Understanding the AI quota

AI features come with a monthly included quota. Each AI action (chat, semantic search, indexing, summaries, participant detection from meeting transcripts) consumes a small, whole number of credits. The quota keeps AI usage predictable — it is deliberately generous.

Viewing your usage

Usage has a page of its own: Management → AI → Usage. On the AI area's overview, the Usage row links straight to it.

There, a bar shows the current month's usage (used vs. available credits, as a percentage). Below it, the usage history lists the recent months. Every month with usage can be unfolded via Details: it shows how many credits and how many calls each AI action accounted for, largest item first — so it is clear what the quota was spent on. The detail rows add up to exactly the month total above them.

Older entries may appear under Unattributed. Those are calls from before the action type was recorded; they count towards the quota as normal.

What happens when it runs out

The limit is a soft limit:

1. From 80% usage, a warning appears.
2. Once the monthly quota is used up, AI keeps running on a grace buffer for a while (a banner points this out). Answers in progress are never cut off mid-stream.
3. Once the buffer is used up too, the limit takes effect. You then have two options:
   • add your own API key (Using your own AI API keys) — AI then runs without the shared quota, or
   • wait for the monthly reset — the quota refills at the start of the month.

Transparency

For each AI action, we record who triggered it, when, and how many credits it cost. This overview is visible to administrators of your organisation and makes the quota transparent.