AI WhatsApp Lead Bot for Argaman Media

ai · whatsapp · crm · leadgen

The problem

Argaman Media's leads arrive on WhatsApp. Nobody watched the inbox outside office hours, so a question about pricing sat until the next morning. And what the prospect said in the chat stayed in the chat: unless someone opened WhatsApp and re-typed it into Pipedrive, the CRM never saw the lead at all.

What was built

An n8n workflow behind the WhatsApp Business number:

  1. Intent classification. An LLM decides whether the message is a price question, a capability question, or a request for a human. Not keyword matching.
  2. Grounded answers. FAQ replies come from the agency's own material, so the bot does not answer for services the agency does not sell.
  3. Qualification. The bot collects the fields that decide whether a lead is worth a call, in the order a person would ask for them.
  4. Booking. Qualified prospects go to scheduling.
  5. CRM write-back. The lead is created in Pipedrive with custom fields filled from the conversation, not as a bare phone number.
  6. Consent and audit trail. Opt-in is recorded and every exchange is stored, which GDPR requires once you are processing personal data.

Anything consultative is handed to a person, with the conversation attached.

Why n8n rather than a chatbot platform

The agency holds the workflow as a file it can version, edit and move. When Pipedrive changes, the workflow changes; there is no vendor between the agency and its own logic.

The cost is that someone has to operate it, so the build includes an error workflow, alerting on failed executions, and scheduled pruning of execution data before it fills the database. A bot that stops answering without anyone noticing costs more than not having one.

Stack

  • Orchestration: n8n (self-hosted)
  • Language model: OpenAI, for intent classification and grounded FAQ answers
  • Channel: WhatsApp Business API
  • CRM: Pipedrive, with custom fields written from the conversation

Credentials

Our Apollo.io node is published on npm as n8n-nodes-apollo (MIT, maintainer arturl95, 14 resources, source public). We hold n8n Verified Creator status.


Want something like this built?

Tell us what the workflow needs to do and we will tell you whether it is a fit, what it would take, and what it would cost.

Start a conversation