Home: Iago Bonifácio

Consulting projects

From low-code flows to an AI-orchestrated lead pipeline

Lead qualification pipelines (WhatsApp capture, AI qualification, CRM sync) built in n8n, Jitterbit and SkyOne, and now the same problem with an agent that decides each step.

A WhatsApp message reaches an AI agent that qualifies it, updates the CRM and logs every decision.WhatsAppcaptureAI agentqualifiesCRMsyncLogevery decision
A WhatsApp message reaches an AI agent that qualifies it, updates the CRM and logs every decision.

Context

Lead qualification (WhatsApp capture, AI qualification, CRM sync) is a problem I have solved with several tools over the years.

Problem

In a visual builder, every new qualification rule means editing a fixed graph. The goal was an agent that reads the message, decides how to qualify it and updates the CRM.

What I did

  • Earlier versions of the pipeline in n8n, Jitterbit and SkyOne: WhatsApp capture, an AI qualification step and CRM sync.
  • Rebuilt it with Claude and the Cursor SDK driving each step, with no visual builder.
  • Same reliability floor: every decision is logged and traceable, with no manual intervention.

Result

Routing no longer depends on a visual builder. A new qualification rule changes in the agent’s instructions, not in the graph.