Chatbot cost in 2026: a scenario bot starts at $1,900, a lead capture bot connected to your CRM runs $4,000–8,500, and an AI assistant on GPT with a knowledge base costs $13,000–27,000. Turnkey chatbot development takes one to six weeks, and we estimate your scenario within 24 hours.
| Bot type | What it does | Cost |
|---|---|---|
| Scenario bot | FAQs, menus, bookings, simple scenarios without integrations | from $1,900 |
| Lead capture bot with CRM | Qualifies leads, hands them off to HubSpot or Pipedrive, notifies managers | $4,000–8,500 |
| AI assistant on GPT | Answers from the knowledge base, free-form conversation, integrations | $13,000–27,000 |
You can ballpark your own scenario in the chatbot cost calculator, and we'll prepare an exact quote within 24 hours of a call.
The bot itself pays off quickly: a working bot saves a business money every month on support, request handling and lead qualification, from part of one employee's workload to that of several people. Prices on the market vary wildly, from “free on a bot builder” to “$70,000 for an AI assistant”. Let's look at what your business actually needs and what it costs.
This article is an honest breakdown of costs by bot type, platform and feature. No marketing haze from bot builders, and no scary price tags from enterprise integrators.
(FAQs, menus, forms)
(CRM, accounting system, payments)
(GPT, RAG, training)
Three types of chatbots and what they cost
Not all bots are the same. A button bot for a pizzeria and an AI assistant for a law firm are different products with different costs, timelines and business impact.
Type 1: Scenario bot (from $1,900)
It follows pre-written scenarios: buttons, menus, dialog branches, forms. It doesn't understand free text, only button taps and choices.
What it can do: answer FAQs (20–50 questions), show a catalog or menu, collect requests (name, phone, topic), book appointments or tables, send notifications (order status, reminders).
Cost by complexity:
Simple ($1,900–3,300): 10–20 scenarios, an FAQ of up to 30 questions, request collection, notifications. Timeline: 1–2 weeks.
Medium ($3,300–6,000): 20–50 scenarios, an interactive catalog, forms with validation, basic personalization (greeting by name), multiple languages. Timeline: 2–4 weeks.
Complex ($6,000–8,500): 50+ scenarios, A/B tests of dialogs, geolocation, dynamic content from a database, analytics by scenario. Timeline: 3–5 weeks.
Good for: small businesses, restaurants, beauty salons, clinics, online stores with routine questions. If 80% of customer requests are the same 20 questions, a scenario bot will handle them for $1,900.
Type 2: Bot with integrations ($5,000–12,000)
The same scenario bot, but connected to your systems: CRM, accounting or ERP system, payment providers, warehouse, calendar. It doesn't just collect requests, it takes action: creates a deal in the CRM, writes off stock, accepts payment, books a time slot.
What it can do on top of a scenario bot: check product availability in real time, accept payments (Stripe, PayPal or your bank's gateway), create and update deals in the CRM (HubSpot, Pipedrive, Salesforce), show order status from your accounting or ERP system, book calendar slots with an availability check.
Cost by number of integrations:
1–2 integrations ($5,000–7,500): bot + CRM + notifications. A typical case: the bot collects a request, creates a deal in HubSpot and notifies the manager in Telegram. Timeline: 3–5 weeks.
3–4 integrations ($7,500–10,000): bot + CRM + payments + a catalog from the database. A typical case: the full ordering cycle in Telegram without going to the website. Timeline: 4–6 weeks.
5+ integrations ($10,000–12,000): bot + CRM + ERP + payments + warehouse + shipping. A mini ERP inside a messenger. Timeline: 6–8 weeks.
Good for: mid-sized businesses with established processes, e-commerce, service companies, companies with a high volume of requests (50+ a day). For more on Telegram bot costs, see our separate breakdown; for widget bots, see website chatbot pricing.
Type 3: AI assistant ($12,000–27,000+)
A bot that understands free text, answers non-standard questions, works with the company's knowledge base and learns from your data. Under the hood: GPT-4, Claude or another LLM plus RAG (Retrieval-Augmented Generation) for working with corporate documents.
What it can do: answer any question from your knowledge base (policies, documentation, a catalog of 10,000+ products), follow the context of the conversation (it remembers what was said 5 messages ago), qualify leads through free-form dialog, draft proposals from templates, and hand off to an operator when it isn't sure of the answer.
Cost by complexity:
Basic AI bot ($12,000–17,000): GPT/Claude + a knowledge base of up to 1,000 documents, free-text understanding, fallback to an operator. Timeline: 4–8 weeks.
