WhatsApp AI Chatbot Development in India
A custom AI chatbot on your WhatsApp number: how it works, what Meta charges as of October 2026, build cost, human handoff, and when a tool is enough.
- Organisation
- WavX Solutions
- Telephone
- +919310079927
Description
Service WhatsApp AI chatbot development
WavX builds AI chatbots on the WhatsApp Business Platform that answer from your own information in English, Hindi and Hinglish, act on your systems, and pass the chat to a person when needed. As of October 2026 Meta charges nothing for free-form replies inside the 24-hour window a customer opens; it charges per delivered template message.
Talk to a technology expert Last updated 2 October 2026
What it is
A WhatsApp AI chatbot answers customers on your business number in ordinary language. It reads the message, finds the answer in your own information, replies in the customer's language, and can do things in your systems: check an order, book a slot, create a lead, share a payment link. When it should not answer, it passes the chat to a person.
It runs on the WhatsApp Business Platform through Meta's Cloud API, not on the WhatsApp Business app on a phone. Setting up the number, the Meta business account and the message templates is covered on the WhatsApp Business API integration page. This page is about the AI on top of it.
How it works
Drawn as a diagram, one message travels through eight steps:
The customer sends a message to your number.
Meta delivers it to your server as a webhook .
The server identifies the customer by phone number and loads the conversation so far and any CRM record.
The message type is handled. Text and button taps are read directly. Images and documents are stored. Voice notes need speech-to-text first, if you choose to support them.
Language and intent are detected. If the customer has asked for a person, or the topic is on your handoff list, the chat goes to step 8.
The answer is prepared. Relevant passages are fetched from your documents ( retrieval-augmented generation ) and sent with the question and the instructions to the model.
The reply is checked and sent through the Cloud API as a free-form message, with buttons or a list where they help.
Handoff. The conversation appears in your team's inbox or helpdesk with its history. The bot stays silent on that chat until a person releases it.
Every step is logged, so a wrong answer can be traced to the message, the passages and the reply.
The rules Meta sets, and what they mean for the bot
A WhatsApp bot is built inside Meta's rules. These are the ones that shape the design, as Meta's documents state them in October 2026.
Rule
What Meta says
What it means for the build
Customer service window
You may reply without a template within a 24-hour window, "which opens and resets with each user message"
The bot can talk freely while the customer is active. After 24 hours of silence it cannot send a free-form message
Templates
"Template messages are the only type of message that can be sent to WhatsApp users outside of a customer service window", and a template must be approved before use
Reminders, order updates and follow-ups are written and approved in advance. The model does not compose them on the spot
Opt-in
You may contact people only if they gave you their number and opted in to receive messages from you
Opt-in is captured at checkout, on forms or in the first chat, and stored
Human escalation
Automation is allowed, but you "must also have available prompt, clear, and direct escalation paths", such as human agent transfer in the chat, a phone number, email or a support form
Handoff is a requirement, not a feature to cut for budget
AI providers
Meta's terms bar AI providers from the platform where the AI is "the primary (rather than incidental or ancillary) functionality" being offered
A bot that serves your own customers about your own business is the intended use. A general-purpose assistant on WhatsApp is not, so the bot is kept to your topics
Both the terms and the messaging policy were last updated on 23 September 2026. Read the current versions before launch, because they change.
What Meta charges
As of October 2026 Meta's pricing documentation states:
Pricing has been per message since 1 July 2025. Guides that describe per-conversation fees describe the older model.
You are charged only when a template message is delivered. All non-template messages are free.
Utility templates delivered inside an open customer service window are free. Marketing templates are always charged.
Rates depend on the template's category and the country calling code of the recipient. Meta publishes rate cards in INR, and the current ones took effect on 1 July 2026. Its change log records a higher marketing rate for India from 1 January 2026.
When a customer arrives from a click-to-WhatsApp ad or a Page button and you reply within 24 hours, a free entry point window opens and stays open for 72 hours.
Message
Who starts
Meta's charge
Bot's free-form reply within 24 hours of the customer's last message
Customer
None
Utility template inside an open window, such as an order confirmation
Business
Utility template outside the window, such as a delivery update next day
Per message, utility rate
Marketing template, such as an offer or a win-back message
Per message, marketing rate, always
Authentication template, such as a one-time password
Per message, authentication rate
So for a support or sales conversation that the customer starts, Meta's fee is nil. The cost of that conversation is model usage and hosting. Meta's fees enter when you message first. This page prints no rupee rates because Meta revises them: read the INR rate card linked from Meta's pricing page. A WhatsApp tool that sits between you and Meta adds its own subscription or markup.
What it costs to run
Cost
Paid to
What drives it
Template messages
Meta
Number of business-initiated messages, by category
Model usage
OpenAI or Anthropic
Messages handled, multiplied by the tokens in each prompt and reply. Anthropic notes that token counts vary by language, so measure with real Hindi and Hinglish messages, not English samples
Speech-to-text, if voice notes are supported
The speech provider
Minutes of audio
Hosting, database, logs
Your cloud account
Traffic, and how long conversations are kept
Inbox or helpdesk for your agents
The tool you choose
Seats
Maintenance
Your development team
Policy changes at Meta, model versions being retired, new questions
What it costs to build
These are planning ranges from the cost model behind the AI cost calculator , not quotes. The model adds the chatbot base to the selected features, multiplies by a data factor, and gives a range from 15% below the subtotal to 25% above.
