Project Description
WhatsApp Voice AI Bot (Multi-language, n8n + OpenAI + WATI)
Project Overview:
We are building a production-level WhatsApp Voice AI system for a ride-hailing company with 900+ drivers. The system must support voice-first communication, multiple local languages, and a combination of structured workflows + AI responses.
This is not a basic chatbot we need a scalable, reliable conversational system.
Scope of Work:
1. WhatsApp Integration (WATI)
Setup WATI and connect WhatsApp Business API numbers
Configure webhooks (send/receive messages)
Support multiple numbers with shared backend
2. Voice AI Pipeline
Voice input → Whisper transcription (Urdu, Pashto, Punjabi, Saraiki)
AI processing (GPT-4o / Claude)
Text → Urdu voice (TTS)
Voice input → voice + text reply
Text input → text reply only
3. Intent Routing System
Build structured flows for:
Driver registration (multi-step)
Bonus/payment queries
Top-ups
Ride/account issues
Office info + FAQs
Angry drivers → instant escalation
4. Hybrid Logic (Flows + AI)
Fixed flows for critical processes (registration, payments, escalation)
AI for general queries (KB-based only, no hallucination)
5. Session & Context
Maintain per-driver conversation memory
Handle multi-step interactions
6. Escalation System
Detect frustration or critical cases
Generate ticket ID
Send full transcript to support via WhatsApp
Allow human agent to continue conversation
7. Reliability
Voice reply must always work (fallback TTS required)
Error handling + retries
Low response time
8. Architecture
n8n (or Make) for workflows
Optional Python for logic/scaling
Design for scaling (100 → 500 msgs/day)
Deliverables:
Fully working WhatsApp AI system
Voice input/output pipeline
Intent routing + flows
Escalation + alerts
KB integration
Tested with real users
Requirements:
Experience with WATI/Twilio (WhatsApp API)
OpenAI / Claude integration
Whisper + TTS experience
n8n / Make workflows
Strong backend/system design
Notes:
Voice UX is critical (low-literacy users)
Focus on reliability and clean architecture
Long-term work possible after delivery
Timeline Total: (12 days)
Part 1: 5 Days
Part 2: 7 Days
Budget full and final:
$100 NZD