Advanced ($17,000–27,000): everything above + multimodality (it understands photos and documents), RAG with fine-tuning, conversation analytics, automatic knowledge base updates. Timeline: 8–12 weeks.
Enterprise ($27,000–70,000): everything above + your own model (a fine-tuned LLM), on-premise deployment (data never leaves your servers), multiple channels (Telegram + WhatsApp + website + email), ERP/CRM integration, sentiment analysis. Timeline: 3–6 months.
Good for: companies with a large volume of requests (200+ a day), law and consulting firms, SaaS tech support, banks and fintech, healthcare companies. For more on applying AI, see our article GPT and business processes.
Cost by platform
The platform affects the price. Not all messengers are equally developer-friendly.
Telegram
Added cost on top of the base price: $0. The cheapest and most convenient platform for bots. A free Bot API, no moderation (unlike WhatsApp) and rich features: inline buttons, Web Apps, payments, groups, channels. Telegram Mini Apps let you embed a full web interface right in the bot.
Limitations: not all of your customers use Telegram, and its reach varies a lot from country to country, so check where your audience actually is.
Added cost: +$1,000–2,500 for setup. The WhatsApp Business API requires business verification and a connection through a provider (Gupshup, Twilio, 360dialog), and you pay Meta's messaging fees, which depend on the country and the message type. At 1,000 conversations a day, those fees become a noticeable monthly line item.
The upside: WhatsApp is the main messenger for B2C audiences aged 35+. For some businesses (delivery, taxis, healthcare), WhatsApp gets a 40–60% higher response rate than Telegram.
Website widget
Added cost: +$1,300–4,000. A chat widget on your website that can hand off to the bot. It includes widget design (matched to your brand), a mobile version, pre-chat forms and display triggers (by time, by scroll depth, on exit intent).
Multiple channels (2+ platforms)
Added cost: +$1,700–5,000 for each platform after the first. One bot logic running in Telegram, WhatsApp, on your website and in social media messaging. It needs a single backend and message routing. Going to two channels doesn't double the base cost: 70–80% of the code is reused.
Hidden costs: what the initial quote leaves out
Development is 50–60% of what a bot really costs. The rest is what not everyone mentions at the start.
Content for the bot
$700–2,500. Answer texts, dialog scenarios and FAQs aren't a matter of “just copying from the website”. Good bot copywriting means short, tight answers, the right tone and variants for different situations. For an AI bot, it means preparing the knowledge base: cleaning, structuring and tagging documents.
Testing
10–15% of the development cost. The bot has to cope with unexpected input: random characters, voice messages, stickers, attachments. Plus load testing: what happens if 1,000 people write at once (a promotion, a mailing).
Hosting and infrastructure
Roughly $10–400 a month. For a scenario bot, a VPS at around $10–20 a month is enough. A bot with integrations: about $30–100 a month (it needs more memory and CPU). An AI bot: about $100–400 a month (the OpenAI/Anthropic API plus your own server for RAG). For AI bots, API fees are a significant expense: at 1,000 requests a day, a GPT-4-class model costs roughly $300–1,000 a month, depending on the model and how much context each request carries.
Support and updates
$500–1,700 a month. Uptime monitoring, updating answers (the business changes, so the bot has to change too), bug fixes, new scenarios, log analysis and optimization. Without support, a bot goes stale in 3–6 months: answers get outdated and integrations break when APIs change.
Analytics
$1,000–2,500 (one-time). A dashboard with metrics: number of conversations, conversion, the most frequent questions, drop-off points, NPS, average response time. Without analytics, you won't know whether the bot works or annoys your customers.
Set aside 40–60% on top of the development cost for hidden expenses. A $5,000 bot will really cost $7,000–8,000 in the first year (development + content + hosting + support). But if the bot takes over the workload of even one operator, the ROI turns positive by month 3–4.
Bot builder vs custom development
“Why pay $5,000 when Chatfuel or ManyChat lets you build it for free?” A fair question. Let's break it down.
Bot builders (free to a few hundred dollars a month)
Pros: a fast launch (1–3 days), no developer needed, a visual scenario editor, built-in analytics.
Cons: limited logic (no complex conditions or loops), a subscription that grows with your contact list and gets noticeable by 10,000 contacts, limited integrations (a CRM, yes; your ERP or custom APIs, no), the bot is tied to the platform (if the builder shuts down, your bot dies), and there's no access to the code (you can't hand it over to another developer).