Scope
How the subtotal is reached
Planning range
Timeline band
AI chatbot on WhatsApp and web, answering from your documents, with handoff and CRM, in English and Hindi
(₹1,50,000 + ₹80,000 + ₹70,000 + ₹50,000) × 1.25 = ₹4,37,500
₹3,71,875 to ₹5,46,875
6–10 weeks
The same, connected to live systems and taking actions such as bookings and orders
(₹1,50,000 + ₹80,000 + ₹70,000 + ₹50,000 + ₹1,10,000) × 1.5 = ₹6,90,000
₹5,86,500 to ₹8,62,500
10–16 weeks
Voice is a ₹1,20,000 line in the same model, added before the multiplier. Assumptions: one number, one brand, a Meta business account that can be verified, and systems with usable APIs. The WhatsApp AI chatbot cost guide prints its own tiers, from simple menu bots upwards.
Hindi, Hinglish and voice notes
Customers in India write Hindi in English letters, switch language in the middle of a sentence, and shorten everything. A bot tested only on tidy English fails on the first real message. WavX builds the test set from the way your customers write, and the bot replies in the language and script the customer used. The guide to Hinglish and multilingual chatbots goes into the detail.
Voice notes arrive as audio files. Transcription makes mistakes with names, amounts and addresses, so anything the bot will act on is read back to the customer for confirmation.
What goes wrong
The window has closed. The bot tries to follow up the next day with a free-form message and the send fails. Follow-ups need an approved template.
A person and the bot both reply. The handoff did not silence the bot on that chat.
Stale answers on price or stock. The bot read a document when it should have read the live system.
A message type it cannot read. A photo of a damaged product, a location pin. The bot must say so and offer a person.
The same webhook arrives twice and the customer gets two replies. Each message ID is processed once.
Messages go to people who did not opt in. That breaks Meta's policy and puts the number at risk.
What it should not do without a human check
Agree to refunds, cancellations with a charge, or discounts that are not in the price list.
Give medical, legal or financial advice.
Decide a payment dispute or a complaint.
Send a marketing broadcast. The wording, the audience and the timing are approved by a person, and the template by Meta.
Act on a voice note or an image without confirming what it understood.
The design of these limits is covered in chatbot guardrails and human handoff.
How it is evaluated
WavX builds a test set from real chats, with personal details removed: the questions customers ask, in the languages they use, with the answer your team would accept, plus cases that must be refused or handed over. The bot is scored against the set before launch and after every change. A first release can run in suggest mode, where the bot drafts and your agent sends, until the logs show which topics it can handle alone.
When a ready-made WhatsApp tool is enough
As of October 2026, Wati, AiSensy and Interakt each sell WhatsApp API access by subscription with a shared team inbox, broadcasts, chatbot builders and AI agents. If your need is broadcasts, a shared inbox and simple reply flows, one of them is quicker and cheaper than a custom build. If you receive a handful of messages a day, a person with the WhatsApp Business app is the right answer.
A custom build is worth it when the bot must read and write your own systems, such as a custom ERP, a booking engine or your own app. It is also worth it when answers must be grounded in your documents with testing you control, when per-seat and per-message markups at your volume exceed the cost of owning the software, or when WhatsApp is one channel of a product you sell.
WhatsApp chatbots are one part of AI solutions and automation at WavX.
Frequently asked questions
Does Meta charge for every reply the chatbot sends?
No. As of October 2026 Meta's pricing documentation says you are charged only when a template message is delivered, and that non-template messages are free. A chatbot answering a customer who wrote to you within the last 24 hours is sending non-template messages. Meta's fees apply when you message first or after the window has closed, which requires a template.
Can this run on the WhatsApp Business app, or do I need the API?
A custom AI chatbot runs on the WhatsApp Business Platform, normally through Meta's Cloud API, because it needs a server that receives each message and sends the reply. The WhatsApp Business app on a phone is for a person answering by hand. The integration page for the WhatsApp Business API covers what the setup involves.
How much does a WhatsApp AI chatbot cost to build?
Our cost model gives a planning range of ₹3,71,875 to ₹5,46,875 for an AI chatbot on WhatsApp that answers from your documents, hands over to staff through a CRM and works in English and Hindi. One that also takes bookings or orders in live systems comes to ₹5,86,500 to ₹8,62,500. These are planning ranges from the AI cost calculator, not quotes.
Can it reply in Hindi and Hinglish?
Yes. It is built and tested on the way your customers write, including Hindi typed in English letters and messages that switch language halfway. It replies in the language and script the customer used. Voice notes need a speech-to-text step, which is an optional extra.
Can a customer still reach a person?
Yes, and Meta requires it. The WhatsApp Business Messaging Policy allows automation only if there are prompt, clear and direct escalation paths, such as transfer to a human agent in the chat, a phone number or an email address. We build in-chat handoff with the conversation history attached.
Sources
Meta for Developers: WhatsApp Business Platform pricing · read 2 October 2026
Meta for Developers: Template fundamentals · read 2 October 2026
WhatsApp Business Messaging Policy (last updated 23 September 2026) · read 2 October 2026
Meta Terms for WhatsApp Business Platform (last updated 23 September 2026) · read 2 October 2026
Meta for Developers: AI Provider pricing on WhatsApp Business Platform · read 2 October 2026
Anthropic: Claude API pricing (how tokens are counted) · read 2 October 2026
Wati: pricing and features · read 2 October 2026
AiSensy · read 2 October 2026
Interakt · read 2 October 2026
Related
WhatsApp Business API integration
WhatsApp AI chatbot development cost in India
AI chatbot development for website and app
AI development cost calculator
Build your own software — your way, your pricing.
WavX Solutions is here to create your own software in a fully custom way, built exactly how you work — with a pricing model that fits your business. Connect now and let's build it.
Contact Now helpwavx@gmail.com