Chatbot Price in India 2026: ₹1,500/Month or ₹35K-₹8L Build
Founder & Lead Developer, Codingclave · 200+ projects since 2017
Chatbot price in India in 2026: the short answer
A chatbot in India costs ₹1,500 to ₹16,999 per month on a subscription platform, or ₹35,000 to ₹8,00,000 one-time for a custom build, plus ₹600 to ₹25,000 a month to host and run it. Those are the three numbers every buyer needs, so here they are before anything else.
The spread is wide because "chatbot" covers a ₹1,500-a-month AiSensy flow bot that sends menu buttons on WhatsApp, and a ₹14 lakh HR assistant that reads 11 years of policy documents for a bank. I have quoted both. At Codingclave we have built over 50 chatbots for Indian businesses since 2024, and this page is the price list I wish clients had before their first call: subscription plans with current vendor prices, one-time build costs by chatbot type, hosting cost per month, LLM API fees, and the honest point where monthly SaaS stops being cheaper than building your own.
Every vendor price below was checked against the vendor's own pricing page or a dated price tracker on 4 October 2026. Where a price depends on exchange rate I have used ₹85 to the dollar.
Chatbot price table by chatbot type (one-time build)
| Chatbot type | One-time build cost | Monthly running cost | Timeline | Best for |
|---|---|---|---|---|
| Rule-based FAQ bot | ₹35,000 to ₹1,25,000 | ₹600 to ₹2,000 | 2 to 3 weeks | Clinics, local services, small stores |
| Intent-based bot (Dialogflow or Rasa) | ₹1,25,000 to ₹3,00,000 | ₹2,000 to ₹8,000 | 4 to 6 weeks | Booking flows, support with 500+ queries a month |
| AI chatbot (GPT or Claude, answers from your documents) | ₹2,50,000 to ₹6,00,000 | ₹8,000 to ₹40,000 | 6 to 10 weeks | Sales, admissions, complex support |
| WhatsApp AI chatbot | ₹2,00,000 to ₹5,50,000 | ₹5,000 to ₹25,000 plus Meta fees | 5 to 8 weeks | D2C, coaching, real estate |
| Enterprise AI assistant | ₹5,50,000 to ₹15,00,000+ | ₹40,000 to ₹3,00,000 | 10 to 20 weeks | Banks, insurers, large e-commerce |
The build cost is a fixed-price figure for a working bot in production. The monthly column is hosting plus LLM API fees plus a maintenance retainer, and each of those is broken out further down. WhatsApp bots also pay Meta's per-message fees, about ₹0.86 per marketing message and ₹0.115 per utility message in India as of 2026, whichever platform you use; from 1 October 2026 Meta charges bot replies inside the 24-hour window at the utility rate too, once a business number has used its 1,000 free service messages for the month.
Chatbot subscription plans in India: monthly SaaS prices vs a one-time build
This is the question behind "chatbot cost per month in India", so here are the plans as published in October 2026. All Indian plans are plus 18 percent GST.
| Platform | Entry plan | Higher plan | Chatbot included? | Notes |
|---|---|---|---|---|
| PayPerWA (ours) | ₹0 subscription, ₹0.20 per message platform fee | Same | Yes, drag-and-drop builder | Meta Cloud API, failed messages refunded to wallet |
| AiSensy | ₹1,500 a month (Basic) | ₹3,200 a month (Pro) | Add-on, ₹2,500 a month for the first 5 flows | 10 percent off on annual billing (AiSensy pricing audit) |
| Interakt | ₹3,499 a quarter (Starter), about ₹1,166 a month | ₹7,699 a quarter (Growth), about ₹2,566 a month | Yes, flow builder | Marketing messages carry a markup over Meta's rate (Interakt pricing audit) |
| WATI | ₹2,499 a month (Growth), 1,000 chatbot sessions | ₹5,999 (Pro, 2,000 sessions) and ₹16,999 (Business, 5,000 sessions) | Yes, with session caps | Roughly 20 percent markup on Meta fees (WATI pricing audit) |
| Tidio Lyro (website AI agent) | $39 a month, about ₹3,300, for 50 AI conversations | Usage-priced, about $0.70 per AI conversation above the bundle | Yes, LLM answers from your help centre | Best for Shopify and small SaaS |
| Intercom Fin (AI agent) | $0.99 per resolved conversation | Copilot add-on $35 per seat a month | Yes, LLM answers | Needs an Intercom seat plan underneath |
Three things to read off that table.
