AI Lead Qualification Bot

When Every Missed Message Is a Lost Commission

In real estate, timing is everything. A prospect who sends a WhatsApp message at 9 PM asking about a 3BHK in Whitefield expects a response within minutes, not the next morning. By then, they've already spoken to three other brokers.

This national brokerage had 15 agents handling an average of 200+ inbound inquiries per day across WhatsApp, their website, and phone. The problem wasn't volume — it was triage. Agents were spending 6+ hours per day fielding basic questions ("What's the price of Plot #47?", "Is the 2BHK still available?") instead of closing deals.

What We Built

Conversational AI Architecture

The bot is built on LangChain with a custom retrieval-augmented generation (RAG) pipeline:

• Knowledge base: Ingested their entire property database (500+ listings) with details like price, location, amenities, availability status, floor plans, and nearby landmarks
• Conversation memory: Maintains context across messages so prospects can ask follow-up questions naturally
• Intent classification: Distinguishes between browsing (early-stage), comparing (mid-stage), and ready-to-buy (high-intent) leads
• Smart routing rules: High-intent leads (budget confirmed, timeline < 30 days, specific property shortlisted) get instantly routed to a human agent with full conversation context

WhatsApp Integration

We used the WhatsApp Business API (via Gupshup) with:

• Template messages for follow-ups: site visit reminders, price drop alerts, new listing notifications
• Rich media support: Floor plans, photos, and location maps sent directly in the chat
• Multi-agent routing: If the bot determines the prospect needs a specialist (e.g., commercial vs. residential), it routes to the right team with context

CRM Sync

Every conversation is logged in their CRM (Zoho) with:

• Lead score (automated based on qualification questions)
• Conversation transcript
• Properties discussed
• Next action (appointment booked, follow-up scheduled, or nurture sequence)

Training the Bot

We didn't just feed it data and hope for the best. The training process involved:

1. Analyzing 500+ real WhatsApp conversations from their top-performing agents to understand objection handling patterns
2. Building a prompt engineering framework that mirrors their sales methodology
3. Creating fallback handlers for edge cases: legal questions, negotiation requests, and complaint escalation
4. A/B testing greeting messages to optimize engagement rates (the winning version had a 34% higher response rate)

Results After 60 Days

The transformation was immediate. The bot now handles the first 2-3 interactions with every prospect — qualifying them, answering property-specific questions, and booking site visits. Human agents only engage when a lead is qualified and ready to move forward.

Sales team efficiency jumped 65% because agents now spend their time exclusively with high-intent prospects instead of answering "What's the price?" for the hundredth time. Booked appointments increased 40% because the bot responds at 2 AM, on Sundays, and during Diwali — when the agents don't.