What moves the price up or down?
- Connections: which systems the agent reads from or changes, and whether those systems give clean access.
- Data: its format, quality and volume, and your permission to use it.
- Controls: access permissions, the points where a person must approve, logging, and any security review.
- Acceptance: the tasks the system must do, how it handles failure, and the evidence you need before it goes live.
A narrow job with one clean connection costs less. A job that touches several systems, needs approvals and must pass a security review costs more.
What does it cost to run once it is built?
Ask for each running cost to be listed: hosting, charges from the AI provider for each use, outside licences, monitoring, maintenance, and the human review that remains. Compare the estimate at your expected volume and at a higher one. Agree who watches usage and who approves extra spending.
Support after handover is quoted with the project.
How do you estimate payback?
Compare the total build and running cost with the net monthly benefit: the hours saved or the revenue added, minus the human work that remains. Treat the result as a scenario until real operating data confirms it. No one can honestly guarantee a payback period before the system has run on your work.
What should a quote include?
A useful quote ties the price to one defined workflow, the systems it connects to and a clear acceptance test. It should list what is excluded, the outside costs, the payment milestones, and what you will own at handover. Our guide on build vs buy for AI agents shows how to compare that quote with a product's three-year cost.
How do you get a quote?
Describe the task, how it runs today and the systems involved, and email it to hello@kin-etiq.ai. Keep sensitive records out of the first message. The custom AI development page explains how a build runs, and the Staffy case study shows the kind of regulated work our founder knows. Staffy is a related company with the same founder.