First, a WhatsApp flow bot is cheap. ₹1,500 to ₹6,000 a month gets you menus, buttons, booking and order-status flows on WATI, AiSensy or Interakt, and PayPerWA removes even the subscription. If your bot does not need to compose answers from your documents, pay monthly and stop reading.
Second, AI answers on SaaS are priced per conversation, and that is where monthly bills surprise people. 3,000 Fin resolutions a month is $2,970, about ₹2,52,000, every month. 5,000 Lyro conversations at $0.70 is $3,500, about ₹3,00,000. At those volumes a one-time build of ₹2,50,000 to ₹6,00,000 pays for itself in two to three months.
Third, the subscription is never the whole bill. Add Meta's per-message fees on WhatsApp (replies included from 1 October 2026, after the first 1,000 a month on each number), agent seats above the plan limit (AiSensy charges ₹750 per extra agent a month), and the platform's markup on Meta rates. Our WhatsApp API pricing comparison works the markups out for each provider.
AI chatbot hosting cost in India per month
Hosting is the part of "AI chatbot hosting cost in India" that people under-budget in one direction and over-budget in the other. The LLM fee is not hosting. Hosting is the server, database and vector store that your bot's code runs on, and it starts at about ₹600 a month.
| Setup | What runs on it | Hosting per month |
|---|---|---|
| Rule-based or intent bot | Node or Python app, JSON or small Postgres | ₹600 to ₹1,000. Hostinger KVM 1 is ₹599 introductory and ₹999 on renewal (Zoutons, August 2026); a DigitalOcean basic droplet is $6, about ₹510 |
| Small AI chatbot | FastAPI app plus Postgres with pgvector on one 4 GB VPS | ₹1,500 to ₹3,000 |
| Mid-market AI chatbot | Managed Postgres, Redis, two app instances, backups, AWS or Azure Mumbai | ₹8,000 to ₹25,000 |
| Managed vector database | Pinecone Standard instead of pgvector | About $50 (₹4,250) a month minimum, added to the above |
| Enterprise | Kubernetes, multi-zone, WAF, audit logging, monitoring | ₹40,000 to ₹2,00,000 |
Two ways we keep this low for clients. Use pgvector inside the Postgres you already pay for instead of a separate vector database until you pass roughly a million chunks. And host in Mumbai regions (AWS ap-south-1, Azure Central India) so the DPDP Act data-residency conversation is short.
SaaS chatbots bundle hosting into the subscription, which is the real reason a ₹1,500-a-month plan is unbeatable for small flow bots.
LLM API fees: the monthly cost of the AI brain
The development fee is one bill; the model's API usage is another, and it scales with conversation volume. These are list prices in October 2026 (OpenAI pricing page; Anthropic pricing page), converted at ₹85 to the dollar, assuming a typical support conversation of 8 exchanges that sends about 12,000 input tokens (your retrieved document passages included) and 1,500 output tokens.
| Model | Price per million tokens (input / output) | Cost per 1,000 conversations | Best for |
|---|---|---|---|
| GPT-4o mini | $0.15 / $0.60 | ₹150 to ₹300 | High-volume FAQ, cost-sensitive |
| Claude Haiku 4.5 | $1.00 / $5.00 | ₹1,200 to ₹2,000 | Fast, strong reasoning for the price |
| Claude Sonnet 5.5 | $2.00 / $10.00 | ₹2,500 to ₹4,000 | Best overall quality for business support |
| GPT-4o | $2.50 / $10.00 | ₹3,000 to ₹4,500 | Balanced quality, good Hindi |
Prompt caching, which both OpenAI and Anthropic offer, cuts the cost of the repeated knowledge-base prefix by up to 90 percent, so real bills land 40 to 60 percent below the table once configured.