When it fits: testing a hypothesis (do you need a bot at all), simple scenarios of up to 20 steps, a budget under $1,500.
Custom development (from $1,900)
Pros: logic of any complexity, integration with any system, the code belongs to you, no per-contact subscription, full control and customization.
Cons: it takes longer (2–8 weeks vs 1–3 days), you need a developer for support, and the upfront cost is higher.
When it fits: a bot with integrations, an AI bot, non-standard logic, 50+ users a day, a bot as a business-critical channel.
A bot builder is for testing an idea. Custom development is for a business tool. If a bot makes or saves you money, it has to be reliable. And a builder's reliability depends on a third-party vendor you have no influence over.
How a bot pays for itself: the ROI math
A bot is an investment. Let's calculate the return.
Savings on support
A support agent costs roughly $2,500–4,500 a month with payroll taxes, depending on the country. One bot replaces 1–3 agents on routine questions. Savings: roughly $2,500–13,500 a month. With the bot at $5,000, payback takes 1–2.5 months.
Higher conversion
A bot answers instantly (vs 5–30 minutes for a human operator). Response speed lifts conversion by 15–30%. For an online store with 10,000 visits a month and a 2% conversion rate, +0.5% conversion means +50 orders. With an average order of $100, that's +$5,000 a month.
Lead qualification
The bot asks 3–5 questions and hands the manager a “warm” lead with context. The manager saves 5–10 minutes on every lead. At 50 leads a day, that's 4–8 hours a day saved, or 0.5–1 manager.
Working 24/7
30–40% of requests come in outside business hours. Without a bot, those are lost leads. With a bot, requests are collected around the clock. For a business with 100 requests a day: +30–40 handled leads = +10–15 sales a month (at a 30–40% conversion rate).
and no sick days
What drives the cost: a checklist for your estimate
To get an accurate estimate, prepare answers to these questions.
Our approach: chatbots from $1,900
At March Code we build bots for Telegram, WhatsApp and websites on Node.js, with Python for the AI components. More on Telegram bots on our Telegram bot development page.
The process: 1) analysis of the task and scenarios (2–3 days), 2) designing the dialogs and integrations (3–5 days), 3) development and testing (1–6 weeks), 4) launch and monitoring (1 week), 5) optimization based on analytics (ongoing).
FAQ
How much does a simple Telegram booking bot cost?
$1,900–4,000. It includes a welcome message, choice of service, choice of date and time from the open slots, contact details collection, booking confirmation and a reminder 24 hours before. If you need integration with Google Calendar or a CRM, add $1,000–2,000. Timeline: 1–3 weeks.
Are a website chatbot and a Telegram bot two separate bots?
Not if it's designed properly. One backend, two front ends (the Telegram Bot API plus a website widget). Cost: the base bot plus $1,300–2,500 for the widget. 80–90% of the business logic code is reused; only the UI layer differs.
How much does an AI bot that answers from the company's knowledge base cost?
$12,000–17,000 for the basic version. It includes connecting GPT-4/Claude via API, loading and indexing the knowledge base (up to 1,000 documents), a RAG pipeline for accurate answers and fallback to an operator when confidence is low. Running costs: roughly $150–600 a month (API and hosting). It pays off if it replaces 2+ support agents.
Can I start with a simple bot and add AI later?
Yes, and that's the best strategy. Launch a scenario bot ($1,900–5,000), collect data on customers' real questions for 2–3 months, then analyze which questions the scenarios don't cover. If that's more than 30%, add an AI module (+$6,500–12,000). The data from the first stage is used to train and test the AI.
Will a bot replace a call center?
Partly. Routine questions (order status, opening hours, product availability) the bot closes 70–85% of the time. Complex cases (complaints, unusual situations, upset customers) need a human. The real-world effect: 40–60% of requests move to the bot, and the load on operators halves. It's not “fire all the operators” but “operators handle complex cases instead of answering ‘where's my order?’”.
What guarantees are there that the bot will run reliably?
An SLA: 99.5% availability (no more than 3.6 hours of downtime a month), bot response time under 2 seconds, 24/7 monitoring with alerts on failures. All of this is part of the support contract ($500–1,700 a month). Without support, the warranty covers only the code (6 months), not the integrations and APIs staying up.
How long does it take to build an AI bot?
A basic AI bot: 4–8 weeks. That breaks down into 1 week of analysis and design, 1 week of knowledge base preparation (the most underestimated stage), 2–4 weeks of development and integrations, and 1–2 weeks of testing and fine-tuning. An advanced bot with custom models: 8–16 weeks.
