AI Knowledge

What AI actually costs in your business

Usage fees are rarely the biggest item. What really determines the bill.

As of 2026For decision-makers without technical background
In short

Setup, ongoing use, maintenance, and internal effort. The last one is almost always forgotten and is often the largest for small projects.

01

The four cost blocks

Setup, ongoing use, maintenance, and internal effort. The last one is almost always forgotten and is often the largest for small projects.

02

Usage fees

Language models are billed by the amount of text processed. For a chatbot with a few hundred conversations per month, that's in the low double digits — much lower than most people expect.

03

What actually costs

Preparing the content, defining the behavior, testing, and the first weeks of fine-tuning. That's work, not a fee, and it happens once.

04

Operation

After the initial phase, ongoing effort is low: occasional additions to the knowledge base, monitoring conversations. For a chatbot, that's a few hours per quarter.

05

The honest calculation

In a typical starter project, about eighty percent of costs go to one-time setup and twenty to two years of operation. Anyone only comparing monthly fees is comparing the smaller part.

06

Why a pilot project often seems more expensive than it is

An initial test run bears the entire setup cost alone, while ongoing costs only spread over many months. Anyone seeing just the first bill thinks the project is more expensive than it actually is over two years — that's why the payback calculation should be in the proposal from the start, not only when asked.

Questions

Frequently asked questions

What does a single language model call cost?

In the range of a few cents to low single-digit cents, depending on the model and amount of text. For a typical chatbot conversation it's a fraction of a cent — only volume makes it noticeable, not individual cases.

Why do offers vary so much in price?

Because they differ in how much goes into setup: How thoroughly is the knowledge base built, how many test rounds with real questions, how precisely is the response behavior defined. A cheap offer usually cuts corners here.

Is it worth running your own local model?

Not for most small businesses — the operational overhead of your own hardware far exceeds the savings on usage fees. It only becomes interesting with very high volume or data that cannot leave the premises.

How do you spot hidden follow-up costs in advance?

By asking about operation in the second and third year, not just the setup price. Anyone who can't or won't answer that hasn't done the math themselves.

Next step

Send us a few sentences about your project.

We read every inquiry ourselves and respond personally. No sales team, no callback form into the void.