Monthly LLM cost by business size
- Small business, 500 conversations a month: ₹100 to ₹2,000
- Growing SME, 5,000 conversations a month: ₹1,200 to ₹20,000
- Mid-market, 25,000 conversations a month: ₹6,000 to ₹1,00,000
- Enterprise, 1,00,000+ conversations a month: ₹25,000 to ₹4,00,000
Five ways to cut the LLM bill
- Route 80 percent of queries to GPT-4o mini or Haiku 4.5. Send only escalations and complex questions to Sonnet or GPT-4o.
- Cache common answers. If "what are your timings" is asked 100 times a day, serve it from a cache, not the model.
- Limit context. Pass the 2 to 3 KB of relevant passages, not 50 KB of knowledge base, with every query.
- Cap response length at about 300 tokens unless the question needs more.
- Pre-filter with rules. Order status, pricing and location lookups go straight to your database before the LLM is involved.
Done properly these cut LLM spend by 60 to 80 percent with no quality loss.
What drives the build cost
Four things move a quote from ₹35,000 to ₹8,00,000.
1. Intelligence level
Rule-based: fixed decision trees. Cheap, brittle, cannot handle a rephrased question.
Intent-based: Dialogflow or Rasa recognises intent across varied phrasing. Medium cost, medium flexibility.
LLM-powered: GPT-4o, Claude or Gemini composes answers from your documents. Highest cost, handles anything a customer types.
Hybrid: rules for structured flows (order status, appointment booking), LLM for everything else. About 70 percent of our clients end up here because it balances cost and capability.
2. Integration depth
A bot that only answers questions is cheap. Each system it has to write to (CRM, payments, calendar, product catalogue, human handoff with full context) adds ₹15,000 to ₹80,000.
3. Channel coverage
Website widget only, or WhatsApp only, is cheapest. Running identically on website, WhatsApp, Instagram DM, Messenger and Telegram can double the build because each channel's API behaves differently.
4. Knowledge base size
Training on 20 FAQs is trivial. Training on 10,000 pages of documentation, product specs and old support tickets needs a vector database, retrieval-augmented generation and careful chunking, adding ₹50,000 to ₹3,00,000.
Tier 1: Rule-based FAQ bot (₹35,000 to ₹1,25,000)
Who it is for: small businesses with a handful of common questions: clinics, small stores, local service providers.
What it does: answers 20 to 50 predefined questions, collects name, phone and requirement, routes complex queries to WhatsApp or a phone number, runs as a website widget or Messenger bot.
What it does not do: remember earlier messages, understand rephrased questions, or write anything beyond its templates.
Stack: React or Next.js widget, Node.js keyword matching, JSON question-answer pairs.
Example: a dental clinic in Hyderabad wanted answers for timings, appointments, prices for RCT, braces and cleaning, and directions. Built in 14 days for ₹45,000. 73 percent of enquiries now get an instant answer; 27 percent reach the receptionist.
Monthly running cost: ₹600 to ₹2,000, hosting only, no LLM fees.
Tier 2: Intent-based bot (₹1,25,000 to ₹3,00,000)
Who it is for: businesses handling 500+ queries a month that need structured flows.
What it does: understands intents across 50 to 200 phrasings, remembers context across turns, books into a calendar, pushes leads to a CRM, fills forms conversationally, works in Hindi and English.
Stack: Dialogflow CX or Rasa, Python FastAPI or Node.js, PostgreSQL for conversation logs, webhooks to CRM and calendar.
