Skip to content
Arshed Ahmed
Back to projects

WhatsApp AI Order Bot

Conversational ordering and support on WhatsApp

AutomationSoftsora (Pvt) Ltd
WhatsApp AI cover (placeholder)

The problem

Customers already use WhatsApp every day, but answering every message by hand does not scale. The goal was to automate customer conversations on WhatsApp so people get answers quickly without waiting for a person.

My role

I developed and deployed the WhatsApp AI chatbot at Softsora using Python, LLM APIs and Twilio.

Stack

  • Python
  • LLM APIs
  • Twilio
  • WhatsApp

Architecture

Customers message a WhatsApp number connected through Twilio. Twilio forwards each message to a Python webhook service, which uses an LLM API to understand the message and write a reply, then sends the reply back through Twilio.

WhatsApp AI system architectureCustomers connects to Twilio. Twilio connects to Python service. Python service connects to LLM API.USERSCustomersWhatsAppCHANNELTwilioWhatsApp APISERVICEPython serviceWebhook + logicAILLM APIIntent + replies
WhatsApp AI system architecture

Key technical decisions

  • WhatsApp through Twilio, so the bot works in an app customers already use, with no new app to install.
  • A Python webhook service that keeps business logic separate from the messaging channel.
  • LLM APIs for understanding free-text messages instead of rigid menu trees.

Result

The bot was deployed and automated customer interactions on WhatsApp.

Links

There is no public link for this project yet. I am happy to walk through it in an interview.

Screenshots

  • WhatsApp conversation with the order bot (placeholder)
  • Order confirmation in WhatsApp (placeholder)