How much does an AI chatbot cost?
A chatbot has two prices: the build, which is scoped work, and the running cost, which is almost entirely model tokens plus messaging fees. The running cost surprises people by how small it is. Ten thousand support conversations of 3,000 tokens in and 700 out each come to roughly $6.50 in model usage on the cheapest current Claude model at list prices.
TL;DR verdict
Budget for the integrations, not the model. At list prices the model is often the cheapest line on the invoice. The money goes into wiring the bot into your CRM, calendar and order data so it has something true to say.
| Cost line | How it is billed | Current figure |
|---|---|---|
| Model, cheap tier | Per million tokens in and out | Claude Haiku 5.5: $0.10 in / $0.50 out per million, for prompts up to 100,000 tokens |
| Model, mid tier | Per million tokens in and out | Claude Sonnet 5.5: $2 in / $10 out per million |
| Model, high tier | Per million tokens in and out | Claude Opus 5.5: $4 in / $20 out per million |
| Model, top tier | Per million tokens in and out | Claude Fable 5.1: $10 in / $50 out per million |
| Worked example | 10,000 support conversations, each 3,000 tokens in and 700 out (our assumption) | ~$6.50 on Claude Haiku 5.5, ~$130 on Claude Sonnet 5.5 |
| Repeated context | Cache read vs standard input | 0.1x the input price or less on a cache hit (0.05x on Sonnet 5.5 and Opus 5.5, 0.025x on Fable 5.1) |
| WhatsApp replies | Per delivered message, by category | Free inside the 24-hour customer service window |
| WhatsApp outbound | Per delivered template message | Billed by recipient country and template category |
Which we'd pick, scenario by scenario
A web chat assistant on your own site
No messaging fees at all, so the running cost is tokens and hosting. At a few thousand conversations a month that comes to a few dollars on the cheapest model and tens of dollars on a mid-tier one. Most of the budget goes into the build.
The same assistant on WhatsApp
Replies inside the 24-hour service window are free, so a support bot that customers message first adds very little. Costs appear when you start the conversation yourself with a template, which is billed per delivered message.
Voice AI on a phone number
The most expensive channel, because speech recognition and speech synthesis are billed per minute on top of the model. It pays off where the alternative is a missed call, which is exactly the situation in our lead-response work.
The line people forget
A long system prompt and a large retrieved context ride along with every message. Caching that repeated block cuts it to a tenth of the input price or less, which matters far more at volume than switching models.
Almost every chatbot quote we are asked to sanity-check looks the same: a large build number, a vague "plus API costs" and no volume assumptions. That is backwards, because the API cost is the one line you can calculate exactly today. Our worked example assumes 3,000 tokens in and 700 out per support conversation: at Anthropic's list prices, ten thousand conversations cost about $6.50 on Claude Haiku 5.5 and about $130 on Claude Sonnet 5.5. You can run that arithmetic for your own volume in a minute.
What varies is everything around the model: whether the bot can read an order, how many channels it answers on, whether it has to hand a conversation to a person midway and keep the context, and whether it works in one language or three. Those decide the build. They also decide whether the bot is useful at all, because a chatbot with no access to your data is a FAQ page that types slowly. When we quote, the model line is usually the smallest number in the estimate, and we say so instead of hiding it in a bundle. See the systems this sits inside on the AI chatbots page, and what the messaging side costs in WhatsApp Business API pricing.
Sources
Frequently asked
How much does an AI chatbot cost per month?
For a web chat assistant, model tokens make up most of the running cost: at list prices, ten thousand support conversations of 3,000 tokens in and 700 out each cost roughly $6.50 on the cheapest current Claude model. Add messaging fees if you run on WhatsApp and per-minute fees if you run on the phone. The build is scoped separately.
Why do chatbot quotes vary so much?
Because the integrations usually drive the cost; the model rarely does. A bot that answers from a help page and a bot that reads your order database, books into your calendar and hands off to a person are the same product only in marketing copy.
How can I reduce the running cost?
Cache the parts of the prompt that never change, which cuts them to a tenth of the input price or less on repeat, and keep retrieved context tight. Both save more than switching to a cheaper model, and neither costs you answer quality.
Is a cheaper model good enough for support?
For routing, extraction and answering from retrieved documents, usually yes, and the price gap between tiers is large. We test on your real conversations before recommending one, because the answer depends on your questions, not on a benchmark.
What does the build itself cost?
It depends on channels, integrations, languages and volume, so we scope it in a free 30-minute audit and send back a written estimate split into milestones. There is no menu price, and a quote given before those questions are answered is padded.
Which process costs you the most?
Thirty minutes, free. You leave with a concrete fix plan and what the process costs you. The plan is yours, even if you build it without us.
Book a free audit