# AI Usage Report — {team_name}

**Period:** {start_date} → {end_date} ({n_days} days)
**Users in scope:** {n_users}
**Total invocations:** {n_invocations}
**Distinct tools / skills observed:** {n_distinct_tools}

> {data_caveats_line}

## Per-user breakdown

| User | Invocations | Sessions | Avg / session | Distinct tools | Effectiveness rate | Single-shot rate |
|---|---:|---:|---:|---:|---:|---:|
{per_user_rows}

**Reading the table:**
- *Effectiveness rate* — % of this user's invocations whose output was consumed in a shipped artifact. `—` means the input lacked the `output_consumed` field.
- *Single-shot rate* — % of this user's sessions with exactly one invocation. `—` means the input lacked `session_id`.
- *High single-shot + low effectiveness* is a starting place to ask about engagement — not a verdict.

## Top tools / skills used (across team)

| Tool | Invocations | % of total |
|---|---:|---:|
{top_tools_rows}

## Where signals are missing

{missing_signals_section}

## How to read this report

This report shows **patterns**, not verdicts. A user with a low effectiveness rate may be:

- Exploring (high volume of throwaway prompts while learning a tool)
- Solving in one shot (one prompt is enough — no follow-up needed)
- Doing AI theater (invoking to be seen invoking)

The data alone cannot distinguish these. Use the report to surface conversations, not to grade individuals.

---
*Generated by `ai-usage-report` on {generated_at}. Input file: `{input_file}`. View: `per-user`.*