← All situations
Situations
Costing an AI project budget you can defend
You have to give a number before you have all the facts, and that number will follow you. Better to build it so you can explain it line by line.
What we observe
- Quotes range from one to five times each other with no scope difference to explain it
- Monthly running cost appears nowhere
- Human supervision time is costed in none of the proposals
- Someone promises a three-year return, with no intermediate checkpoint
What usually blocks
An AI budget reads on two lines you must never merge: setup is paid once, run comes back every month for as long as the system runs. A quote that does not separate them usually hides the second.
The forgotten line is almost always the same. Human review time belongs to running cost, and on a supervised agent it often weighs more than tokens.
What we do
- 01We cost setup and run separately, with the assumptions written next to them
- 02We measure real consumption over two weeks rather than estimating it
- 03We add supervision time to running cost, not to good intentions
- 04We state the payback period over twelve months, with a checkpoint at three
The method in full
The reference guide on the subject, with its definition, its sources and its guardrails.
Scope your first AI agent
20 minutes to review your tools, data and the first useful case. No jargon, no commitment.