Do you need a WhatsApp chatbot, and should it use AI?

A WhatsApp bot is worth it only when the same questions pile up, and AI earns its extra cost only once those questions come in free text or in more than one language.
The short answer
A WhatsApp chatbot is worth setting up when the same handful of questions arrive often enough that answering them by hand eats into real work, and it is not worth it when enquiries are few or every one is different. If a bot does make sense, a simple rule based one that follows set menus answers most of what a small shop or clinic gets asked, and an AI bot earns its extra running cost only once questions come in free text, in more than one language, or faster than a person can keep up. This guide walks you through that decision for an Indian small business, with the real running costs and with Tamil, Hindi and English in mind.
What can a WhatsApp chatbot actually do for a small business?
A WhatsApp bot answers common questions on its own: opening hours, location, what you sell, a price list, a booking, the status of an order, and simple lead capture before a person takes over. There are two kinds. A rule based bot follows a fixed menu or set keywords, so the customer taps a button or picks from a list and gets a prepared reply. An AI bot reads what the customer actually types, including typos and loose wording, and answers from a knowledge base you give it.
One thing is true of both, and it decides where a bot can run at all. To send automated replies to customers at any scale, a bot sits on the WhatsApp Business Platform Cloud API, the official route for programmatic messaging, rather than the WhatsApp app on a phone (Meta's Cloud API overview). A customer also has to have messaged you or opted in first, so a bot is for answering and following up, not for cold outreach.
When is a simple rule based bot enough?
A rule based bot is enough when your questions are predictable. A clinic that is asked for its timings, its address and how to book an appointment, or a shop asked for its catalogue and whether an item is in stock, can answer almost everything with a tidy menu of buttons. The replies are fixed, so the bot never invents an answer, it is cheap to keep running, and it is quick to set up. For a lot of small businesses that is the whole job done.
It runs into its limits in two places. The first is when a customer ignores the menu and types a real sentence in their own words, because a keyword bot either matches the wrong rule or gives up. The second is language. A customer who writes in Tamil, or in Tamil typed using English letters, will not line up with English keywords, and the bot stalls. If most of your enquiries are predictable, a rule based menu is the sensible first step and often the last one you need.
When does an AI bot earn its running cost?
An AI bot is worth its extra cost once the questions stop being predictable. It reads free text, copes with synonyms and spelling, holds the thread of a short conversation, and answers from your own information rather than a fixed menu. That matters when you get a steady stream of open ended questions, when customers write in mixed languages, or when enquiries arrive faster than your team can reply.
The trade off is honest and worth stating. An AI bot can give a confident answer that is simply wrong, so it needs guardrails: a clear boundary on what it may answer, a tested knowledge base, and an easy way to hand off to a person when it is unsure. It also carries a small cost for every question it processes, on top of the platform it runs on. An AI bot that is set up carelessly is worse than a good rule based one, so the question is never only whether you can afford AI, it is whether your enquiries actually need it.
What does it really cost to run in India?
Two bills hide behind every WhatsApp bot, and the second is the one that surprises people. The first is the platform or builder you use, usually a monthly subscription. The second is Meta's own message charge, which is separate and which a monthly platform price does not include.
Meta charges for WhatsApp messages on a per message basis, a model it moved to on 1 July 2025, sorted into categories for marketing, utility, authentication and service messages, and every business phone number gets a free tier of 1,000 service messages a month before service messages are charged (Meta's published WhatsApp pricing). Timing matters too. When a customer messages you, a 24 hour customer service window opens, and inside it you can reply in free text, but once it closes you can only send a pre approved template message, which is charged (WhatsApp's conversation rules). An AI bot adds one more line to the bill, the model's own small charge for each question it reads and answers. The useful way to compare options is to ask for all three parts, the platform fee, the expected Meta message charges for your volume, and the AI cost per question, rather than just the headline subscription.
