From File step to read PDF and DOCX text.
3. Write a fixed rubric and score with an OpenAI node returning per-criterion JSON.
4. Add a Code node to parse the JSON and compute the weighted total.
5. Log every candidate to Airtable or Google Sheets with the reasons.
6. Add an IF node on the threshold: invite plus Google Calendar plus Slack above, delayed decline below.
7. Keep a human reviewing the shortlist, then start from a template instead of wiring the scheduling loop by hand.

The screening half of this connects to the wider funnel covered in automating your hiring pipeline with n8n, and the people-ops side continues into n8n HR automation once a candidate becomes a hire.