Skip to main content
Back to case studiesAI / Automation

Lead Qualification & Follow-Up — Rules-Based Scoring, Not a Guess

An n8n workflow that enriches an inbound lead from its company site, scores it against a 10-rule config block, logs a clean CRM record with a plain-English explanation, and routes to a real booking link, an AI-drafted follow-up, or a human — the last of which happens for anything ambiguous or flagged as spam, never a silent discard.

n8nFirecrawlGemini AIAirtableResendCal.com
3 routing branches tested live, real CRM record on every run

The Problem

Most 'lead scoring' automations either hand the number to an LLM (unauditable, drifts silently) or auto-book everyone above an arbitrary line (books the wrong leads, burns the sales team's time). Neither one is trustworthy enough for a business to actually hand its inbound pipeline to.

Our Approach

  • 1Built the score as a deterministic rules engine in a single Code node — 10 additive/subtractive rules (service area, building type, size, frequency, urgency, decision authority, contact completeness, a real Firecrawl enrichment signal, referral source), never an LLM guessing at a number, so every score is auditable back to the exact rule that produced it.
  • 2A spam/junk override still routes to a human instead of discarding — verified live by sending deliberately low-quality test payloads and confirming a real CRM record and alert still get created for each one.
  • 3Booking confirmation closes the loop for real: a second webhook receives Cal.com's actual `BOOKING_CREATED` event and marks the CRM record booked, tried first via the lead ID threaded through the booking link's own metadata, falling back to an email lookup if that's missing — both paths tested against the real Airtable base.

What This Achieves

  • Every one of the three routing branches (human review, auto-book, AI follow-up) confirmed against real Firecrawl, Gemini, Airtable, and Resend calls — not simulated, not mocked.
  • A salesperson opens the CRM record and sees exactly why a lead scored what it did, in plain English, not a black-box number.
  • The scoring thresholds live in one editable config block, so the same workflow retargets to a different vertical or business without touching the routing logic.

Share this case study

Next Automation Case Study

Customer Support Agent — Grounded Answers, Verified Actions, Real Handoffs

Have something similar in mind?

Scope Your Project