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

# AI governance

> Discover the AI models used across your organization and govern their usage, estimated cost, calling identities, and lifecycle risk.

AI Governance brings together the AI models used across your connected cloud and AI platforms. Explore which models are in use, how many tokens they consume, their estimated cost, and the people or applications calling them. Track upcoming model retirements to plan transitions and reduce the risk of disruption.

Open **AI Governance** from the main navigation to explore three tabs: **Token Usage**, **Models In Use**, and **Principals**.

## Token usage

The **Token Usage** tab shows token consumption over time, broken down into input, output, cache read, and cache write tokens. Use the chart to identify which models drive usage and when consumption changes.

Switch between **This Month** and **Last Month**, and filter by vendor to narrow the view.

## Models in use

The **Models In Use** tab lists discovered models with their usage, estimated cost, calling principals, and lifecycle status. Models with no usage in the selected period remain visible.

Select a model to review its attributes, lifecycle timeline, and token usage, including invocation counts.

## Lifecycle status

| Status           | Meaning                                                                                        |
| ---------------- | ---------------------------------------------------------------------------------------------- |
| Active           | The model is current and fully supported.                                                      |
| Legacy           | A newer version is available.                                                                  |
| Extended Support | The model is supported under an extended-support arrangement, which may carry additional cost. |
| EOL              | The model has reached End of Life.                                                             |

## Principals

The **Principals** tab shows the identities calling your models, along with their token consumption, models used, estimated cost, and last activity.

| Type             | Description                                                                                                                             |
| ---------------- | --------------------------------------------------------------------------------------------------------------------------------------- |
| User             | A human identity.                                                                                                                       |
| Service Account  | A non-human application identity.                                                                                                       |
| Managed Identity | An Azure managed identity.                                                                                                              |
| IAM Role         | An AWS IAM role.                                                                                                                        |
| API Key          | Usage authenticated with a shared API key rather than an individual identity. For Azure, Draftt displays the caller's masked source IP. |
| Unresolved       | An identity that could not be resolved to a readable name. Draftt displays the provider's object ID instead.                            |

Identity visibility depends on the provider and the access configured for your integration.

## Supported providers

Connect a supported provider to discover models and collect usage data.

| Provider         | Integration       | Coverage                                                                                                                                      |
| ---------------- | ----------------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
| AWS Bedrock      | [AWS](/aws)       | Model inventory, token usage, estimated cost, and calling identities.                                                                         |
| Azure AI Foundry | [Azure](/azure)   | Model inventory, token usage, and estimated cost. Identifying calling principals requires diagnostic logging and Microsoft Graph permissions. |
| OpenAI           | [OpenAI](/openai) | Model inventory, usage, and lifecycle risk.                                                                                                   |

Estimated costs, where available, are based on published model pricing.

## Identity attribution

Identity attribution links model usage to the person or workload making each call.

* **AWS Bedrock:** Calling identities are attributed automatically once AWS is connected.
* **Azure AI Foundry:** Token usage and estimated cost are collected through the integration. Calling identity attribution additionally requires diagnostic logging and Microsoft Graph permissions. See [Enable AI model identity attribution](/azure#enable-ai-model-identity-attribution-optional) in the Azure integration guide.

## Related policies

Draftt's lifecycle policies evaluate AI models against vendor end-of-life dates and version currency. You can also create custom policies using AI model attributes to apply your organization's standards.

See [Governance policies](/governance-policies).