Example: a Bengaluru fitness studio chain wanted free-trial booking, fee answers, class schedules and rescheduling. Built in 38 days for ₹1,95,000. 65 percent of trial bookings now arrive through the bot with no human touch.
Monthly running cost: ₹2,000 to ₹8,000, hosting plus Dialogflow calls above the free tier.
Tier 3: AI chatbot with GPT or Claude (₹2,50,000 to ₹6,00,000)
This is where "we want a ChatGPT-type chatbot" usually lands.
Who it is for: coaching institutes, insurance agencies, real estate firms, B2B SaaS: anyone with complex questions, a large knowledge base, or leads to qualify.
What it does: natural conversation with GPT-4o or Claude Sonnet 5.5, answers from your own policies, catalogue and documentation, copes with typos and Hinglish, qualifies leads with follow-up questions, hands over to a human with the full transcript, optional voice input, multiple languages out of the box.
Stack: OpenAI or Anthropic Claude via API, Pinecone or Supabase pgvector, LangChain or a custom pipeline, Python FastAPI or Node.js, React widget or WhatsApp.
Example: a coaching institute franchise needed admissions answers across 40 branches: fees, courses, faculty, results, facilities. We built a retrieval-based GPT-4o chatbot trained on their 380-page prospectus, FAQ and historical WhatsApp logs. ₹4,20,000, 9 weeks. It now closes 71 percent of enquiries without a human, and enquiry-to-enrolment conversion rose 34 percent against the old "call the branch" flow.
Monthly running cost: ₹8,000 to ₹40,000, mostly LLM usage at ₹0.20 to ₹4.50 per conversation depending on the model.
Tier 4: WhatsApp AI chatbot (₹2,00,000 to ₹5,50,000)
A WhatsApp AI chatbot adds Meta's Business API integration on top of the AI work. Our WhatsApp Business API guide covers the setup steps.
What it does: GPT-quality answers inside WhatsApp, rich media (images, PDFs, video, location), order status, booking and payment flows, handover to a shared team inbox, campaign triggers based on conversation outcome, Hindi, English and regional languages.
Example: a Gurugram real estate firm wanted to qualify Facebook-ad leads on WhatsApp. The bot asks budget, location, timeline and loan need conversationally, routes hot leads to senior agents and cold ones to a nurture sequence. ₹3,10,000, 7 weeks. Lead-qualification time fell 82 percent.
Monthly running cost: ₹5,000 to ₹25,000 for LLM and hosting, plus Meta's per-message fees via PayPerWA, which adds a ₹0.20 per message platform fee on top, or via a subscription BSP (WATI vs Interakt vs AiSensy compared).
Tier 5: Enterprise AI assistant (₹5,50,000 to ₹15,00,000+)
Who it is for: banks, NBFCs, insurers, hotel chains, airlines, large e-commerce: anywhere a wrong answer or an outage has a real cost.
What it does: heavily prompted or fine-tuned model for strict brand and compliance tone, deep integration with core banking, ERP, inventory or order management, voice as an IVR replacement, multi-language at human-agent quality, role-based internal bots (HR, IT support), sentiment tracking and automated QA, SOC 2 or ISO 27001 hosting, 99.9 percent uptime SLA.
Stack: GPT-4o or Claude with evaluation suites, enterprise Pinecone or self-hosted Weaviate, Python microservices on AWS or Azure, Kubernetes with auto-scaling, encryption, audit logging and PII redaction.
Example: a private-sector bank wanted an internal HR assistant for leave, payroll, benefits and policy questions across 4,200 employees. ₹14,20,000 over 18 weeks, trained on 11 years of HR policy documents. Target was a 60 percent drop in HR helpdesk tickets; actual was 68 percent in the first six months.
Monthly running cost: ₹40,000 to ₹3,00,000 for LLM usage, infrastructure and monitoring.
