AI Lead Qualification Assistant with Live Scoring (n8n)

A bookkeeping firm needed a way to qualify inbound leads without a person manually asking the same four questions on every call.

I built a lead-qualification assistant that collects revenue, transaction volume, business type, and timeline through a natural conversation, scores each lead from 0 to 3 using admin-editable thresholds, and saves a clean structured record. Thresholds can be updated directly from the owner panel, with the new scoring logic applied to the very next lead and verified end to end.

Design principles

Behavior is configuration

Scoring thresholds are stored as database values, not hardcoded in the flow. An owner can update them, and the new settings apply immediately to the next lead — no redeploy, no touching the workflow itself.

Conversation as the interface

The four qualifying details are collected naturally through dialogue, one step at a time, instead of a form. That keeps the intake experience simple for the lead while still gathering everything the scoring model needs.

State lives outside the conversation

A lead’s progress is tracked server-side across messages rather than in the chat history itself. Once a qualification is completed, the system starts a fresh flow for the next lead, which makes it suitable for continuous intake rather than one-off sessions.

Scoring grounded in collected data

The priority score is calculated directly from the qualification fields gathered during the conversation — revenue, transaction volume, business type, and timeline — not from a separate judgment call layered on top.

Verify through real behavior

Dynamic scoring was tested by changing a live threshold in the owner panel and running a new lead through the flow, confirming the updated configuration actually changed the resulting priority score, not just the number on a settings screen.

Stack

n8n for the conversation and scoring flow, GPT for the natural-language exchange, and Supabase for storing thresholds and qualified leads.

Try it

A live version of the assistant is running at /chatbots/leadqual.