Hiring Funnel Automation
An online maths school hiring tutors in Indonesia at volume. Leads came from a Facebook quiz into a Google Sheet. Conversion from lead to working tutor was around 0.1%, so the channel only works if processing a lead costs almost nothing.
Problem
Everything after the sheet was manual: open a row, write to the person on WhatsApp, wait, remind, get a CV, read it, decide. Leads waited for hours, went cold, and recruiters hated the channel. We needed a first reply in minutes and a recruiter only for people worth talking to. I decided to build it myself instead of bringing in a contractor.
Solution
An n8n + WhatsApp bot that takes each lead from the sheet, knocks out obvious misfits, asks the rest for a CV, scores it with an LLM and books the good ones onto a group interview.
- leads invited to a group interview
- 16% → 26%
- leads invited to a group interview
- leads over one weekend
- 1,200
- leads over one weekend
- LLM scoring for three months
- $0.90
- LLM scoring for three months
How it works
Intake, every 10 minutes
New quiz rows are synced, normalized and checked: name, phone, teaching experience. No teaching experience means a quiet rejection, no message.
First touch
A WhatsApp template introduces the role and the next steps (CV review, then an interview) and asks for a CV as a PDF.
Inbound routing
A PDF goes to text extraction and scoring. No PDF yet: a short follow-up asking for the CV "specifically in PDF format", at most twice. Already in review: "thanks for applying, a recruiter will contact you if your profile fits".
Scoring
The LLM gets the role, the must-haves (teaching experience, maths background) and the pluses (online teaching, a relevant degree), and returns a 1–10 score, a short summary and red flags. The threshold is one the recruiting team can move.
Invite and remind
Every 15 minutes good candidates get a booking link for a group interview (Cal.com and a shared Google Calendar), reminders every 30 minutes until they book, and one more 24 hours before.
Dashboard
Pipeline by status (waiting for CV, ghosted after 2 reminders, needs review, invited, rejected, scoring error), AI summaries, HR assignment, one-click upload to Huntflow, and a "needs attention" panel when scoring fails.
Outcome
- Share of leads invited to a group interview: 16% → 26%.
- 1,200 leads handled over one weekend with nobody at a keyboard.
- 87 of 153 invited candidates booked a group interview within 2 hours.
- When we reused the bot for sales hiring, 76% of invited candidates showed up to the group interview, vs 43% and 28% in the two months before.
- LLM scoring cost about $0.90 for three months.
What I did
- Designed the flow and the knockout rules with the recruiting team.
- Built the workflows in n8n: lead intake, state routing for inbound messages, timers for invites and reminders, booking.
- Wrote the CV scoring prompt and the decision logic.
- Built a recruiter dashboard on top of it, then connected Huntflow.
- Handed it over to the recruiting team with docs, three times, as the team changed.
Screens
My role
Designed and built it myself, then trained and handed it over to the recruiting team.
Stack
n8n · Google Sheets · WhatsApp Business API · OpenAI · Cal.com + Google Calendar · Huntflow · Next.js dashboard on Vercel.
Next case
AI Hiring Copilot (Clarifit)
