Transparent prices. No surprises.
We build every solution bespoke, but we want you to have a clear idea of the investment before the first call.
So you know even before the call.
Guideline prices, not a fixed price list. The final price is always individual — based on scope, integrations and required capabilities.
- How it works WEB ASSISTANT
- How it works VOICE ASSISTANT
- How it works PHONE ASSISTANT
- How it works EMAIL ASSISTANT
- How it works LIVECHAT (NO ASSISTANT)
- How it works SOCIAL MEDIA
- How it works SELF-SERVICE CENTRE
- How it works GUIDED · CUSTOM
How the price is built: you pay once for building and deploying the assistant. After that it is a fixed monthly fee — running the language model is included in it and we keep it under limits, so it will not run away from you as your traffic grows. We agree the figure in advance, calculated on your own numbers.
Where the monthly figure comes from.
The monthly figure comes from how many questions you handle and how demanding they are. Here is a rough calculation — treat it as a ballpark, as it depends heavily on the model chosen. We will give you the exact figure after testing on your own scenarios.
Rough estimate only, rate ≈ 24 CZK/USD. We measure real usage per request and recalculate it in the ENIQ portal. Not an offer or a binding price.
How token usage is calculated (details)
Inference is billed separately for input and output tokens, always according to the specific model chosen. Input tokens include the user’s question, system instructions, conversation history and any context from documents. Output tokens are the model’s answer.
We assume a more capable open-source model — this type of assistant needs a larger context, better multilingual handling, reasoning and more stable answers than a smaller everyday model.
Indicative consumption per request: a simple question roughly 1,000–3,000 input + 300–800 output tokens; a normal question with documents roughly 4,000–10,000 input + 500–1,500 output; a more complex request roughly 10,000–25,000+ input + 1,000–3,000 output.
Request price = (input tokens / 1,000,000 × the model’s input price) + (output tokens / 1,000,000 × the model’s output price). Example: a model at USD 0.60/1M input and USD 3.60/1M output, a request of 6,000 input + 1,000 output ≈ USD 0.0072.
We will know the final figures after testing on real scenarios. Token measurement for every request is part of the rollout, so you will see consumption and cost calculated precisely in the ENIQ portal.
Simple economics, no magic.
An assistant starts at roughly 4,500 CZK and runs around the clock — outside office hours too, no question is left without a response.
From roughly 4,500 CZK + monthly by real usage. No payroll, sick days or overtime.
Team time spent on repetitive work — and inquiries that would otherwise go unanswered and leave for a competitor.
For most companies the cost pays off in time saved and inquiries captured. We'll calculate the exact impact on your numbers during the consultation.
Everything you need, in one flat rate
No hidden fees. Everything from the first consultation to ongoing tuning is included.
Free initial consultation
Analysis and solution design
Implementation and testing
Ongoing support and optimization
The risk is ours.
We guarantee what we genuinely control — the deadline, the quality of the answers and the fact that the assistant never invents one. We cannot honestly promise you revenue; that rests on your range, your prices and your traffic. What we can promise is this.
It does not go live until it passes your test
You put together a set of real questions from your own operation. Until the assistant handles them the way we agree on beforehand, it does not go near your customers — and you are not charged for unfinished work.
We launch on time, or the deployment is on us
You get the date in writing after the consultation. If the delay is on our side, you do not pay for the deployment.
We keep tuning free of charge until it fits
Live traffic always shows what the test did not catch. In the first month we watch the conversations daily and keep tuning — and tuning stays part of the service afterwards too; we never bill it as extra work. When the assistant has no basis for an answer, it hands the case to a person instead of inventing one.
If it does not hold the level, you get everything back
You can follow how the assistant is doing in your dashboard — it is measured on the questions it was trained for. If it does not hold the agreed level after the first month of live operation, we refund the full build fee and everything you paid for that first month. No haggling.
And no lock-in: you can end the service whenever you like. You take the knowledge base and the conversation history with you.
We'll give you a price based on your numbers.
In a free consultation we'll walk through your operations and prepare a concrete proposal with an indicative price — no obligation.
Book a free consultation