Employee AI usage metrics
Employee AI usage metrics, one prompt at a time.
An employee AI usage report answers who used an agent, which agent it was, and how many tokens that prompt consumed. It is the view a company asks for when several coding agents are already in use and the invoices do not name the person.
What managers can total
- Tokens per employee, summed from that person’s usage rows.
- Tokens per agent, so Claude Code and Cursor are visible side by side.
- Turns, counted as one row per completed prompt.
- Cache share, from cached input tokens beside raw input tokens.
A later prompt does not replace an earlier one. Each reply keeps its own counts. If a prompt never arrives, that turn is missing and the other turns remain.
What this report leaves out
The employee row does not need the prompt text, the source files, or the model’s reply. Those stay with the agent. The company record is the identity, the agent name, the time, and the token counts. That is enough to see adoption and spend, and it is the boundary the demo is built around.