> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nebuly.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Spend

**AI Spend** shows how much your organization spends on AI and what that spend is used for. Instead of looking only at tokens, Nebuly measures spend at the **task level**, connecting AI costs to the work users executed.

You can track spend over time and break it down by **task, model, team, market, or tenure** to understand where your AI budget goes and where costs can be optimized.

## Why measure spend per task

Token usage explains how much model capacity was consumed, but it is difficult to use as a measure of business activity.

A **task** groups the interactions required to complete a piece of work. Measuring spend at this level gives you a clearer view of what your organization is paying to accomplish.

<img src="https://mintcdn.com/nebulyai/tyxr6Gd4zuVUslkl/images/image-54.png?fit=max&auto=format&n=tyxr6Gd4zuVUslkl&q=85&s=d75bb78044efb08a94092743356180eb" alt="Image" width="1375" height="356" data-path="images/image-54.png" />

## Total AI spend

**Total AI spend** shows how much your organization spent on AI during the selected period.

Click **Breakdown** to see how that spend is distributed across groups such as **Team, Market, or Tenure**. This helps you identify which parts of the organization account for the largest share of AI spend.

## Average AI spend per task

**Avg. AI spend per task** shows the average cost of completing a task across the AI models and providers used by your organization.

The distribution chart compares spend per task across teams and provides two reference points:

* **Organization average** shows the average spend per task across the organization.
* **Top quartile** shows the spend per task reached by the most cost-efficient 25% of teams.

Use **Distribution** to see the full breakdown and identify the teams whose task costs differ significantly from the rest of the organization.

A higher spend per task is not necessarily inefficient on its own. Some teams may perform more complex or compute-intensive work, so this metric is most useful when considered alongside **task type, task volume, and error rate**.

## AI spend over time

The **AI spend over time** chart shows how costs change across the selected period.

You can switch between two views:

* **Total AI spend** shows the amount spent during each period.
* **AI spend per task** shows how the average cost of completing a task changes over time.

In the spend-per-task view, the range around the line shows the difference between the teams with the lowest and highest spend per task. This helps you see whether costs are becoming more consistent across the organization or whether large differences remain.

## AI spend distribution

The **AI spend distribution** section shows how your spend is distributed across a selected dimension.

Use the dropdown to group spend by **Task, Model, Team, Market, or Tenure**.

<img src="https://mintcdn.com/nebulyai/73dNoiWDlX9XFjz2/images/Screenshot-2026-09-23-at-15.15.35.png?fit=max&auto=format&n=73dNoiWDlX9XFjz2&q=85&s=54674020aeffdda3fea3907d262551a6" alt="Screenshot 2026 09 23 At 15 15 35" width="1391" height="494" data-path="images/Screenshot-2026-09-23-at-15.15.35.png" />

The distribution bar shows each item's share of total AI spend. The table provides the underlying metrics:

<table>
  <colgroup>
    <col width="119" />

    <col width="569" />
  </colgroup>

  <thead>
    <tr>
      <th>Metric</th>
      <th>What it shows</th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td>**Total spend**</td>
      <td>Total AI cost for the item and its share of overall spend</td>
    </tr>

    <tr>
      <td>**Spend / task**</td>
      <td>Average AI cost of executing one task</td>
    </tr>

    <tr>
      <td>**# Tasks**</td>
      <td>Number of tasks completed</td>
    </tr>

    <tr>
      <td>**Users**</td>
      <td>Number of users associated with those tasks</td>
    </tr>

    <tr>
      <td>**Error rate**</td>
      <td>Share of tasks that resulted in an error</td>
    </tr>
  </tbody>
</table>

Click a column header to sort the table.

These metrics are most useful when read together. **High total spend with low spend per task** can indicate a frequently used task with relatively low unit cost. **High spend per task combined with a high error rate** can point to an area worth investigating.

## Team view

Select **Team** from the left sidebar to view AI Spend grouped by team.

From there, click a team to open its dedicated **Team** view. The team selector at the top lets you move between teams and compare their total spend. Use **Map to tags** to define which user tags correspond to each team, **in case you're not using the Team attribute**.

The **Team avg. AI spend per task** view places teams on the same spend-per-task scale. It shows the selected team's position alongside the **organization average**, making it easier to compare task costs across teams.

The rest of the page uses the same metrics as the organization-level view, including **Total AI spend**, **Avg. AI spend per task**, and **AI spend over time**, filtered to the selected team.

<img src="https://mintcdn.com/nebulyai/73dNoiWDlX9XFjz2/images/Screenshot-2026-09-23-at-15.20.21-1.png?fit=max&auto=format&n=73dNoiWDlX9XFjz2&q=85&s=22394bac5215c35616f23dda63bfc44e" alt="Screenshot 2026 09 23 At 15 20 21" width="1375" height="356" data-path="images/Screenshot-2026-09-23-at-15.20.21-1.png" />

## Build your own reports

AI Spend metrics are also available in Nebuly's reporting engine.

You can combine **AI spend**, **spend per task**, **task volume**, **model**, and user attributes with your existing filters and groupings to investigate specific cost patterns.

For example, you can build reports to understand:

* Which tasks account for the largest share of AI spend
* How spend per task differs across teams or models
* Which models are used for the most expensive tasks
* How AI costs change over time
* Where high spend per task coincides with high error rates


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.