WhatsApp AI Sales Assistant

Context

The customer acquisition system created a steady stream of WhatsApp inquiries. However, each new conversation still required a person to answer recurring questions, gather project details, assess early intent, and record relevant information before a consultation could happen.

I built a WhatsApp AI Sales Assistant to support that first stage of the journey. The system is in final pre-launch testing. It is not yet live, so this case study does not claim production impact or performance metrics.

What I built

• A WhatsApp AI assistant for early-stage customer conversations.
• An n8n workflow with conditional routing and escalation to a human team member.
• A knowledge layer using Google Docs and conversation context through Simple Memory.
• Google Sheets integration for lead records and operational data.
• A human-handover path for consultation-ready or exception-based conversations.

Workflow

Customer WhatsApp message
  ↓
n8n workflow
  ↓
Conversation and customer lookup
  ├─ Existing context: Simple Memory + Google Sheets
  └─ Service knowledge: Google Docs
  ↓
Conditional routing
  ├─ FAQ or early inquiry → AI response
  ├─ Data collection / qualification → structured follow-up
  └─ Consultation-ready or exception → human handover
  ↓
Lead record update in Google Sheets

The goal is not to replace human consultation. The system is designed to make early conversations more consistent and hand the right context to a person at the right time.

Outcome

The assistant is in final pre-launch testing. The implementation has been built and tested, but it has not been launched on a live production number. As a result, there are no customer-volume, conversion, response-time, or revenue claims in this case study.

Area	Before	Current state
First response	Handled manually by the team	Workflow built and undergoing final pre-launch testing
Lead context	Collected manually during conversations	Designed to capture context through structured routing and Google Sheets updates
Escalation	Managed manually	Designed to hand over qualified or exception-based conversations to a human team member

Visuals

Tools used

n8n · WhatsApp · OpenRouter · DeepSeek · Google Sheets · Google Docs · Simple Memory

Once the system is launched, the next step is to validate its performance in real conversations: response quality, reliable data capture, appropriate human handover, and its effect on the speed and consistency of first responses.