How much a chatbot costs and what you pay for

What makes up the price

The actual work performed barely depends on which model you choose. The process goes as follows:

• Audit. We review incoming requests from recent months and calculate what share of them is standard. This takes two days and allows us to define a realistic timeline.
• Knowledge base. The most time-consuming part. Answers are compiled from your web pages, internal guidelines, and price lists; anything not written down is extracted through conversations with your team.
• Channel connection. Website widget, Telegram, or an application. One channel is included in the base package; each additional one is billed separately.
• Pilot. A week where the bot handles a fraction of the traffic while we review logs and fill in the gaps for things it didn't know.
• Post-launch adjustments. Two weeks are included in the package.

Why “from 5 000 ₴” is not the real price of a bot

For that amount, you can set up a widget and connect a model. From there, the bot starts answering based on whatever it finds online or invents on the spot. The very first mistake regarding price or delivery time will cost more than the money saved.

The difference lies not in the model, but in where the answer comes from. A bot without a knowledge base is just a confident-looking chat. A bot with a knowledge base answers strictly from your documents and cites the exact source.

What makes a bot more expensive

The price does not increase with the volume of requests, but with the number of places the bot needs to query:

• integration with CRM, billing or an accounting system — two-way sync, not read-only;
• each additional channel: Instagram, Viber, WhatsApp, custom app;
• a second language means translating the knowledge base, not just adding a toggle switch;
• voice channel: speech recognition and synthesis are a separate scope;
• populating the knowledge base on our end if you don't have time to write documentation.

All of this is listed in the add-ons table on the “Pricing” page — with rates and timelines, so nothing needs to be calculated verbally during meetings.

When a bot will not pay off

If there are few requests or every request is unique, there is nothing to automate. We calculate this during the audit and state it plainly: if the bot covers less than a third of inquiries, taking on the project makes no sense.

Another scenario is when a company lacks anyone who can answer questions about internal processes. In that case, there is no source material to build a knowledge base, and starting with a bot is the wrong move.