People ask this first, and they are right to. An AI agent is a program with one job. It reads and decides one small thing.
Then it waits for a person before anything leaves the building. The model inside it is the cheap part.
What sets the price
The price follows the workflow around the agent. Where does the work come in? Which systems does it touch? Who approves it, and on what screen?
A clean workflow with 1 input and 1 approval is a small build. A workflow that reads scanned pages across 4 years of records is a large one.
- A clean workflow
- 1 input and 1 approval
- Scanned pages to read
- 4 years of records
On a quoting engine we built for a precision shop in North America, the agent itself was the smaller part.
The work was reading 332 pages of the shop's old quotes and its own registers, so its figures came from the shop's own history.
How we price it
It starts with a 30 minute call. We listen to the workflow and say whether a system fits. You get a fixed price within a day, everything included.
Does a system fit?
Everything included
Priced once
Runs and improves it
Builds start at $500. We price once, after the call, so the number rests on your workflow and not on a guess.
What it costs to run
Running cost is the model calls plus the server, and both scale with volume. We log the cost of every run from day 1, so you see the bill per run first.
Where it runs
It runs on servers you own, in your accounts. If we part ways, the system stays with you and keeps working. You never pay to move it out.