Build and Sell WhatsApp Chatbots to Small Businesses
यह पोस्ट हिंदी में भी उपलब्ध हैहिंदी में पढ़ें →

Build and Sell WhatsApp Chatbots to Small Businesses

Ram Ashare··7 min read

Short answer: A WhatsApp chatbot is an automated reply flow that answers a business's most repeated questions: hours, price, location, booking. No-code builders handle most of it. Building them for small businesses works as a freelance service with a setup fee plus a small monthly maintenance retainer.


Nine days after I switched on my first bot, the receptionist at the dental clinic called me and said, "Please turn this off, it has made my job harder, not easier."

I had thought I built something impressive. It worked. It replied. Messages went out within a second of arriving. But the replies were vague enough that patients read them, got confused, and then called the clinic anyway. So I had inserted a step between the patient and the answer, and solved nothing.

That project paid Rs 4,500. I spent the next three weeks fixing it for free.

What clients actually want, which is less than you think

Say the word chatbot and people picture AI. Most small businesses do not want AI and would be worse off with it.

What they want is this: somebody types "what are your timings" on WhatsApp and gets a correct answer immediately. That is it. A menu flow. Press 1 for hours, 2 for address, 3 to book. This is a decision tree and the idea is older than most of the people selling it.

Of my seven clients, two have any AI in their bot, and only because their incoming questions were genuinely unpredictable.

Honestly, the technical part is the easy half. The hard half is sitting with a business owner and working out which questions actually arrive. They cannot tell you. They will say "people ask everything." Then you borrow their phone, scroll through the last two hundred incoming messages, and find that a hundred and forty of them ask three things.

Those three things are the bot.

The eleven-option disaster

I skipped that scrolling step with the clinic. I assumed I already knew what a dental patient would ask.

So I built a menu with eleven options. Eleven. On a phone screen a WhatsApp message with eleven numbered lines fills the whole display, and the reader scrolls once and decides calling would be faster.

Second mistake: every reply I wrote was a paragraph. "Our clinic operates from 10 in the morning and our doctors are available…" Nobody wanted that. They wanted "10am to 7pm, closed Sunday."

Third mistake, and this is the one that caused the phone call: I put a booking option in the flow, and the booking went nowhere. The bot said "your request has been received." Nobody received it. The receptionist had no way of knowing a request existed. So patients turned up saying they had booked on WhatsApp, and there was no record.

That is why her job got harder. I had generated ghosts for her to deal with.

The fix took three weeks. Eleven options became four. Replies became one line each. And I deleted the booking option entirely, replacing it with "to book, call this number." That sounds like a downgrade and it was the change that made her happy.

Sometimes removing the feature is the fix.

One real earning story like this every week, on WhatsApp.

Join Free

Picking a tool without recommending one

I have abandoned three tools and settled on one. I am not naming it, because pricing in this space changes every few months and I would rather nobody sign up on my say-so.

What matters is whether the tool runs on the official WhatsApp Business API. Some cheap ones work by hooking into a normal WhatsApp account on a phone and automating it. That breaks Meta's rules, and the client's number can get banned. If a client's business number goes dark because of something you set up, that is the end of your referrals, permanently.

Second thing to check: does it give you an agency view. One login, all your clients' bots. Without that you are juggling separate logins per client, and by the third one it becomes genuinely annoying in a way that makes you avoid doing maintenance.

Third: whose card is on file. Get the client's on there from day one.

Getting the first few clients

Client one was my neighbour's clinic and barely counted as paid work. Client two was his referral.

Client three came differently, and this is the method that has held up. I was scrolling a local gym's Instagram and noticed that under almost every post, people had commented "fees?" and nobody had replied. Dozens of them, going back months.

I sent a DM: "You have over 40 comments asking about fees with no reply. I can set up automatic answers on WhatsApp. Fifteen minutes for a demo if you want."

He watched the demo. He said yes. Rs 9,000.

That works because you are not sending a pitch, you are handing someone a screenshot of their own leak.

But to be fair about the numbers, the hit rate is not good. I have approached roughly 60 businesses this way. Seven became clients. Most simply never replied, and a handful said the thing everyone says, which is that the budget will open up next quarter.

