AI Usage
AI Usage (Overview → AI Usage) shows how much AI your clients use across the whole platform, which models and features they use, and what it costs you.
The problem it solves
Every client shares the one AI key you saved under AI Provider Setup, so you pay the AI bill. If one client uses far more than their plan expects, you lose money. This page shows you who it is before the bill arrives.
What you see
Use the From and To dates to choose the period.
| Card | What it means |
|---|---|
| Tokens | Total tokens used, split into "in" (what was sent to the AI) and "out" (what it wrote back). A token is a small piece of a word. |
| AI runs | How many times the AI was used: answers, drafts and indexing of documents. |
| Estimated cost | What those runs cost, worked out from the usage log. |
| Clients using AI | How many clients used AI in the period, out of all clients. |
| Knowledge bases | How many knowledge bases exist across all clients. |
| Documents indexed | How many documents are ready to answer from, out of all uploaded. |
Below the cards you will find Tokens per day, By model, By operation (what the AI was used for) and Top clients by cost.
How to use it
- Open Top clients by cost. Is one client far above the rest?
- Open that client's page and compare their usage with their plan's Max AI cost / month limit.
- If the limit is too high or missing, change the plan. See Subscription Plans.
When a client reaches the plan's monthly AI cap, the AI stops answering for them until the next month, and conversations go to a human instead.
Costs are estimates from the price list built into Threadline. They are not your provider's invoice. Speed and error rate are not shown because the usage log does not record them.
Only staff with View AI usage across clients can open this page. Each client sees only their own use, under Reports → Usage Report in their workspace.