MAKIN 
How much does an AI chatbot cost in India?
“Chatbot” covers three different things, and they cost differently. A rule-based bot follows menus and keywords. An AI assistant answers in its own words from your documents or data. An AI agent goes further and acts: it reads an email, updates a record, drafts a reply, and asks a person before doing anything risky.
Our packages for each, with the starting price, the first things each includes and the delivery time, read from the price list:
- WhatsApp & Support Automation — from ₹12,000
- Tally AI Assistant — from ₹22,000
- AI Support Copilot — from ₹45,000
- AI Agent Workflows — from ₹85,000
- Enterprise RAG Knowledge Assistant — from ₹1,75,000
What drives the cost of building one?
The model is rarely the expensive part of the build. The work is in connecting it to your data, limiting what it may do, and proving it behaves before customers see it.
- Where it lives. WhatsApp, a website chat, your helpdesk or Tally: each channel is its own integration.
- What it knows. A fixed set of answers, your resolved tickets, your documents, or live data from your systems. Answers grounded in your documents, with a citation on every answer, are the core of our Enterprise RAG Knowledge Assistant.
- What it may do. Answering is cheaper to build than acting. Agents that call tools and APIs need approval gates for risky actions and an audit log of every decision; both are in AI Agent Workflows.
- How you know it works. An evaluation set tested before go-live, in both our RAG and agent packages, and refusal behaviour when the answer is not in your documents, in the RAG assistant.
- Hand-off to people. Intent routing with a human hand-off is part of WhatsApp & Support Automation; the AI Support Copilot drafts replies for your agents rather than sending them itself.
- Language. The Tally AI Assistant takes questions in plain English or Hindi and cites every answer to vouchers and ledgers.
What does an AI chatbot cost to run?
AI model providers charge per token: small pieces of text, counted separately for what you send (the question, your instructions and any documents retrieved) and for what the model writes back. Prices differ by model, and larger models cost more per token.
As an example of published prices, Anthropic lists the following rates per million tokens for its Claude API 1. Google publishes per-million-token prices for its Gemini models on Google Cloud’s Gemini Enterprise Agent Platform too, with separate rates by model and by length of input 2. Prompt caching lowers the price of input that repeats on every request, such as your instructions.
| Model | Input | Output | Cached input (cache hits) |
|---|---|---|---|
| Claude Haiku 4.5 | $1 | $5 | $0.10 |
| Claude Sonnet 5.5 | $2 | $10 | $0.20 |
What does one conversation cost?
Take a support reply that sends 3,000 tokens to the model (instructions, the customer’s message and a few retrieved paragraphs) and gets 400 tokens back. At the Claude Haiku 4.5 prices above, that is 3,000 × $1 ÷ 1,000,000 plus 400 × $5 ÷ 1,000,000: $0.005 per reply. On Claude Sonnet 5.5 it is 3,000 × $2 ÷ 1,000,000 plus 400 × $10 ÷ 1,000,000: $0.010. Ten thousand such replies a month come to about $50 or $100 in model charges, before any caching.
Two things move that number most: how much text goes with every request, and which model answers. Retrieval that sends three relevant paragraphs instead of a whole manual, and a smaller model for routine questions, keep the bill small. Our agent package includes cost and token monitoring, and the RAG assistant includes usage and cost dashboards, so the running cost is visible every month.
On WhatsApp, Meta charges for each message on top of the AI cost. Free-form replies inside the 24-hour customer-service window are service messages, free up to 1,000 a month per business number and charged after that, as the table below shows. Our WhatsApp API pricing guide covers the rest.
| Category | Rate per message |
|---|---|
| Marketing | ₹0.8631 |
| Utility | ₹0.115 |
| Authentication | ₹0.115 |
| Service | ₹0.115 |
What do typical AI budgets look like?
These examples add up the starting prices on our cards; they are not bundle discounts. Model usage and WhatsApp messages are billed by those providers on top.
- WhatsApp support that captures leads into a CRM: WhatsApp & Support Automation, Sales CRM Starter — from ₹30,000 together
- Support copilot plus WhatsApp automation: AI Support Copilot, WhatsApp & Support Automation — from ₹57,000 together
- Agent that reads documents into your systems: AI Agent Workflows, Document & Data-Entry Automation — from ₹1,01,000 together
- Document assistant with a support copilot: Enterprise RAG Knowledge Assistant, AI Support Copilot — from ₹2,20,000 together
How do you compare AI chatbot quotes?
AI quotes are hard to compare because the expensive parts are invisible in a demo. These questions bring them out.
- What does it know, where does that data come from, and who keeps it current?
- What may it do on its own, and where does a person approve?
- How was it tested before launch, and what does it do when it does not know?
- Which model provider does it use, and who pays the token bill?
- Can you see usage and cost every month?
- Which providers will process your data, and are they named in the proposal?
How we worked out these figures
Every MakinForU price on this page is read from our live price list when the page is built, so it always matches the package cards. Example budgets add those starting prices together; they are not bundle discounts.
Model prices are copied from the providers’ own pricing pages on the date shown, and the worked example is plain arithmetic on them; your own cost depends on volumes, prompt length and the model chosen. Meta’s WhatsApp rates come from the rate file our WhatsApp calculator uses, with the rate card’s date and the date it was checked.
Sources
- Claude API pricing, Anthropic documentation, checked 7 Oct 2026
- Agent Platform pricing (generative AI), Google Cloud, checked 7 Oct 2026