Custom chatbot vs monthly SaaS: 3-year cost at 5,000 conversations a month
The question I get most, with October 2026 prices and ₹85 to the dollar.
| Approach | Year 1 | Year 2 | Year 3 | 3-year total |
|---|---|---|---|---|
| WATI Pro flow bot, no LLM answers | ₹1,00,000 | ₹1,00,000 | ₹1,00,000 | ₹3,00,000 |
| Intercom Fin, 60 percent resolved (3,000 × $0.99 a month) plus seats | ₹31,00,000 | ₹31,00,000 | ₹31,00,000 | ₹93,00,000 |
| Custom AI chatbot (Codingclave) | ₹3,50,000 build + ₹1,50,000 running | ₹1,50,000 | ₹1,50,000 | ₹8,00,000 |
Running cost for the custom bot is LLM usage on a mixed GPT-4o mini and Sonnet routing (about ₹6,000 a month), hosting (₹2,000) and a maintenance retainer (₹4,500).
The honest reading: if a flow bot does the job, SaaS at ₹3 lakh over three years is the cheapest option and I will tell you so on the call. The moment you need AI answers from your own documents, per-resolution SaaS pricing climbs past a custom build within months. Below about 500 AI conversations a month, a SaaS AI agent is still cheaper; above 1,500, custom wins inside year one. For a longer version of this trade-off see chatbot cost in 2026: SaaS vs custom.
Timelines: first call to production
| Tier | Development | Testing | Training and tuning | Go-live | Total |
|---|---|---|---|---|---|
| Rule-based FAQ | 2 weeks | 3 days | 3 days | 2 days | about 3 weeks |
| Intent-based | 4 weeks | 1 week | 1 week | 3 days | about 6 weeks |
| AI chatbot | 6 weeks | 2 weeks | 2 weeks | 1 week | about 11 weeks |
| WhatsApp AI | 5 weeks | 1.5 weeks | 1.5 weeks | 1 week plus Meta verification | about 10 weeks |
| Enterprise | 12 weeks | 3 weeks | 3 weeks | 2 weeks | about 20 weeks |
Five mistakes that inflate chatbot cost
- Over-engineering version one. 200 intents, 15 integrations and 6 languages on day one. Ship 20 intents in English, then add what usage data asks for.
- Paying for the wrong model. GPT-4o on FAQ traffic that GPT-4o mini handles identically costs 15 times more per token.
- No fallback path. When the AI cannot answer, where does the customer go? Without a handover, every edge case becomes a complaint.
- Skipping knowledge-base cleanup. Messy PDFs, an outdated website and four "official" FAQ versions. Clean this first; it decides 80 percent of answer quality.
- Measuring nothing after launch. Questions drift. Without a monthly review, accuracy drops 10 to 15 percent a year.
How to bring the price down
- Start narrow. "Answer pricing questions" ships in 3 weeks for ₹60,000. "Handle all support" takes 12 weeks for ₹4,00,000.
- Bring clean material. A good FAQ, product docs or chat transcripts make development about 30 percent faster.
- Pick one channel. WhatsApp or website first; omnichannel can double the cost and v1 rarely needs it.
- Skip the custom interface. A standard widget or WhatsApp works. Custom chat UIs add ₹40,000 to ₹1,20,000 and rarely lift conversion.
- Build in India. US and UK agencies quote $15,000 to $80,000 (₹12 lakh to ₹68 lakh) for the same scope. Indian teams, ours included, deliver at roughly a third, with Hinglish, Indian names and DPDP Act compliance handled natively.
What changed in AI chatbots between 2024 and 2026
If your last evaluation is older than 12 months, these are the shifts that change the price maths.
1. Model quality crossed the support threshold
GPT-4o, GPT-4.1 and Claude Sonnet 5.5 now resolve 50 to 65 percent of FAQ traffic end to end, against 15 to 25 percent in 2023. In 2026, a rule-based bot is rarely the right first build; go AI-powered or do not build.