| Your enquiry pattern | The likely right choice | What running it looks like |
|---|---|---|
| A few enquiries a week, mostly different each time | Probably no bot yet | A shared inbox and a couple of saved quick replies do the job, and a monthly platform fee would mostly sit idle |
| Steady enquiries every day, mostly the same set of questions like hours, prices, booking and location | A rule based menu bot | A platform subscription plus Meta's per message charge, with fixed answers that never get invented |
| High volume, open ended questions, often in Tamil, Hindi or mixed text | An AI bot | The platform subscription and Meta's message charge, plus the model's own per question cost, and someone to keep its answers accurate |
How does it handle Tamil, Hindi and English?
This is where the two kinds of bot part company. An AI bot reads Tamil, Hindi and English comfortably, and it also copes with the way people really type in India, switching between languages in one message and writing Tamil or Hindi words in English letters. A rule based keyword bot does not, because it is matching set words, and a message in another script or in transliterated text rarely matches.
There is a practical way around this even without AI. A rule based bot that leads with buttons and menus sidesteps the language problem entirely, because the customer is choosing an option rather than typing a sentence, and you can label those buttons in whichever languages your customers use. So if your customers write freely in Tamil or Tanglish, that points towards either a button led rule based bot or a proper AI bot, and away from a keyword bot that only listens for English words.
How do you start small and grow it?
The low risk path is to start narrow and let demand decide the rest. Begin with a rule based menu that answers your top few questions and captures a lead, watch for a month which questions people ask that the menu cannot handle, and add an AI layer only where free text questions are genuinely piling up. That way you pay for AI when your enquiries have earned it, not before.
The same thinking applies to everything around the bot, from how it connects to your stock or booking system to who checks its answers. We take the same plain approach to building reliable customer facing systems: decide the job first, keep it simple, and add complexity only where it pays for itself.
Frequently asked questions
Do I need the official WhatsApp Business API for a bot?
Yes, for an automated bot that messages customers at any scale. The bot runs on the WhatsApp Business Platform Cloud API, which is the official programmatic route (Meta's Cloud API overview). The ordinary WhatsApp or WhatsApp Business app on a phone is for replying by hand, not for running a bot.
Will a bot annoy my customers?
It will if it traps them in a loop or cannot pass them to a person. A good bot answers the easy questions instantly, says plainly when it does not know, and offers a human handover at any point. Used that way, it tends to make customers happier, because they get an answer at once instead of waiting.
Can a bot take payments or bookings?
Yes. A rule based bot can take a booking for fixed slots or share a payment link through a menu, and an AI bot can do the same while also understanding a free text request. What a payment needs is a secure link or an integration with your booking or payment system, not AI by itself.
How accurate is an AI bot, really?
It depends almost entirely on how it is set up, not on the model behind it. An AI bot answering from a clean, specific knowledge base with clear limits is dependable, while one pointed at the open internet with no boundaries will sometimes answer confidently and wrongly. The safeguard is to scope it tightly and let it say when it needs to connect you to a person.
Ask Prognesoft whether a bot is worth it for your enquiry volume
Tell us roughly how many enquiries you get, what people usually ask, and which languages they write in, and we will tell you plainly whether a bot is worth it, whether a simple rule based one is enough, or whether an AI bot would pay for itself. See how we approach AI chatbots and automation, then ask us about a WhatsApp bot for your business.
- A WhatsApp bot is worth setting up only when the same questions arrive often; for a few mixed enquiries a week, a shared inbox and a couple of saved replies do the job
- A rule based menu bot answers predictable questions like hours, prices and bookings cheaply and never invents an answer, which suits most small shops and clinics
- An AI bot earns its extra running cost once questions come in free text, in Tamil, Hindi or mixed script, or faster than a person can reply, and it needs guardrails and a human handover
- Budget for two separate costs, the platform subscription and Meta's own per message charge, plus a per question cost for AI; Meta gives every business number a free tier of service messages each month (https://developers.facebook.com/docs/whatsapp/pricing/)