> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://help.sigmacomputing.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://help.sigmacomputing.com/_mcp/server.

# Configure AI features for your organization

> Configure AI features for a Sigma organization to accelerate insights, automate tasks, and enable Assistant, agents, and more.

Several features in Sigma use AI to accelerate insights and tasks that you might perform in Sigma:

* **Sigma Assistant** is a natural language query (NLQ) interface that allows you to ask questions about your data and interact with AI-generated responses. See [Ask natural language queries with Sigma Assistant](/docs/ask-natural-language-queries-with-assistant).
* **Sigma agents** let you build agentic solutions in Sigma workbooks capable of analyzing data, calling warehouse agents, triggering action sequences, and more. See [About Sigma agents](/docs/sigma-agents).
* The **formula assistant** uses AI to write new formulas, correct formula errors, and explain existing formulas. See [Use AI with formulas](/docs/use-ai-with-formulas).
* **Explain this chart** uses AI to instantly generate a description of any chart. See [Explain charts with AI](/docs/explain-visualizations-with-ai).

The use of AI features is subject to the following [disclaimer](/docs/notice-for-enabling-ai-enabled-features-in-sigma).

To use AI-enabled features in Sigma, do the following:

1. [Set up an AI provider](#set-up-an-ai-provider)
2. [Manage access to AI features](#manage-access-to-ai-features)
3. If Sigma Assistant is enabled, [configure data sources](#configure-data-sources-for-sigma-assistant)
4. (Optional) [Review data access for AI features](#review-data-access-for-ai-features)

## User requirements

You must be assigned the **Admin** [account type](/docs/create-and-manage-account-types).

## Set up an AI provider

When you set up an AI provider in Sigma, you connect Sigma to either an external AI service or a data platform with access to specific AI models that Sigma can use. The models that Sigma uses for AI-powered features depend on your AI provider. You cannot choose which model Sigma uses, and Sigma might update the supported models over time.

* To connect to an AI model hosted by your data platform, see [Configure a warehouse-hosted model as AI provider](/docs/configure-warehouse-ai-model-integration). When you choose the option to configure a warehouse-hosted model, your data is managed fully in your data platform.

* To connect to an externally hosted model, see [Configure an external AI provider](/docs/manage-external-ai-integrations).

Sigma uses a reasoning model for AI-powered features. The following table lists the reasoning model that Sigma currently uses for each provider. When a model is scheduled to be deprecated or replaced, the relevant model and date are also listed:

<table>
  <thead>
    <tr>
      <th>
        AI provider
      </th>

      <th>
        Model status
      </th>

      <th>
        Model
      </th>

      <th>
        Date
      </th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td rowspan="3">
        [BigQuery](/docs/configure-warehouse-ai-model-integration#bigquery-requirements)

         

        <br />

        (warehouse-hosted)
      </td>

      <td>
        Current
      </td>

      <td>
        Gemini 2.5 Pro
      </td>

      <td />
    </tr>

    <tr>
      <td>
        Deprecated
      </td>

      <td />

      <td />
    </tr>

    <tr>
      <td>
        Planned
      </td>

      <td>
        Gemini 3.5 Flash
      </td>

      <td />
    </tr>

    <tr>
      <td rowspan="3">
        [Databricks](/docs/configure-warehouse-ai-model-integration#databricks-requirements)

         

        <br />

        (warehouse-hosted)
      </td>

      <td>
        Current
      </td>

      <td>
        Claude Sonnet 4.5
      </td>

      <td />
    </tr>

    <tr>
      <td>
        Deprecated
      </td>

      <td />

      <td />
    </tr>

    <tr>
      <td>
        Planned
      </td>

      <td>
        Claude Sonnet 4.6
      </td>

      <td />
    </tr>

    <tr>
      <td rowspan="3">
        [Snowflake](/docs/configure-warehouse-ai-model-integration#snowflake-requirements)

         

        <br />

        (warehouse-hosted)
      </td>

      <td>
        Current
      </td>

      <td>
        Claude Sonnet 4.5 (AWS-hosted accounts) 

        <br />

         GPT-5.1 (Azure-hosted accounts)
      </td>

      <td />
    </tr>

    <tr>
      <td>
        Deprecated
      </td>

      <td>
        Claude Sonnet 4 (AWS-hosted accounts)
      </td>

      <td>
        April 3, 2026
      </td>
    </tr>

    <tr>
      <td>
        Planned
      </td>

      <td>
        Claude Sonnet 4.6 (AWS-hosted accounts)
      </td>

      <td />
    </tr>

    <tr>
      <td rowspan="3">
        [OpenAI](/docs/manage-external-ai-integrations#add-openai-as-an-ai-provider)

         

