AI Ecommerce Customer Support System — V1

The problem

Routine shipping, policy, and order questions require information scattered across business tools. Cases needing staff attention must reach the right people with useful context.

The solution

A coordinated n8n workflow receives WhatsApp messages through Evolution API, lets an AI agent select the appropriate tool, and routes a structured result into a customer reply, a clarification request, or an escalation.

Architecture with a purpose.

WhatsApp is the customer channel, Evolution API connects messages to n8n, and the main agent coordinates specialized tools. Workflow routing keeps the final outcome explicit.

{
  "status": "reply | clarification | escalation",
  "customer_reply": "…",
  "reason": "…",
  "priority": "…"
}

Clear roles. Useful tools.

• Retrieve shipping, returns, and store policies from Google Sheets.
• Read Shopify order information to answer order-related questions.
• Identify cases needing human support and notify staff on WhatsApp.
• Return status, customer_reply, reason, and priority in structured output.

From request to response.

1. Customer sends a WhatsApp message.
2. Evolution API passes the message into the n8n workflow.
3. The AI agent interprets the request and selects a policy, order, or escalation tool.
4. Structured output is validated and routed by workflow logic.
5. The customer receives a reply or clarification; staff receive escalated cases.

Technology stack

Challenges & learning.

Design challenges

The design needs to connect open-ended customer messages with predictable workflow branches. Clear tool responsibilities and structured output make those boundaries explicit.

What I learned

This project brought tool selection, API integration, sub-workflows, and structured agent responses into one support system.

Result & current status

A completed support workflow that retrieves business information and routes customer requests. No client performance metrics are claimed.

Future improvements.

Evolve toward persistent customer context and controlled order operations through V2. Validate additional edge cases and improve escalation context.