Charging monthly, which I resisted for no reason

For the first three clients I charged once and walked away. Build, invoice, done.

Then three months later the calls started. Timings changed, update it. New offer, add it. Each one was a ten-minute job so I did it free, and ten-minute jobs across several clients became a chore I quietly resented.

From the fourth client I split it: a setup fee, then Rs 1,200 a month. Maintenance means I check the flow once a month, make small edits, and fix anything that breaks.

Clients accept it without blinking. It costs me around 25 minutes per client per month. Five are on it now.

And I had this backwards going in. I assumed selling a retainer would be the hard part of the conversation. It turned out to be the part that reassured them, because what small business owners are actually afraid of is that the system will quietly stop working and nobody will notice for a month.

Where the hours really go

Building the bot takes me two to three hours now. Everything around it takes longer.

Reading their old messages: an hour. Finalising the answers with the owner: usually two meetings, because in the first one they say everything looks perfect and then send fourteen changes on WhatsApp three days later. Testing: an hour, where I message my own bot in every wrong way I can think of.

Do not skip the testing. People type "hlo" and "helo" and "HELLO" and "hi bro" and sometimes just send a voice note, which a bot can do nothing with. I keep one line for that case: "Voice notes are read by our staff, someone will reply shortly." Small thing. It is the kind of small thing that makes an owner believe you thought about their customers.

One more rule I never break now. Every bot gets an option that leads to a human. People who dislike bots need an exit, and owners relax when they can see that nobody gets trapped.

Two things I still have not worked out

Whether this scales. Every client needs so much conversation that I suspect I hit a wall around twelve or thirteen. A template for the clinic category should fix that, since their questions overlap almost completely, and I have been half-building one for months without finishing it.

And the AI question. The two bots that use it work well, until they don't. One of them once told a patient about a treatment that clinic does not offer. I told the client immediately and he laughed, but my stomach dropped. Now the AI bots are locked to answering only from documents the client gave me, and they refuse rather than improvise.

That dental clinic bot is still running, about a year and a half later. The same receptionist messages me herself when the timings change. Last month she mentioned that a patient asked who replies so fast, and she told them, "that's our system."

Our system. I thought about that for a while.

Frequently Asked Questions

Do you need to code to build WhatsApp chatbots?

Not for the common cases. My first three bots were built entirely in a no-code flow builder with drag-and-drop steps. I needed a developer exactly once, when a client wanted the bot to write into their existing booking software, and I paid a freelancer Rs 4,000 for that piece.

What do people actually charge for this in India?

I charged Rs 4,500 for my first one, which was too low for the work it turned into. My third was Rs 11,500. These days a straightforward FAQ bot is somewhere between Rs 8,000 and Rs 14,000 for me, plus Rs 1,200 a month for maintenance.

Who pays for the WhatsApp API charges?

The client should, from their own card. Meta charges per conversation and the rate differs by conversation type. One clinic client of mine runs about Rs 700 to Rs 900 a month. I made the mistake of paying from my card for two clients early on and forgot to invoice it back, which cost me Rs 1,600.

Which businesses say yes fastest?

Anyone drowning in repeat questions: clinics, coaching centres, salons, gyms, small showrooms. Four of my seven clients are clinics. Restaurants have said no the most, because they assume the ordering conversation already happens inside a delivery app.

How long before you can sell your first one?

Learning the tool took me under a week. Learning to figure out what a business actually needs is still in progress. My first bot was bad enough that the client's receptionist called nine days in and asked me to switch it off.

👤

Ram Ashare

Founder, Simple Kamai

Testing online earning methods in India since 2023 — freelancing, digital products, affiliate marketing, and more. Only writing about what has actually worked.

Learn more →

Free: The First Earning Checklist (7-Day Action Plan)

Subscribe and get instant access to the 7-day checklist that takes you from "I'll start soon" to your first proposal sent. Plus one tested earning tip every week — no fluff.

Join WhatsApp Channel

Get weekly earning tips

Join Free →

Share this. It might help someone.