        <br />

        (external)
      </td>

      <td>
        Current
      </td>

      <td>
        GPT-5.4

        <br />

        GPT-5.1
      </td>

      <td />
    </tr>

    <tr>
      <td>
        Deprecated
      </td>

      <td>
        GPT-4.x
      </td>

      <td>
        February 2, 2026
      </td>
    </tr>

    <tr>
      <td>
        Planned
      </td>

      <td>
        GPT-5.6
      </td>

      <td />
    </tr>

    <tr>
      <td rowspan="3">
        [Gemini](/docs/manage-external-ai-integrations#add-google-gemini-as-an-ai-provider)

         

        <br />

        (external)
      </td>

      <td>
        Current
      </td>

      <td>
        Gemini 2.5 Pro
      </td>

      <td />
    </tr>

    <tr>
      <td>
        Deprecated
      </td>

      <td />

      <td />
    </tr>

    <tr>
      <td>
        Planned
      </td>

      <td>
        Gemini 3.5 Flash
      </td>

      <td />
    </tr>

    <tr>
      <td rowspan="3">
        [Anthropic](/docs/manage-external-ai-integrations#add-anthropic-as-an-ai-provider)

         

        <br />

        (external)
      </td>

      <td>
        Current
      </td>

      <td>
        Claude Sonnet 4.5
      </td>

      <td />
    </tr>

    <tr>
      <td>
        Deprecated
      </td>

      <td />

      <td />
    </tr>

    <tr>
      <td>
        Planned
      </td>

      <td />

      <td />
    </tr>

    <tr>
      <td rowspan="3">
        [Amazon Bedrock](/docs/manage-external-ai-integrations#add-amazon-bedrock-as-an-ai-provider)

         

        <br />

        (external)
      </td>

      <td>
        Current
      </td>

      <td>
        Claude Sonnet 4.5
      </td>

      <td />
    </tr>

    <tr>
      <td>
        Deprecated
      </td>

      <td />

      <td />
    </tr>

    <tr>
      <td>
        Planned
      </td>

      <td>
        Claude Sonnet 4.6

        <br />

        Claude Haiku 4.5
      </td>

      <td />
    </tr>
  </tbody>
</table>

If you use [Azure OpenAI](/docs/manage-external-ai-integrations#add-azure-openai-as-an-ai-provider), you can choose which model to use. Sigma recommends using the same model listed as **Current** for OpenAI.

Sigma uses an embeddings model to create embeddings of your documents for [semantic search](/docs/search-in-sigma). The following table lists the embeddings model that Sigma currently uses for each provider:

| AI provider                                                                                             | Currently supported embeddings model |
| :------------------------------------------------------------------------------------------------------ | :----------------------------------- |
| [BigQuery](/docs/configure-warehouse-ai-model-integration#bigquery-requirements) (warehouse-hosted)     | `gemini-embedding-001`               |
| [Databricks](/docs/configure-warehouse-ai-model-integration#databricks-requirements) (warehouse-hosted) | `GTE v1.5`                           |
| [Snowflake](/docs/configure-warehouse-ai-model-integration#snowflake-requirements) (warehouse-hosted)   | `snowflake-arctic-embed-l-v2.0`      |
| [OpenAI](/docs/manage-external-ai-integrations#add-openai-as-an-ai-provider) (external)                 | `text-embedding-3-small`             |
| [Azure OpenAI](/docs/manage-external-ai-integrations#add-azure-openai-as-an-ai-provider) (external)     | `text-embedding-3-small`             |
| [Gemini](/docs/manage-external-ai-integrations#add-google-gemini-as-an-ai-provider) (external)          | `gemini-embedding-001`               |
| [Amazon Bedrock](/docs/manage-external-ai-integrations#add-amazon-bedrock-as-an-ai-provider) (external) | Amazon Titan Text Embeddings         |

If you use [Anthropic](/docs/manage-external-ai-integrations#add-anthropic-as-an-ai-provider) as your AI provider, configure a separate **Embeddings provider** in **Administration** > **AI settings**. Select another provider from this table. Anthropic does not provide an embeddings model that Sigma can use, and you cannot bring your own embeddings model.

When you update the embeddings provider, Sigma starts a reindex of searchable sources in the background.

## Manage access to AI features

You can control who can access different AI features by enabling feature permissions on different account types assigned to users. See [Configure permissions for AI features](#configure-permissions-for-ai-features).

