How much does an AI chatbot cost in India in 2026?

In 2026, an AI chatbot for an Indian business typically costs ₹30,000–75,000 for a rule-based FAQ bot, ₹1–4 lakh for a business-grade assistant wired into your website, WhatsApp, and CRM, and ₹5–25 lakh for an AI-grade build that answers from your own documents and hands off to a human cleanly. Running it adds ₹5,000–50,000 a month — the number most quotes omit. Which tier you're buying decides the bill; the technology mostly doesn't.

Those are typical Indian market ranges, not quotes. The chatbot is the most quoted and least specified project our AI and automation practice gets asked to price: the same word covers a ₹40,000 menu bot and a ₹20 lakh support system, and many proposals never say which one they contain. So before anyone argues about a number, pin down the tier.

A habit that keeps the whole conversation honest: a chatbot is priced by three words — knowledge, channels, actions. What must it know, where must it live, and what may it do on its own? Every figure below is one of those three wearing a price tag.

What are the three tiers of chatbot, really?

Treat these as planning bands for work delivered by an established Indian firm at production standard. Freelance builds can come in lower; enterprise deployments with compliance reviews go higher.

• Rule-based FAQ bot — scripted flows and keyword matching, 20–50 fixed answers, one channel: ₹30,000–75,000, live in one to two weeks. Honest and cheap, but it answers only what it was told, in the words it was told.
• Business chatbot — website widget plus WhatsApp, lead capture into your CRM, forms and order-status lookups, basic language understanding: ₹1–4 lakh, three to six weeks. The tier most small businesses actually need first.
• AI-grade assistant — a language model grounded in your own documents and catalogue (RAG), so it answers questions nobody scripted, in the customer's words and language, with sources cited, guardrails, and clean human handoff: ₹5–25 lakh, six to twelve weeks. Deep ERP integration or a regulated industry pushes past this band.

For international buyers, the same bands translate to roughly $400–900, $1,200–5,000, and $6,000–30,000 at 2026 rates. The tell for which tier you need is simple: if the right answer changes when your data changes — prices, stock, policies — you're shopping in the third tier, whatever the brochure calls it.

What actually moves the price?

Four drivers explain most of the spread inside every band. The model provider's fee is rarely one of them.

• Knowledge — a bot that must know 20 FAQs is a different machine from one that must know 500 products with variants and weekly price changes. The state of that knowledge matters as much as its size: clean web pages are cheap to ground on; a decade of scanned PDFs needs preparation that is real engineering time.
• Channels — the website widget is the easy case. WhatsApp Business API adds an approval process, template rules, and Meta's per-message fees; Instagram and voice each bring their own review cycles and edge cases. Every added channel multiplies testing, not just reach.
• Actions — answering is cheap; doing is not. The moment the bot checks an order in your ERP, books a slot, or raises a ticket, every connected system adds discovery, error handling, and testing — the same integration economics that governs all software builds, and usually the biggest line in a tier-three quote.
• Handoff and guardrails — deciding what the bot must never answer, when it must summon a person, and what that person sees when it does. This is design work, not configuration, and it is what separates a chatbot customers tolerate from one they thank.

What does a chatbot cost to run each month?

Budget against four meters. Model usage — the API cost of the conversations themselves — lands between ₹2,000 and ₹15,000 a month for most small-business assistants, rising with traffic and with how much retrieval each answer needs. Channel fees come second: Meta bills WhatsApp per message, small individually and meaningful at volume. Hosting and monitoring for a modest deployment run ₹1,500–8,000. The fourth meter is the one nobody budgets: upkeep. Prices and policies change, and someone must review transcripts monthly to catch what the bot fumbled — a bot nobody reviews decays into a complaint generator.

All told, most small and mid-sized deployments land between ₹5,000 and ₹50,000 a month. We state the run cost next to the build cost in every estimate, for the same reason our software cost guide does it for servers and support: we run our own AI in production every day — DigiSign's Content Studio — so the monthly meters are not hypothetical to us.

Should you build a chatbot or subscribe to one?

If your needs are generic, don't build. Chatbot SaaS platforms at ₹2,000–10,000 a month handle FAQ flows, lead capture, and standard WhatsApp journeys well, and you can be live this week. The subscription is right while your questions match its templates — and you'll know you've outgrown it when every request to their support team starts with 'is it possible to…'.

Building wins in three situations: when the bot must know your data deeply (catalogues, policies, contracts — the RAG tier), when it must act inside your systems, or when volume turns per-conversation pricing into a bigger bill than owning the bot. There's a quieter fourth: ownership. A built bot's prompts, integrations, and conversation history are assets you keep — the same IP terms we tell every software buyer to demand apply here, because switching costs are exactly where bad chatbot vendors hide.

What results should you actually expect?

Calibrate against checkable patterns, not vendor decks. An assistant grounded in your documents with human handoff typically deflects 40–60% of routine tickets, and speeds up the rest because the person inherits the conversation and its sources instead of starting cold. What you should not expect is 90% automation in month one; anyone promising it is pricing a demo, not a system. The honest curve is strong on the repetitive majority immediately, then better every month as transcript reviews feed the knowledge base. Four numbers tell you the truth throughout: deflection rate, handoff quality, how often customers rephrase, and how often your staff overrule the bot.

How do you pilot one without betting the budget?

Start where the evidence already sits: your inbox and call log. The twenty questions your team answers most often are the pilot scope — not 'everything a customer might ask'. A two-to-four-week pilot puts a grounded assistant on one channel, with human handoff from day one, measured on live traffic against the four numbers above. In the Indian market, pilots of this shape commonly land in the low lakhs — a line item, not a bet — and they end with the only estimate that means anything: real deflection on your real questions, priced with its run cost.

And if a ₹3,000-a-month subscription covers your twenty questions, the honest advice is to take it and revisit in a year. The expensive mistake in chatbots is rarely choosing the wrong vendor — it's buying tier three to solve a tier one problem.