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

# AI ROI overview

> See what your AI coding tools cost, what your engineers ship with them, and where the money goes.

**AI ROI** answers one question: is the money you spend on AI coding tools turning into shipped work? It joins two things nOps already sees, your **AI coding spend** (Cursor, Claude Code, GitHub Copilot) and your **merged pull requests** (GitHub, GitLab, Bitbucket), and shows the result for the whole organization, for each team, and for each developer.

> [!NOTE]
>
> AI ROI is in preview and is not available to every organization yet. To turn it on, contact your nOps account team.

## What you get

> [card] [Executive Summary](https://help.nops.io/ai-roi/executive-summary.md)
>
> Three scores (Velocity, Productivity, Efficiency), how they move over time, and every team side by side.

> [card] [Engineering Overview](https://help.nops.io/ai-roi/engineering-overview.md)
>
> One team at a time: where its AI spend goes, its people, every merged PR and what it cost.

> [card] [Developers](https://help.nops.io/ai-roi/developers.md)
>
> A developer's own scores, why they are what they are, cost and quality by project, and how they use AI.

Each tab is built for one audience. You only see the tabs your role allows; see [Who sees what](https://help.nops.io/ai-roi/access.md).

## Find what you need

| I want to… | Read |
| - | - |
| Turn AI ROI on and connect the right tools | [Set up AI ROI](https://help.nops.io/ai-roi/setup.md) |
| Know what a score means and how to raise it | [Scores](https://help.nops.io/ai-roi/scores.md) |
| Understand a number: cost per PR, rework, output, "Written by AI" | [How the numbers work](https://help.nops.io/ai-roi/how-the-numbers-work.md) |
| Put people on teams, or tie a GitHub login to a person | [Teams and people](https://help.nops.io/ai-roi/teams.md) |
| Decide who can see whose numbers | [Who sees what](https://help.nops.io/ai-roi/access.md) |
| Get exact cost per PR and work types for every session | [AI ROI Agent](https://help.nops.io/ai-roi/agent/overview.md) |
| Find a quick answer | [AI ROI FAQ](https://help.nops.io/ai-roi/faq.md) |

## How it works

**1. nOps reads your AI coding spend**

Spend per person per day comes from your [Cursor](https://help.nops.io/integrations/cursor-admin-api.md) and [Claude Enterprise](https://help.nops.io/integrations/claude-enterprise.md) integrations. GitHub Copilot spend comes in through the [GitHub](https://help.nops.io/integrations/github.md) integration. Chat use of Claude and Bedrock API spend are not coding spend and are left out.

**2. nOps reads what your engineers merged**

Pull requests, their commits and the files they changed come from [GitHub](https://help.nops.io/integrations/github.md), [GitLab](https://help.nops.io/integrations/gitlab.md) or [Bitbucket Cloud](https://help.nops.io/integrations/bitbucket). nOps reads metadata only: it never stores file contents.

**3. nOps matches people to their code host accounts**

Each person's work email is matched to their GitHub, GitLab or Bitbucket login, mostly automatically. You can fix the rest by hand. See [Teams and people](https://help.nops.io/ai-roi/teams.md).

**4. Spend is tied to the work it produced**

Each merged PR gets an AI cost and a work type (Build, Fix, Understand, Docs, Ship). With the [AI ROI Agent](https://help.nops.io/ai-roi/agent/overview.md) installed, the cost is exact. Without it, nOps estimates it from the author's spend while the PR was in the making. See [How the numbers work](https://help.nops.io/ai-roi/how-the-numbers-work.md).

**5. The page rolls it up**

Scores, trends and tables for the organization, each team and each person, for the date range you pick.

## What you need

| | Required? | What it adds |
| - | - | - |
| An AI coding tool: Cursor, Claude Enterprise, or GitHub Copilot | **Required** | AI spend per person |
| A code host: GitHub, GitLab or Bitbucket Cloud | **Required** | Merged PRs, AI cost per PR, time to merge |
| Teams | Optional | The Engineering Overview, team scores and team comparisons |
| Linear or Jira | Optional | A work type for PRs whose title or branch doesn't name one |
| The [AI ROI Agent](https://help.nops.io/ai-roi/agent/overview.md) on developers' machines | Optional | Exact AI cost per PR, and a work type for spend that never reached a merged PR |

The [Setup tab](https://help.nops.io/ai-roi/setup.md) walks you through each one and tells you what is missing.

## Open AI ROI

1. Sign in to [nOps](https://clara.nops.io/dashboard).
2. In the sidebar, open **AI ROI** (below **AI Visibility**).
3. Pick a date range at the top of the page. AI ROI opens on **this quarter** to date. Besides the usual presets, it offers **This quarter**, **Last quarter** and **Last 4 quarters**, or any custom range of up to a year.

> [!TIP]
>
> Every "vs" figure on the page compares with the same number of days into the previous quarter (for example, the first 45 days of Q4 against the first 45 days of Q3). For a range that spans more than one quarter, it compares with the period of the same length just before.

## Your data and privacy

* **No prompts, no code.** AI ROI never reads prompt text, AI responses or file contents. Work types come from PR titles, branches, ticket labels and, with the agent, a label chosen on the developer's own machine. See [What the agent sends and keeps](https://help.nops.io/ai-roi/agent/privacy.md).
* **Small groups stay private.** Totals for a team with fewer than five people using AI are shown only to people who may already see that team's members.
* **Read only.** nOps never writes to your code host or your AI tools for AI ROI.

## Related

* [AI overview](https://help.nops.io/ai/introduction.md): the AI spend, seat and budget side of nOps.
* [AI Tracking](https://help.nops.io/ai/tracking.md): model and token usage across providers.


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