Usage: what you have consumed
The /utilisation page
Making an AI answer costs computation, and that computation is counted. This page tells you what you have consumed since the first of the month, and where the company's plan stands. This is the meter that will refuse a question the day it is full: a wall you don't see coming always feels like a breakdown.
What you have used, and what remains: nothing is hidden.
A dial, not a bar
The dial shows your share of what the company has spent. It is not a personal quota, nobody has one, the ceiling belongs to the company. It is the proportion that falls to you.
What counts, and what doesn't
| Consumes | Consumes nothing |
|---|---|
| Asking a question and getting the answer | Re-reading your old threads |
| Your scheduled tasks, even while you sleep | Searching your conversations |
| The AI colleague runs you launch | Looking at your images, filing a document |
Only making the model answer costs anything. Browsing, reading, tidying, filing: none of that touches the meter. Tom can explore the tool a whole afternoon without eating into anything.
The word "token"
It is the unit in which a model reads and writes. An English word is worth roughly one and a third tokens. What you send counts, and so does what it answers. You do not have to calculate anything: the figure is there so that the end of the month can be seen coming, not so that people censor themselves.
Whatever has no author, a connected product, a connector, a background job, is charged to the company and appears on nobody's screen. An administrator finds it under Consumption.
In closing
This counter is not there to slow you down but to avoid the invisible wall. A glance mid-month is enough: if the bar climbs too fast, we talk about it before you are blocked.
Commenting is not enabled on this course.