"How much does an AI chatbot cost?" comes up in almost every consultation - and almost always the expectation is a single number. There isn't one, just as there is no single price for a car. Instead of a number, we'll give you something more useful: the structure of a quote, the list of things that push it up, and a simple calculation of when the subscription pays for itself.
Why there is no single price
An AI chatbot isn't a "switch on and go" plugin. It's an assistant that knows your catalogue, checks availability, understands your delivery and returns rules and, sometimes, order statuses. The scope of that knowledge - and the number of systems it has to reach into - determines the price far more than the "AI engine" itself.
That's why a sensible quote always starts with a conversation about the process, not the technology. The same AI model can cost very different amounts depending on how much real work it takes over.
What goes into the quote
The cost splits into two parts - one-off and recurring:
- Implementation (one-off): process analysis, integration with your catalogue, stock levels and orders, "teaching" the assistant your knowledge (FAQs, policies, product specifics), testing and go-live.
- Subscription (monthly): the AI model and conversation costs, EU hosting, maintenance, monitoring of answer quality and ongoing updates to the assistant's knowledge.
With us, ecChat is billed as a predictable subscription - no paying "per message", which makes cost planning easier.
What really drives the cost up
- Number of integrations. The catalogue alone is one thing; stock levels, order statuses, ERP or courier systems are further layers of work.
- Languages. Multilingual support (e.g. a Polish store plus the UK market) increases the scope of testing and configuration.
- Complexity of the range. Technical products require precise knowledge - otherwise the assistant "guesses".
- GDPR requirements and hosting. Data in the EU, a data processing agreement, a retention policy - standard with us, but worth pricing in consciously.
When the subscription pays off
The simplest calculation: count how many enquiries reach your customer service team each month and how many of them are repetitive questions (availability, delivery, "does it fit…"). Those are exactly the questions the assistant takes over.
The "Cieplik" assistant handles around 819 conversations a month, resolving 88% of cases without a human agent. 43% of conversations take place outside working hours - when the customer service team wouldn't have answered anyway.
On top of the team's time savings comes sales: conversations outside working hours are orders that previously often slipped away. At that volume, the subscription usually pays for itself faster than the cost of one extra customer service hire.
FAQ
Do I pay per conversation?
With ecChat you pay a predictable subscription, not micro-payments per message - so the cost stays stable even during seasonal traffic spikes.
Will a chatbot replace my customer service team?
It doesn't replace them - it takes the load off. It handles repetitive questions (88% of cases at Ciepło24) and passes the harder ones to a human with the full context of the conversation.
How long does implementation take?
With a standard catalogue and typical integrations, usually 2-4 weeks - and most of that time goes not on technology but on gathering and organising the knowledge the assistant needs to have.