How to Set Up a WhatsApp AI Bot for Your Indore Business in 2026

Over 500 million Indians use WhatsApp. For Indore businesses — whether a coaching institute on MP Nagar, a real estate developer on Hoshangabad Road, or a clinic near Bombay Hospital — WhatsApp is where leads arrive first. Not Instagram. Not a website contact form. WhatsApp.

The problem is not the lead volume. It is what happens next. A human can handle 3 to 5 simultaneous WhatsApp conversations. After business hours, they handle zero. An AI bot handles 500 simultaneously, at 3 AM on a Sunday, in Hindi, Hinglish, and English — without missing a single enquiry, without a bad mood, and without variable quality based on who is on shift.

This guide covers the complete process for setting up a production-grade WhatsApp AI bot: from choosing the right API route to the specific metrics you should be watching after launch. Every step is based on what AdsVerse has built for 113+ clients across Indore and Madhya Pradesh.

Step 1: Choose Your API Route — BSP or Direct Meta Cloud API

Before anything else, you need to decide how you will access the WhatsApp Business Platform API. There are two routes — and this decision shapes every subsequent step.

The 9-Step Setup Process

1. Verify your Meta Business Account Go to business.facebook.com and create or log into your Meta Business Account. Complete business verification by submitting your GST certificate, shop act licence, or bank statement with business name. Verification takes 2 to 5 business days in India. This step is non-negotiable — without a verified Meta Business Account, API access is not available. If your Meta Business Account is already verified for running Facebook or Instagram ads, you can skip directly to Step 3. Tip: Use your official business name exactly as it appears on your GST certificate. Mismatches between your submitted name and the document are the most common reason for verification delays.
2. Register a dedicated WhatsApp Business phone number Get a phone number that is not currently registered on WhatsApp or the WhatsApp Business App. A new SIM card is the simplest option — Jio, Airtel, or Vi all work. You can also use a landline number that receives OTPs via call. Virtual numbers from providers like Twilio are supported. This number becomes your business-facing WhatsApp identity. Choose it carefully — changing the number after setup disrupts all active conversations and requires re-verification. Tip: If you want to migrate an existing number, remove it from your current WhatsApp app first and wait 24 hours before registering it on the API. Skipping this wait causes registration failures.
3. Apply for WhatsApp Business Platform API access In Meta Business Manager, navigate to Business Settings → WhatsApp Accounts → Add. Register your phone number, verify it via OTP, and submit your display name and business category for Meta review. Approval takes 1 to 3 business days. Once approved, Meta provides your API credentials: a permanent Access Token and a Phone Number ID. Store these securely — the Access Token is your master key to the entire bot system and cannot be recovered if lost without re-verification. Tip: Set your WhatsApp display name to exactly your business brand name. Meta rejects names that look like generic keywords — "Best Coaching Indore" will be rejected; "Sharma Academy" will be approved.
4. Map your lead qualification conversation flow before building anything This is the step most businesses skip — and it is the single most important one. Before configuring any tool, document your bot conversation as a written flowchart: the opening greeting, the 3 to 5 qualification questions the bot asks every lead, the exact scoring logic (what combination of answers makes a lead hot, warm, or cold), and the exit paths for each score category. For an Indore real estate client, qualification questions might be: budget range, property type (residential or commercial), area of interest (Vijay Nagar, Hoshangabad Road, or elsewhere), and buying timeline. For a coaching institute: target exam, current preparation stage, batch start preference. Tip: Limit qualification to 3 to 5 questions maximum. Beyond 5 questions, drop-off rates increase sharply. Get the minimum viable information to score the lead, not a full application form.
5. Set up your webhook server to receive Meta API events Meta pushes all incoming WhatsApp messages to a webhook URL you provide — a server endpoint that receives the message data and decides how to respond. You need a server running HTTPS with a valid SSL certificate. Options: a VPS (DigitalOcean, AWS Lightsail — starts at ₹600/month), Railway.app for Node.js deployments, or Google Cloud Run for serverless deployments. Your webhook receives a JSON payload for every message, processes it, and calls the Meta API to send a reply within 15 seconds before Meta times out the session. Tip: Use ngrok during development and testing to expose your local server to Meta's webhook callbacks without deploying to a live server yet.
6. Integrate Gemini API for natural language processing Create a project in Google AI Studio (aistudio.google.com), generate a Gemini API key, and integrate it into your webhook handler. When a message arrives, your server sends the full conversation history plus the new message to Gemini with a system prompt that defines your bot's persona, business context, qualification criteria, and response boundaries. Gemini returns a natural language response that your server then sends back to the lead via the Meta API. The system prompt is the most critical configuration — it defines everything the bot knows and how it behaves. For an Indore coaching institute, the system prompt includes: what courses are offered, batch timings, fee structure, and the 5 qualification questions to ask in sequence. Tip: Always instruct Gemini in your system prompt to respond in the same language as the user. This enables automatic Hindi, Hinglish, and English switching without any extra code.
7. Build n8n workflows for automated lead routing and CRM entry Install n8n on your server (self-hosted is free; n8n Cloud starts at ₹1,500/month) and build workflows triggered by your webhook when a lead completes the qualification flow. Hot lead workflow: create a CRM entry in Zoho, HubSpot, or Google Sheets with the lead's name, number, conversation summary, and score → send a WhatsApp or Slack notification to your sales team with the full context → assign a follow-up task due in 2 hours. Cold lead workflow: enrol the contact in a nurture sequence — an automated WhatsApp message at day 3 ("Still thinking about it? Here is what other [Indore businesses / students] have said"), day 7, and day 14 before marking the lead as inactive. n8n handles all of this without a single manual step. Tip: Log every lead — hot and cold — to a Google Sheet with timestamp, score, and conversation summary. This creates a searchable lead history and the data source for your monthly performance review.
8. Test with 20+ real conversation scenarios before going live Run a minimum of 20 test conversations covering: the expected ideal path (lead answers all questions perfectly), common Indian language variations (Hindi, Hinglish), unexpected inputs (lead sends a voice message, sends an image, answers out of sequence), edge cases (lead asks a question the bot is not designed to answer, lead says they already spoke with someone from your team), and error conditions (API timeout, Gemini rate limit). Verify that every hot lead correctly appears in your CRM. Verify that every cold lead enters the nurture sequence. Verify that your sales team receives the correct notification format. A production bug in a WhatsApp bot is visible to real customers — fix it before launch, not after. Tip: Ask 2 or 3 actual team members to test the bot as if they were customers. Internal testers who know the expected flow miss edge cases that real leads will hit.
9. Go live and monitor four key performance metrics Launch your bot and track these four metrics daily for the first 30 days. Response rate: the percentage of leads who reply to the bot's opening message — benchmark is 55 to 70%. Qualification completion rate: the percentage of respondents who answer all qualification questions — benchmark is 40 to 60%. Hot lead rate: the percentage of completed conversations scoring as qualified — this varies by business and campaign quality. Conversion rate: the percentage of hot leads that become actual sales — this is your true north metric. If response rate is low, rewrite your opening message. If completion rate is low, your questions are too many or too intrusive. Optimise based on data, not assumptions. Tip: Review the bot's conversation logs weekly for the first month. Real conversation data reveals misunderstandings, missing FAQs, and scoring logic errors that no testing scenario can fully predict.

How a Completed WhatsApp AI Bot System Works

Common Mistakes That Kill WhatsApp Bot Performance

FAQ — WhatsApp AI Bot Setup for Businesses