The first automation most Indian businesses buy is a chatbot for their website. It is also, almost always, the automation installed on the surface their customers visit least. Your buyers are on WhatsApp, on Google Maps, on Instagram, on the phone. Your website is where they go once, to confirm you exist, before messaging you somewhere else. The bot sits in the corner of that page like a receptionist in a building nobody enters.
I am not guessing at this. I have built these systems, and I have opened the dashboards afterwards. The pattern repeats: a business pays for a chatbot subscription for a year, renews it without looking, and at no point does anyone ask the only question that matters. How many real conversations did this thing actually have?
You can answer that question in under an hour, before your next renewal. This post shows you how.
Why the website got the bot first
The website gets the bot first because that is what vendors sell first. A website widget is the easiest thing for a chatbot company to install. There is no WhatsApp Business API approval, no template registration, no Meta verification. The vendor drops one script tag on your site, shows you the little chat bubble, and the demo looks impressive because in the demo, someone is talking to it.
So the decision gets made backwards. Nobody asked "where do our enquiries actually arrive?" The question that got asked was "where is it easiest to put a bot?" Those are different questions with different answers, and the gap between them is where the subscription money goes.
Think about how your own customers reach you. If you run a distribution business, your retailers WhatsApp their orders. If you run a clinic, patients call or walk in from a Google Maps listing. If you run a services firm, referrals come by phone and email. In every one of these, the website is a brochure. People check the address, maybe the phone number, and leave. A brochure does not need a receptionist.
A worked example: the distributor
Take a business every Indian owner recognises. An FMCG distributor covering a few districts. Retailers place orders the same way they have for years: a WhatsApp message to the salesman, often a photo of a handwritten list, sometimes a voice note. Payment reminders go out on WhatsApp. Scheme details go out on WhatsApp. During GST filing week, the accountant is chasing missing invoices, on WhatsApp.
This distributor gets pitched a chatbot. The vendor puts it on the company website, the one that was built years ago and last updated when the address changed. The bot can answer questions about products, working hours, and coverage area.
Now walk through who visits that website. Retailers do not. They have the salesman's number saved. New retailers hear about the distributor through the company's sales officer or through other shopkeepers, and their first touchpoint is a phone call. The people who do land on the website are job applicants, the occasional brand manager doing due diligence, and the owner's nephew checking whether the site is still up.
The bot is fluent, polite, available at three in the morning, and talking to nobody. Meanwhile the actual queue of customer conversations, the WhatsApp inbox where orders arrive and sit unanswered between two and five in the afternoon because the salesman is on the road, has no automation at all. The money went to the empty room.
How to check the real number before you renew
Do not ask the vendor whether the bot is working. Ask the dashboard. Every chatbot platform keeps a conversation log, because that log is how they justify the renewal. Here is the check.
Step one: find the conversations report. Log into the chatbot dashboard. Look for a section called Conversations, Chats, Sessions, or Inbox. Set the date range to the last three months, not the last week. A short window hides the emptiness; one busy day can make a dead month look alive.
Step two: separate real conversations from noise. The headline number is inflated by design. Most platforms count a "conversation" the moment the widget loads or the bot fires its automatic greeting, even if the visitor never types a word. Filter for conversations where a human actually sent at least one message. Then subtract your own noise: your staff testing it, the vendor's team testing it, you checking it after reading this post. What is left is the true count.
Step three: read what people actually asked. Open the transcripts of the real conversations. This is the most honest ten minutes you will spend on your marketing stack. You will usually find three kinds of messages: people asking for the phone number so they can leave the website and call you, people asking something the bot could not answer, and job seekers. Count how many transcripts end with the bot successfully resolving something a human would otherwise have handled. That number is the bot's entire contribution.
Step four: compare it to one day of WhatsApp. Open WhatsApp Business on the phone where enquiries land. Count yesterday's incoming customer conversations. Just yesterday. Now put that single day next to three months of website bot transcripts. For most businesses I have looked at, this comparison is not close, and it settles the renewal decision on its own.
Step five: check who even visits the site. If you have Google Analytics or Search Console connected, look at monthly visitors and which pages they see. If most sessions are a few seconds on the homepage, you have confirmed the surface itself is quiet. No bot, however clever, generates conversations on a page nobody opens.
Sometimes the honest answer is: don't automate this
Here is the part the industry will not tell you, because everyone in it is paid when you automate more, not less.
If your website gets a trickle of visitors, the correct amount of website automation is none. Not a better bot, not an AI upgrade to the existing bot, not a redesign with the bot moved to the centre of the page. None. Put a clear phone number and a WhatsApp click-to-chat link on the site and let the website do its actual job, which is being a brochure. A link that opens WhatsApp costs nothing, needs no renewal, and delivers the visitor to the channel where you already answer people.
The same honesty applies on WhatsApp itself. Even where the volume genuinely is, not everything should be handed to a bot. Order confirmations, catalogue requests, working hours, payment reminders: automate those, they are repetitive and low-stakes. But a retailer angry about a short supply, or a new customer with a big first order, should reach a human quickly. A bot that stalls a hot enquiry with menu options does more damage than no bot at all. Automation should absorb the repetitive middle of your inbox, not fence off the conversations that make you money.
If a vendor never tells you "this channel does not need automation", you are not talking to an adviser. You are talking to a renewal.
What to do this week
One action. Log into your chatbot dashboard, set the range to the last three months, and count the conversations where a real customer typed a real message and the bot resolved it without a human stepping in. Write that number down next to what the renewal costs you.
If the number is healthy, keep the bot and ignore me. If the number is close to zero, cancel the renewal, add a WhatsApp click-to-chat link to your website in its place, and spend the money where your customers already are. Either way, you will have made the decision from your own data instead of the vendor's demo, and that habit is worth more than any single subscription.
I write about automation and the systems I actually run at architmittal.com. Originally published there.
