AI Customer Service Automation 🤖

Turning "DM for Orders" into a scalable AI-powered system

🧠 Overview

Designed an AI-assisted workflow for a WhatsApp-based hoodie business handling ~300 daily messages. Improves response time, consistency, and scalability.

🚨 Problem

⚙️ Solution

AI generates responses → automation handles simple queries → human reviews complex ones.

Automation Logic: Simple queries are auto-answered, complex queries are escalated to a human.

🔄 Workflow

📩
Message
🤖
AI
👀
Review

Send

🧩 System Architecture


Customer

WhatsApp

Automation

AI Engine

Decision

Storage

🧠 Prompt Engineering

You are a helpful customer service assistant for a hoodie business in Johannesburg.

Rules:
- Be friendly and professional
- Keep responses short
- Include price when asked
- Mention location (Johannesburg)
- Mention delivery
- Encourage ordering

Example:
"Hi 😊 The hoodie is R350. We have stock available and we are based in Johannesburg. We also deliver nationwide 🚚."

📊 Performance Metrics

300+

Messages / Day

80%

Automated

Instant Replies

Sales Impact

These metrics are illustrative and based on a typical small business scenario handling high-volume WhatsApp enquiries. They demonstrate the potential impact of AI-assisted responses on response time and message handling capacity.

📈 Impact Simulation (Before vs After AI)

💬 Live AI Interaction

Hi, how much is this?
AI is typing...

📈 Impact

🙋 My Role

Back to Portfolio