2. Retrieval-augmented generation made hallucination rare
With Pinecone or pgvector, the bot answers only from your documents. Hallucination fell from 8 to 15 percent of answers (2023) to under 1 percent (2026). RAG is standard, not optional.
3. WhatsApp plus AI is the Indian default
Indian customers choose WhatsApp over web chat three to five times more often. The common stack is WATI, Interakt or AiSensy (or PayPerWA) for the channel, OpenAI or Claude for answers, and Pinecone or pgvector for retrieval. We have built 25+ of these; 60 to 75 percent FAQ deflection on WhatsApp alone is typical.
4. Multi-agent flows with LangGraph
For research, draft, review and escalate workflows, a single agent is not enough. LangChain with LangGraph orchestrates multi-step agents; we use it in legal contract review, healthcare triage and financial advisory bots.
5. Per-resolution cost collapsed
- 2023: $0.30 to $1.00 per resolution
- 2024: $0.10 to $0.30
- 2026: $0.03 to $0.10 with caching and smart routing
A SaaS company handling 5,000 tickets a month runs an AI chatbot for ₹15,000 to ₹50,000 a month against ₹3,00,000 for the equivalent human team.
6. Voice agents are production-ready
ElevenLabs Conversational AI runs real-time voice agents with sub-100 ms latency that handle interruptions. Virtual receptionists and voice support cost ₹40,000 to ₹1,00,000 a month in production; see our AI voice agent development guide.
7. Indian languages at near-native quality
GPT-4o and Claude handle Hindi, Tamil, Telugu, Bengali, Marathi, Kannada and Gujarati well enough for Tier-2 and Tier-3 audiences, and switch language mid-conversation.
8. Intercom Fin made "buy" simple and expensive
Intercom Fin is a drop-in AI agent at $0.99 per resolution on top of your seat plan. Buy it if you already run Intercom and your support is standard FAQ. Build if you need your own tools, compliance control or predictable cost at volume. Full comparison: AI customer support for Indian SaaS, build vs buy.
Next steps before you ask for a quote
- Define the one problem to solve. One, not ten. Scoping accuracy improves tenfold.
- Choose the channel. Website, WhatsApp or both.
- Gather the knowledge base. FAQ, product information, policy documents. Cleaner input, cheaper build.
- Estimate monthly conversations. This decides whether LLM fees matter and whether SaaS or custom is cheaper.
- Ask for a fixed price. Hourly billing on chatbot projects invites scope creep.
Ready to price your chatbot?
At Codingclave we build AI chatbots for healthcare, education, real estate, D2C, logistics and insurance businesses, on fixed-price, fixed-timeline engagements.
- See our custom software services
- For WhatsApp bots without a subscription, see PayPerWA
- Read the WhatsApp Business API guide for the channel setup
- Or book a free 30-minute consultation and get a realistic figure for your scope
Founder note: happy to take a 15-minute WhatsApp call to scope what you actually need before you spend anything. Message +91 92771 84741. If a ₹1,500-a-month plan fits your size, I will say so instead of selling you a custom build.
Related resources
- AI Customer Support for Indian SaaS 2026: Build vs Buy
- WATI vs Interakt vs AiSensy 2026 Comparison
- WhatsApp Business API Complete Guide India 2026
- WhatsApp API Pricing India 2026: 5 BSPs Compared
- OpenAI GPT Integration Services
- Claude (Anthropic) Integration Services
- Intercom Integration Services
- Pinecone Vector Database Integration
- LangChain Integration Services
Need an AI chatbot built for your business? Get a free 30-minute strategy call and we will review your use case, your knowledge base and whether to build or buy.
Or WhatsApp Ashish for a quick reply.
