AI usage
Every run leaves one recorded row: which workspace, who, how much.
No spend without a row
Everything a language model does — reading a document, an assistant answer, transcribing a meeting, a weekly summary, a question about a meeting transcript — writes one row into the usage ledger.
Each row says which workspace it belongs to and who spent it. Spend that is not attributed to a workspace is spend nobody can ever explain.
Who can see the usage report depends on role.
How to keep usage down
The biggest saving is not uploading the same document twice. Retrying a document that lost pages only re-runs those pages.
Pasted text costs nothing; if you already have the text, bring it in through the text door.
Meeting segments that have been transcribed once are never transcribed again, even if the meeting run resumes.
Read this next
- Account & plansPlans and modulesWhat is in which plan, and what it looks like when something is not.
- DocumentsWhen a page fails to readTotal loss, a few missing pages, and the difference between the two kinds of retry.
- DocumentsPaste text instead of a scanWhen you already have the text, nothing needs to be pulled back out of an image.
Still stuck?
If the answer wasn't here, write to us. It is a real inbox and it gets answered.