WhatsApp and website chatbots that answer from your own documents
Your customers ask the same questions on WhatsApp and on your website. A bot can answer them from the documents you approve and can show the source it used. When it has no good answer, it is set up to say so and offer a person.
Bots we build
- WhatsApp bots on the official WhatsApp Business Cloud API, for enquiries, order questions and bookings.
- Website chat bots that sit on your site (websites) and pass enquiries to your team.
- Internal knowledge bots. We can build one for staff that answers from your SOPs, price sheets or HR policies.
Bots use the same rules for when to pass a chat to a person as our voice agents. Two sector examples: admissions enquiries (education) and order questions (online retail).
How answers stay tied to your documents
We load the documents you approve, such as FAQs, product sheets or a prospectus. The bot is set up to answer from them, and we can build it to show which document it used. We can test the bot against real questions your team writes, including questions it should refuse. When nothing in the documents fits, the bot is set up to say so and offer a person.
Handing a chat to your team
We can set up a shared inbox where staff reply in the same chat, with the bot’s conversation above. We built a shared WhatsApp inbox into Microsive ATS, where recruiters reply with the candidate’s profile beside the chat (Microsive ATS).
WhatsApp rules we design around
Meta allows free-form replies within 24 hours of the customer’s last message (WhatsApp Cloud API). After that window, your business can only start a conversation with an approved template. We plan reminders and follow-up messages around those templates before launch.
Which AI model it runs on
We can build on OpenAI or Anthropic models, or on an open model you run yourself through Ollama. The choice depends on your documents, your budget and where your data should stay. For larger document sets, we can add retrieval so the bot pulls the right passages first (RAG). Whichever model we pick, the chat records and your documents stay in an account you own.
Built by us
We built the support chatbot inside Microsive ATS. It finds its answers with PostgreSQL full-text search.
Sample bot
Interactive simulation · sample data
Sample Appliances is a made-up store. The bot answers from the two sample documents beside the chat. Pick a question to see what it does with it.
Choose what happens
Pick one above. Nothing here is live: no call is placed, no message is sent and nothing is saved.
FAQ
Can the bot take orders or payments?
We can add a Razorpay payment link (Razorpay) to a chat flow, so the customer pays on a Razorpay payment page. What the bot may confirm on its own is agreed in the scope.
Who owns the WhatsApp Business account?
You do. The number and the WhatsApp Business account sit on your own Meta account, and we build inside them. If that number is already running on the WhatsApp Business app, it has to move to the Cloud API, and it stops working in the app once it does.
Where are chats and documents stored?
In an account you own, in a region we choose with you. That covers the conversation records and the documents the bot answers from.
Should we build a bot or a voice agent?
If customers phone you, start with inbound calls. If they message, start with a bot. Our guide compares the two (voice agent or AI bot).
Which language will the agent speak?
The one your callers use. Language is a project decision rather than a fixed list: tell us who the agent will talk to, and we test a model in that language before we commit to it in the scope. That includes languages spoken outside India.
What do customers ask you over and over?
Paste a few real questions into the enquiry. Someone from our team replies within one business day.