WhatsApp support assistant for an online store

The support team answered the same questions about sizes, delivery and order status all day, and messages after office hours waited until the next morning.

A conversational assistant trained on the product catalog and store policies answers in English and Bangla, checks order status through the store API, and passes complex cases to a human with the full conversation.

1. Inputs01

   • WhatsApp messages
   • Product catalog
   • Store policies

2. Intelligence02

   • Conversational model
   • RAG knowledge base
   • Intent detection

3. Integrations03

   • Store order API
   • Helpdesk inbox
   • Webhook handler

4. Outputs04

   • Instant answers
   • Order status replies
   • Human hand-over

1. 01 · Message

   Customer asks

2. 02 · Detect

   Intent found

3. 03 · Retrieve

   Knowledge search

4. 04 · Act

   Order lookup

5. 05 · Reply

   Answer sent

6. 06 · Escalate

   Human if needed

• OpenAI
• WhatsApp API
• Pinecone
• Node.js
• Shopify API

1. 01

   Discovery

   Collected the 50 most common customer questions.

2. 02

   Design

   Defined tone, languages and when to hand over to a human.

3. 03

   Build

   Built the knowledge base, order lookup and WhatsApp flow.

4. 04

   Launch

   Ran with the team watching, then opened to all customers.

• Sample · not verified

  —

  Messages answered automatically

  Replace with your measured result.

• Sample · not verified

  —

  After-hours coverage

  Replace with your measured result.

• Sample · not verified

  —

  Support workload

  Replace with your measured result.