AI Token Usage Report
The AI Token Usage Report shows how many AI tokens your company has used over a date range, broken down by model and — for administrators — by user, along with an estimated US-dollar cost. It gives you one place to review AI spend across every AI feature in the desktop app.
Opening the report
The report opens as a tab in the desktop app. You can reach it two ways:
- From the Reports menu on the company workbench toolbar, choose AI Token Usage Report.
- From the AI Order Reader, open the Action drop-down and choose View AI Token Usage — handy for checking usage right after reading orders.
Choosing a date range
Pick a From and To date at the top of the tab and click Run. The report covers every AI call recorded between those dates (inclusive). It opens showing the last 30 days by default.
What the report shows
A summary line at the top gives the totals for the selected range:
- Calls — the number of AI requests made.
- Input, Output, and Total tokens — the tokens consumed.
- Estimated cost — the total estimated US-dollar cost.
Below the summary, a By model table breaks usage down per provider and model:
| Column | Meaning |
|---|---|
| Provider / Model | The AI provider (OpenAI or Anthropic) and model used. |
| Input / Output / Total | Tokens consumed by that model. |
| Calls | Number of requests that used that model. |
| Est. cost | Estimated US-dollar cost for that model over the range. |
Who can see what
The report automatically scopes to your role:
- All users can see their own AI usage — the totals and per-model table reflect only the calls you made.
- Administrators see usage for the whole company, plus an additional By user table showing how much each team member used, so you can see where AI spend is going.
Estimated cost
Costs are an estimate computed from published per-model rates and captured at the time each call was made, so historical figures stay accurate even if provider prices change later. The authoritative charge always comes from your AI provider's own billing — Backbone MRP uses a bring-your-own-account model and adds no charge of its own.