Track daily job applications from Gmail to Google Sheets via GPT-4o-mini with approval

Who's it for

Active job-seekers who apply to 10–30 roles a week and lose track of which application is at what stage. Built for anyone tired of manually updating a spreadsheet every time a recruiter replies — or worse, realizing their tracker has been lying to them for two weeks.

What it does

This workflow runs every morning at 9 AM, scans your Gmail for job-application emails, and uses GPT-4o-mini to extract three things from each one:

• Company (the actual hiring company, not the ATS platform)
• Position (cleaned of boilerplate like "the role of")
• Status (Applied, Interview, Offer, Rejected, or Other)

It then writes structured rows into a Google Sheet — one row per application, identified by a Match Key of Company | Position. New applications get appended automatically. Status changes for existing rows (Applied → Interview, Interview → Offer / Rejected) are gated behind a one-tap email approval, so an LLM misclassification can never silently demote a row.

How it works

Daily 9 AM Trigger ↓ Fetch Gmail (filtered by ATS senders & job-keyword subjects, last 2 days) ↓ Process Email Data (clean & truncate body to 4000 chars) ↓ Extract Key Information (GPT-4o-mini, structured JSON output) ↓ Format Data (build Match Key) ↓ Filter Valid Job Updates ──→ Other ──→ "Other app" sheet tab ↓ Branch by Status ├─ Applied ──→ Track New Application (append row) └─ Interview / Offer / Rejected ↓ Aggregate into single digest ↓ Email & Wait for Approval (sendAndWait) ↓ Merge → Split → Update Application (match by Match Key)

Key design choices:

• Match Key, not Thread ID, drives deduplication. Recruiters often start new threads for interview scheduling, but the same Company | Position should stay one row.
• Approval gate only on updates, not appends. New applications go in directly to keep your inbox quiet. Only status changes — where misclassification matters — require your tap.
• Aggregation before approval. You get one digest email per morning, not one per status change. Skim, approve, done.

Setup

Prerequisites:

• n8n 1.x (cloud or self-hosted)
• Gmail account connected via OAuth2
• Google Sheets account (same OAuth2 credential works)
• OpenAI API key with gpt-4o-mini access

Steps:

1. Create your tracker sheet. Make a Google Sheet with two tabs:

• Sheet1 — columns: Thread ID | Match Key | Date | Company | Position | Status | Modified Date
• Other app — same columns (used for unclassifiable emails)

1. Import this template. Open n8n → Workflows → Import from File. n8n will flag missing credentials.

2. Configure three credentials:

• OpenAI → used by the Open AI node
• Gmail OAuth2 → used by Fetch Job Application Emails and Email Updates & Wait for Approval
• Google Sheets OAuth2 → used by all three Sheets nodes

1. Replace placeholders (search the workflow for these strings):

• YOUR_GOOGLE_SHEET_ID → your real sheet ID (from the sheet's URL)
• [email protected] → the address that should receive approval digests

1. Test it manually. Click "Execute workflow" and verify the Gmail node returns recent emails, the LLM extracts clean JSON, and a sample appears in your sheet.

2. Activate. Toggle the workflow to active and it will run daily at 9 AM.

Customization

• Schedule. Change Daily 9 AM Trigger to any cron expression.
• Lookback window. The Gmail filter ends with newer_than:2d. Bump to 7d if your daily run sometimes fails — the workflow handles re-processed emails idempotently because Match Key updates are no-ops on identical data.
• Languages. The prompt in Extract Key Information includes English and French status phrases. Add your language's equivalents to each section (Rejected, Applied, etc.).
• ATS coverage. The Gmail query covers ~25 ATS providers (Lever, Greenhouse, Ashby, Workday, SmartRecruiters, Welcome to the Jungle, etc.). Add or remove domains to match your job market.
• Approval timeout. Add options.limitWaitTime.values.resumeAmount (in hours) to the approval node if you want stale digests to auto-cancel.

Honest caveats

• It's only as accurate as the emails it reads. Recruiters don't follow a standard format. Misclassifications happen — the human-in-the-loop approval step is your safety net. Read the digest before tapping approve.
• Status downgrades are possible. A delayed "thank you for applying" email arriving after an interview invite will offer to demote the row. The approval gate exists for exactly this case.
• Privacy. Email subjects and bodies are sent to OpenAI for processing. Don't use this if your job-search emails contain NDAs or confidential info — swap in a local LLM instead.
• Cost. Roughly $0.01–0.05 per day at typical job-search volume. Gmail and Google Sheets API calls stay within free quotas.

Why I built this

Manual spreadsheet tracking lasted me two weeks before it became the chore I'd skip on busy days. By week three the sheet was lying to me — half the "Applied" rows were rejections I hadn't logged yet. So I built this. It's not a system of record; it's a guided draft I skim once a week. The bones might be useful to someone else.