On this page
A WhatsApp chatbot in plain words
A WhatsApp chatbot is software connected to your official WhatsApp number that reads incoming messages and replies automatically. The good ones don’t pretend to be human and don’t try to handle everything. They guide the customer with reply buttons and list menus, answer what can be answered from reliable information, collect the details your team would ask for anyway, and hand the conversation to a person when it needs judgement.
Chatbots can be rule-based (fixed menus and flows), AI-assisted (a language model answers open questions from your own content), or a mix. For most businesses the mix works best: structured flows for the common paths, and an AI layer only where customers type free-form questions.
Where a chatbot earns its place
A chatbot is worth building when the same few questions make up most of your incoming chats. Some examples:
- Restaurant — menu, opening hours, table booking and order status without a staff member picking up the phone during service.
- Real estate agency — project details, brochures, site-visit booking and a quick qualification on budget and location.
- Clinic — doctor availability, appointment booking and pre-visit instructions.
- Coaching institute — course list, batch timings, fee structure and admission forms.
- Manufacturer — dealer enquiries routed by region, catalogue downloads and order status for distributors.
If your chats are mostly unique, complex negotiations, a chatbot will add friction rather than remove it. We’ll say so if that’s the case.
How we design the conversation
We start from real conversations, not a blank flowchart. With your permission we read a sample of recent chats and group them by what the customer actually wanted. That tells us which paths deserve a flow, which questions need a person, and where customers currently wait longest.
Principles we follow
- Buttons over typing wherever the answer is one of a few options.
- Short messages — one idea per message, not a wall of text.
- An exit at every step — the customer can always reach a person.
- No dead ends — unrecognised messages get a helpful reply and a menu, not silence.
- Collect once — details the bot gathers are passed to the agent, so the customer never repeats them.
Human handover and the 24-hour window
When a customer messages you, WhatsApp opens a 24-hour customer service window in which you can reply freely. The chatbot works inside that window, and the handover to a person is timed so an agent can respond before it closes. If a reply is needed later — a quote that takes two days, for example — it must go out as an approved template.
Handovers go to a shared inbox or your CRM with the full conversation attached. Agents see what the bot asked, what the customer answered and why the bot stepped aside.
Tools and technology
Bots run on the official WhatsApp Cloud API with webhooks into a backend we build, typically in Node.js or Python, with PostgreSQL for conversation state and Redis for queues. Where forms suit better than chat, we use WhatsApp Flows for structured screens inside WhatsApp. AI answers use models such as OpenAI’s GPT or Anthropic’s Claude, grounded in your own documents — see AI Chatbot Development for how that retrieval works.
Before launch we run every flow against a test number, including the awkward cases: a customer who types instead of tapping a button, sends a voice note, switches language halfway or goes quiet and comes back the next day. Each conversation is logged with its outcome, so after launch you can see which paths finish cleanly and which ones people abandon.
What affects timeline and cost
- Number of flows and how many branches each has.
- Whether answers come from fixed content or need an AI layer with its own testing.
- Integrations — booking system, store, CRM or database lookups for live status.
- Languages — each language needs its own content and testing.
- Whether a new API setup is needed first (WhatsApp Business API Integration).
Mistakes that make chatbots unpopular
- Hiding the option to talk to a person.
- Long numbered text menus where buttons would do.
- Letting an AI model answer prices or policies it wasn’t given.
- Launching and never looking at where customers drop off.
We review conversation logs after launch and adjust flows as patterns emerge. Browse the rest of our WhatsApp automation services, read the WhatsApp automation guide, or start a conversation with us.
Frequently asked questions
Can the chatbot answer in Hindi, Gujarati or other languages?
Yes. Fixed flows are written in each language you need, and AI-assisted answers can reply in the customer’s language. Each language is tested separately before launch.
Will customers know they are talking to a bot?
They should. We make it clear the first replies are automated and that a person is available, which sets the right expectations and avoids frustration.
Can the chatbot send messages to people who haven’t contacted us?
Only to people who have opted in, and only through approved templates when outside the 24-hour customer service window. A chatbot is mainly for replying, not cold outreach.
Can we change the flows ourselves later?
Text and simple options can be made editable from an admin screen. Structural changes, new integrations or new branches are usually a small development task.
How is a WhatsApp chatbot different from a chatbot on our website?
The logic can be similar, but WhatsApp adds its own rules — the 24-hour window, template approval, opt-in — and its own interface elements like reply buttons and list menus. Conversations also continue over days, so the bot has to remember where each customer left off.
Talk to us about whatsapp chatbots
Automated conversations that qualify, answer and route customers on the app they already use.