Last updated: 4 October 2026, with vendor prices re-checked against WATI, AiSensy, Interakt, Tidio, Intercom, OpenAI and Anthropic pricing pages. Reviewed by Ashish Sharma, founder of Codingclave Development LLP, Lucknow. 50+ AI chatbot projects shipped across SaaS, e-commerce, healthcare, legal and EdTech since 2024. Top Rated on Upwork. 4.9/5 Google rating across 76 reviews.
Frequently asked questions
A chatbot in India costs ₹1,500 to ₹16,999 per month on a subscription platform (AiSensy Basic ₹1,500, WATI Growth ₹2,499, WATI Business ₹16,999, all plus GST), or ₹35,000 to ₹8,00,000 one-time for a custom build. A rule-based FAQ bot is ₹35,000 to ₹1,25,000. An AI chatbot that answers from your own documents with GPT-4o or Claude is ₹2,50,000 to ₹6,00,000. Enterprise assistants with core-system integrations start at ₹5,50,000. On top of the build, budget ₹600 to ₹25,000 a month for hosting and ₹200 to ₹20,000 a month for LLM API usage, depending on conversation volume. These are the ranges we have quoted across 50+ chatbot projects since 2024.
As of October 2026, AiSensy charges ₹1,500 (Basic) and ₹3,200 (Pro) per month plus GST, with chatbot flows as a ₹2,500 per month add-on. Interakt is ₹3,499 per quarter (Starter) and ₹7,699 per quarter (Growth), roughly ₹1,166 and ₹2,566 a month. WATI is ₹2,499 (Growth), ₹5,999 (Pro) and ₹16,999 (Business) per month. PayPerWA has no subscription at all and charges ₹0.20 per message as a platform fee. Western AI agents are priced per use: Tidio Lyro starts at $39 a month for 50 AI conversations and Intercom Fin charges $0.99 per resolved conversation. Every WhatsApp plan also passes on Meta's per-message fees, about ₹0.86 per marketing message and ₹0.115 per utility message, and from 1 October 2026 bot replies inside the 24-hour window are charged at the utility rate too, after each business number's first 1,000 a month.
For a custom chatbot, the monthly bill has three parts. Hosting is ₹600 to ₹3,000 for a small VPS and ₹8,000 to ₹25,000 on AWS or Azure Mumbai for mid-market traffic. LLM API usage is ₹150 to ₹300 per 1,000 conversations on GPT-4o mini and ₹2,500 to ₹4,500 per 1,000 on Claude Sonnet 5.5 or GPT-4o, so a business handling 5,000 conversations pays ₹1,200 to ₹20,000 a month depending on the model. Maintenance and monitoring, if you outsource it, is ₹5,000 to ₹15,000 a month. A small business therefore runs a custom AI chatbot for ₹3,000 to ₹10,000 a month; a mid-market company for ₹20,000 to ₹60,000. WhatsApp bots add Meta's per-message fees on top, including replies beyond each business number's first 1,000 a month.
Hosting for an AI chatbot in India starts at about ₹600 a month. A rule-based bot runs on a single small VPS: Hostinger's KVM 1 plan is ₹599 a month introductory and ₹999 on renewal (Zoutons price tracker, August 2026), and a DigitalOcean basic droplet is $6, about ₹510. An AI chatbot with a vector database fits on a 4 GB VPS with Postgres and pgvector for ₹1,500 to ₹3,000 a month. Once you need a managed database, Redis, two app instances and backups on AWS or Azure Mumbai, hosting is ₹8,000 to ₹25,000 a month. A managed vector database such as Pinecone's Standard plan adds about $50 (₹4,250) a month minimum. Enterprise deployments on Kubernetes with monitoring and a WAF run ₹40,000 to ₹2,00,000 a month. SaaS chatbots include hosting in the subscription.
