Token usage
Total, input and output tokens over time, by platform and model. Cache reads are shown separately.
InputOutputCache
Understand how your organization actually uses AI
Usage, risk, and adoption analytics drawn from metadata-only telemetry, so leadership sees the real picture without ever reading employee content.
Example: a bar chart of AI token usage over the last seven days, highest midweek and low at the weekend. Hovering a day shows its tokens and cost.
Tokens, last 7 days
1.9M tokens
Why it matters
You pay for AI you can't see.
AI runs in browsers, IDEs, CLIs and desktop apps. It is paid per seat and per token. Without one view, nobody can say who uses what, or what it costs.
What you see
One view of AI usage and cost
Tokens, cost, people and runtimes, side by side. Built from metadata only.
Cost
Pay-as-you-go usage and subscription seats in one view, over time and by provider.
UsageSubscriptionsProviders
Top groups and users
See which groups and people use AI the most, and who spends the most.
GroupsUsersTop spenders
Runtimes
The browsers, IDEs, CLIs and desktop apps people use AI in, ranked by active users or by data sent to AI.
BrowserIDECLIDesktop app
Questions it answers
Answers to budget and adoption questions
Are we paying for seats nobody uses?
Idle seats are counted, along with what that idle spend adds up to.
Which teams spend the most?
Cost breaks down by group and by user, next to totals by provider.
Where do our AI rules matter most?
Runtimes are ranked by active users and data sent, so you see where AI is used most.
Do long conversations inflate the numbers?
No. Tokens re-sent from cache are shown separately and don't count toward your usage.
Cost
What AI costs, and what sits idle
Costs combine pay-as-you-go usage with subscription seats. See cost over time and by provider, the top groups and top spenders, and how many paid seats sit idle and what that idle spend adds up to.
Privacy
Metrics, not transcripts
Every number is built from metadata: tokens, models, runtimes, costs and policy outcomes. Prompts and responses never leave the device, so the data that answers budget and adoption questions never becomes a record of what people wrote.
Use cases
See it in action
Analytics, in detail
No. Analytics are built from metadata only. Prompts and responses never leave the device.
No. Tokens re-sent from cache are shown separately and don't count toward your usage.
Yes. Costs break down by group and by user, next to totals by provider and over time.
Say yes to AI.
Set the rules once. Alesia enforces them on every device, and prompts stay on your devices.