WhatsApp Chatbot

WhatsApp AI Customer Support Bot — OpenAI + WhatsApp Business API + NestJS

Turn your WhatsApp Business number into a 24/7 AI support agent. It answers FAQs from your custom knowledge base, gracefully hands off to a human when it's out of its depth, and logs every conversation to your database — so nothing falls through the cracks.

Build My WhatsApp Bot

What We Build

A full-stack WhatsApp chatbot with AI at its core — not keyword matching or decision trees. Real language understanding powered by OpenAI, grounded in your specific business knowledge.

Custom Knowledge Base (RAG)

We ingest your FAQs, product documentation, policy pages, and support scripts into a vector database (Pinecone or pgvector). The bot retrieves relevant context before generating each answer — ensuring accuracy over creativity.

WhatsApp Business API Integration

We connect your WhatsApp Business number via an official BSP (WATI, 360dialog, or Twilio). Incoming messages are received via webhook, processed by the NestJS backend, and replied to within seconds.

OpenAI Conversational Engine

GPT-4o handles multi-turn conversations with full context of the conversation history. The system prompt enforces your brand tone, response length limits, and fallback behaviours — producing consistent, professional answers.

Human Handoff Logic

When confidence is low or the user asks for a human, the bot gracefully escalates — transferring the full conversation to a live agent in your shared inbox (WATI, Respond.io, or Chatwoot) with a single click.

Conversation Logging & Database

Every message is stored in PostgreSQL with user ID, timestamp, message content, intent classification, and resolution status. Optional admin panel for reviewing, filtering, and exporting conversation data.

Our Development Process

1. Knowledge Base Audit

We collect your FAQs, support docs, and product information. We clean, structure, and chunk them for optimal vector search performance.

2. NestJS Backend Setup

Scaffold the NestJS application with webhook endpoints, OpenAI service, vector search integration, and PostgreSQL database schema.

3. WhatsApp BSP Configuration

Set up your WhatsApp Business account with a BSP, configure webhook verification, and handle incoming message events for text, media, and quick replies.

4. AI Prompt & Handoff Logic

Engineer the system prompt, build the RAG retrieval pipeline, implement confidence scoring, and wire up the human escalation flow.

5. Testing & Production Deploy

Test with real WhatsApp messages across 50+ scenarios. Deploy to Railway or AWS, set up monitoring, and hand off with full documentation.

Tech Stack

Backend & AI
  • ✓ NestJS (Node.js)
  • ✓ OpenAI GPT-4o API
  • ✓ Pinecone / pgvector (RAG)
  • ✓ PostgreSQL / Supabase
Messaging & Infrastructure
  • ✓ WhatsApp Business API (WATI / 360dialog)
  • ✓ Respond.io / Chatwoot (handoff)
  • ✓ Railway / AWS (deployment)
  • ✓ Redis (session caching)

Why Choose Shapesky for WhatsApp Bots

RAG Not Just Prompts

We ground the bot in your actual knowledge base — not a generic LLM. Answers are accurate and hallucination risk is minimised.

Graceful Escalation

The bot knows when it's out of its depth and hands off cleanly — building customer trust instead of eroding it with bad answers.

Production NestJS Architecture

Not a weekend hack. Clean, modular NestJS code with proper error handling, logging, and test coverage — ready for production traffic.

Conversation Audit Trail

Every interaction is logged so you can review quality, identify gaps in your knowledge base, and continuously improve the bot.

Frequently Asked Questions

We build a custom knowledge base from your FAQs, product docs, and policy pages. The bot uses OpenAI with RAG (retrieval-augmented generation) to pull the most relevant answers from your knowledge base before responding.

The bot detects low-confidence situations and hands off to a human agent via a shared inbox (WATI, Respond.io, or Chatwoot). The full conversation history is transferred so the agent has full context.

All conversations are logged to a PostgreSQL or Supabase database with timestamps, message content, user ID, and resolution status. An optional admin dashboard lets you review and export conversations.

A single-language bot with a knowledge base up to 50 FAQs and human handoff starts at $2,500. Multi-language bots with custom dashboards and CRM integration range from $4,500 to $10,000.

Ready to Give Your WhatsApp a Brain?

Let's build a WhatsApp support bot that answers your customers' questions 24/7 — accurately, consistently, and on brand.

Get in Touch