Pay monthly if your bot follows fixed flows (menus, booking, order status) and you handle under 1,500 conversations a month. A WATI Pro or AiSensy Pro plan with chatbot flows costs ₹70,000 to ₹1,00,000 a year all in, and nothing custom beats that. Build custom when you need the bot to answer from your own documents, integrate with your CRM or ERP, or when a per-resolution AI agent gets expensive: Intercom Fin at $0.99 per resolution costs about ₹2,50,000 a month for 3,000 resolutions, while a custom AI chatbot costs ₹2,50,000 to ₹6,00,000 once and ₹10,000 to ₹40,000 a month to run. Above about 1,500 AI conversations a month, custom wins inside the first year.
The WhatsApp chatbot market in India splits into three groups. Business Solution Providers with flow builders: WATI (₹2,499 to ₹16,999 a month), AiSensy (₹1,500 to ₹3,200 a month plus a ₹2,500 chatbot add-on), Interakt (₹3,499 to ₹7,699 a quarter), Gupshup, Kaleyra and Infobip for enterprise volumes, and PayPerWA, our own pay-per-use platform with no subscription and a drag-and-drop chatbot builder. Conversational AI platforms with LLM answers: Yellow.ai, Verloop and Haptik, usually quoted per enterprise. Custom development: agencies like Codingclave that build a GPT or Claude powered bot on Meta's Cloud API and connect it to your systems. Most SMBs start in the first group and move to the third when they need answers from their own knowledge base.
A rule-based FAQ chatbot with 20 to 50 answers takes about 3 weeks from first call to go-live. An intent-based bot on Dialogflow or Rasa with booking and CRM integration takes about 6 weeks. An AI chatbot that answers from your knowledge base with retrieval-augmented generation takes 9 to 11 weeks, including two weeks of testing against real customer questions. A WhatsApp AI chatbot takes about 10 weeks because Meta's Business API verification and template approval sit on the critical path. Enterprise assistants with core-system integrations and compliance reviews take 16 to 20 weeks. The single biggest delay in every project is a messy knowledge base, so clean your FAQ and policy documents before development starts.
For English customer support, GPT-4o and Claude Sonnet 5.5 are the quality benchmark in our tests. For Hindi and Hinglish, GPT-4o, Claude Sonnet 5.5 and Google Gemini all answer at near-native quality in 2026. For Tamil, Telugu, Bengali, Marathi and Kannada, test two models on 50 of your real customer messages before committing; Gemini and GPT-4o have been strongest for us on south Indian languages, but the gap changes with each release. For cost-sensitive deployments, GPT-4o mini and Claude Haiku 4.5 give 80 percent of the quality at roughly a tenth of the price (GPT-4o mini is $0.15 per million input tokens against $2.50 for GPT-4o, per OpenAI's pricing page). Route simple questions to the cheap model and escalate hard ones.
Usually that search means OpenAI's personal ChatGPT plans, not a business chatbot. In India, OpenAI launched ChatGPT Go at ₹399 a month in August 2025, ChatGPT Plus is $20 a month and ChatGPT Pro is $200 a month; check OpenAI's pricing page for the current rupee figures. Those plans give one person a chat window. They do not sit on your website or WhatsApp, do not answer from your documents, and cannot talk to your CRM. A business chatbot that does those things is either a SaaS subscription of ₹1,500 to ₹16,999 a month or a custom build of ₹35,000 to ₹8,00,000, and it uses the same models through the API, where you pay per token instead of per seat.
Three layers. First, retrieval-augmented generation: the bot answers only from passages pulled out of your own documents, which cut hallucination in our projects from 8 to 15 percent of answers in 2023 to under 1 percent in 2026. Second, confidence scoring: when retrieval finds nothing relevant, the bot says so and hands the chat to a human instead of guessing. Third, an evaluation suite of 100 to 300 real customer questions with approved answers, run before every prompt or model change so regressions are caught before customers see them. We use Pinecone or pgvector for retrieval, LangChain for orchestration and Langfuse for tracing. Budget 10 to 15 percent of the build cost for this layer; it is the part clients most often try to skip.