Free tool
Chatbot running cost calculator
What a bot costs to run each month: AI usage, WhatsApp fees, platform and care, and the cost of each conversation. Starts with example figures; change them to yours.
Your numbers
Conversations
0 for a fully scripted bot.
AI usage
US dollar list prices, standard tier, as published on 15 November 2026.
Question, history and your content.
Check today’s rate; this is an assumption.
WhatsApp messages you send
£0.0458 each (Meta UK rate).
£0.0159 each: confirmations, reminders.
Everything else, per month
Reviews, updates, testing.
Per-resolution fees, phone numbers…
Results
Running cost a month
£52
£625 a year
Cost per conversation
£0.05
everything included
AI cost per AI conversation
£0.00
21,300 tokens
AI usage
£2.05
a month
WhatsApp fees
£0
a month
AI usage is often the smallest line. Platform fees, messaging and the time to keep answers right usually cost more.
How platforms charge for AI answers
Many helpdesk AI agents charge per resolved conversation instead of per token. List prices seen on 15 November 2026:
| Product | Charge for AI answers |
|---|---|
| Intercom Fin (opens in a new tab) | $0.99 per outcome (a resolution, or a handoff it carries out) |
| Zendesk AI agents (opens in a new tab) | £1.30 per automated resolution beyond those included (committed), or US$2.00 pay as you go |
| HubSpot Customer Agent (opens in a new tab) | 50 HubSpot credits per resolved conversation, about $0.50 at $0.01 a credit |
| Freshdesk Freddy AI Agent (opens in a new tab) | $49 per 100 sessions after the first 500 |
Per-resolution prices add up
At a dollar a resolution, 1,000 resolved conversations a month is about $1,000 before seats. Token prices for answering from your own content are usually pennies per conversation, but you need a platform and care around them.
Questions people ask
What is a token?
The unit AI models are priced in: roughly three quarters of a word in English. Each reply sends the model the question, the conversation so far and the passages of your content it answers from (tokens in), and gets the answer back (tokens out).
Why are tokens in so much higher than tokens out?
Each reply carries the conversation so far and several passages from your documents, so a 50-word answer can need a few thousand tokens of reading. Prompt caching, which most providers offer, cuts the price of the repeated part.
Is a bigger model worth it?
For answering from your own content, small models are often enough, and they are many times cheaper. Test with your real questions: if the small model gets them right, use it. Some bots use a small model for most replies and a larger one for hard questions.
When do WhatsApp fees apply?
Meta charges per template message a business sends: marketing, utility and authentication messages. Replies within 24 hours of a customer’s message are service messages; since 1 October 2026 each number gets 1,000 free a month. The rates here are Meta’s UK list rates.
What about platform per-resolution fees?
Many helpdesk AI agents charge per resolved conversation instead of per token. If you use one, put the fee times the conversations it resolves in “Other”, and set the AI share to 0.
Plan the bot behind the numbers
Pick the jobs, write the answers and set the handoff in the free bot designer, then test it in your browser.
Sources
- Anthropic, Claude API pricing (opens in a new tab)
- OpenAI, API pricing (opens in a new tab)
- Google AI Studio, Gemini API pricing (opens in a new tab)
- Meta for Developers, WhatsApp Business Platform pricing (rate cards from 1 October 2026) (opens in a new tab)
- Intercom pricing (opens in a new tab)
- Zendesk UK pricing (opens in a new tab)
- HubSpot, Customer Agent and Prospecting Agent: now you pay when the task is complete, 13 April 2026 (opens in a new tab)
- Freshdesk pricing (opens in a new tab)
Links checked 15 November 2026.