### Configure permissions for AI features

To control which Sigma users can access specific AI-powered features, enable or disable the relevant permissions for specific account types:

| AI feature                   | Account type permission that enables use |
| :--------------------------- | :--------------------------------------- |
| Sigma Assistant <sup>1</sup> | Use Assistant                            |
| Formula assistant            | Use AI with formulas                     |
| Explain this chart           | Explain charts with AI                   |

<p class="footnote">
  1

  Depending on what you want to do with Sigma Assistant, the ability to use it can require additional account type permissions and document access. For example, to use Assistant to build a dashboard, a user must be able to edit a workbook draft, which requires the **Create, edit, and publish workbooks** permission and **Can edit** workbook access.
</p>

For more details, including which license tiers have access to these permissions, see [License and account type overview](/docs/account-type-and-license-overview).

## Configure data sources for Sigma Assistant

When you configure a data source for Sigma Assistant, you make it available by default for conversations with Assistant. Users can also choose specific data sources that they have access to when asking natural language queries, even if those sources are not configured.

Select data models to take advantage of semantic modeling, relationships, and metrics. You can also select Snowflake semantic views, tables from connected data platforms, or datasets (deprecated).

To configure data sources for Assistant:

1. Go to **Administration > AI settings**:
   1. In the Sigma header, click your user avatar to open the user menu.
   2. Select **Administration** to open the Administration portal.
   3. In the side panel, select **AI settings**.

2. Select the **Assistant** tab.

3. In the **Configured sources** section, click **Add source**.

4. In the **Add configured sources** modal, search for and select one or more data models, tables, Snowflake semantic views, or datasets.

   When you select a data model as a configured source, only the published version of the data model is selected. You cannot select the tagged version of a data model.

5. Click **Save** to configure the selected sources for Assistant. The configured data sources are automatically synced and added to the semantic search index used by Assistant to select data sources.

If you change data sources or metadata in your data platform after you configure those data sources, updates are available to Assistant within 2 hours. If you want the updates to be available immediately, return to this page and select **Manual sync** to update the representation of your data sources in the semantic search index.

## Review data access for AI features

AI features such as Sigma Assistant respect the access granted to users for data sources. Each user can see and query only the data sources that they have access to. If you want to restrict what Assistant uses as a data source for answers for a specific user or team, restrict user or team access to the data source at the data source, schema, or table level. By default, Assistant only answers questions using pre-configured data sources. To configure those data sources, follow the steps to [configure data sources](#configure-data-sources-for-sigma-assistant).

To manage user access to data, review the access set on a specific connection or data model:

* [Manage access to data and connections](/docs/manage-data-permissions).
* [Manage document access](/docs/folder-and-document-permissions) for data models and datasets.
* Within a data model, you can also control visibility of a source: [Enable or disable an element as a data source](/docs/create-and-manage-data-models#enable-or-disable-an-element-as-a-data-source).

### How Sigma Assistant selects data sources when answering user questions

When using Sigma Assistant, users can choose a specific data source that they have access to, or Assistant can choose a configured data source to use in the conversation.

If a source is not specified, Assistant selects a source from the configured data sources based on the following factors:

* **The user's access to the data source**: Assistant respects the access granted to the user asking the question and does not expose data they are not authorized to use.
* **The relevance of the data to the question**: Assistant uses any metadata that exists to describe data sources and data columns, such as data model descriptions or column descriptions. If no other metadata has been defined, Assistant can infer relevancy from column names.
* **The known validity of the data**: For example, if a data model has an "endorsed" badge applied, Assistant is more likely to use that data source. Similarly, if someone in your organization has created and shared metrics and data models, Assistant is more likely to select one of those when answering rather than using a table directly from your connected data platform.
* **The frequency of usage of the data source**: Assistant is more likely to reference frequently used data sources in its answers.

Adding metadata to your data sources for Assistant is not required, but it is recommended.

Assistant uses any available metadata, such as certification badges, source descriptions, table descriptions, and column descriptions, to determine relevancy to a user's question, but can also determine relevancy using column names alone.

To add metadata to data sources, see the following:

* Add badges and descriptions to connection tables. See [Annotate tables with metadata](/docs/manage-data-catalog#annotate-tables-with-metadata).
* Add badges and descriptions to data models. See [Edit a data model description](/docs/create-and-manage-data-models#edit-a-data-model-description) and [View and manage document badges](/docs/view-and-manage-document-badges).

## Related resources

* [Ask natural language queries with Sigma Assistant](/docs/ask-natural-language-queries-with-assistant)
* [Use AI with formulas](/docs/use-ai-with-formulas)
* [Explain charts with AI](/docs/explain-visualizations-with-ai)
* [Configure an external AI provider](/docs/manage-external-ai-integrations)
* [Notice for enabling AI-enabled features in Sigma](/docs/notice-for-enabling-ai-enabled-features-in-sigma)
* [Configure a warehouse-hosted model as AI provider](/docs/configure-warehouse-ai-model-integration)