M↓ View as Markdown ↗

Models, Metrics and Glossary access¶

Each AI Analyst only works with the parts of your semantic layer you give it. Three tabs on the Analyst in Studio control this:

Tab	What it controls
Models	Which models the Analyst can query and, optionally, which dimensions and measures within them.
Metrics	Read-only. Shows the metrics the Analyst can use, which follow from its model access.
Glossary	Which glossary terms the Analyst can see.

Availability

Who can do this: members with the Edit capability in the active Workspace can view and change access. Members without it see the Overview, Models, Metrics and Glossary tabs greyed out.

Why configure access?¶

Not every AI Analyst needs all of your data. A narrower scope gives it fewer chances to pick the wrong model or field. It also enforces governance: the Analyst only sees and queries the models and fields you grant, and this is enforced when queries run, not only in the UI.

New Analysts start with no access

A new AI Analyst can't query any data until you grant it models: the Models tab starts in Choose specific models with nothing selected. It also sees no glossary terms: the Glossary tab starts in Only terms with these tags with no tags selected. A duplicated Analyst copies the model and glossary access of the Analyst it was copied from.

Models tab¶

Choose an access mode¶

At the top of the tab, the Access Mode card shows how this Analyst gets its models. To change it:

In Studio, open the AI Analyst and select the Models tab.
On the Access Mode card, click Edit.

Choose a mode:

Mode	What the Analyst can use
Choose specific models (default)	Only the models you tick. You can narrow each model to specific dimensions and measures. Selecting no models gives no access.
Models with any of these tags	Every model that carries at least one of the selected tags. A new model with a selected tag is included automatically. Selecting no tags gives no access.
All models except these tags	Every model except those carrying a selected tag. Selecting no tags gives access to all models.

In the two tag modes, pick one or more tags under Tags.
Click Save changes.

Tag modes stay up to date on their own. In Models with any of these tags, adding a selected tag to a model grants access right away, and removing it takes access away. In All models except these tags, adding an excluded tag to a model removes access, and removing it restores access. In tag modes the model list below the card is a read-only preview headed N models matched, and the Analyst gets every field of each matched model.

Grant specific models and fields¶

In Choose specific models mode, models are listed with a checkbox each. The counter above the list shows N of M selected, where M is the number of models in this Workspace (or your plan's models-per-analyst limit, if one applies). You can filter the list with Search models….

Tick the checkbox next to each model the Analyst may use. A newly ticked model is granted with all of its fields.
To narrow a model, click the arrow next to its name to expand it.
Under Dimensions and Measures, untick the fields the Analyst should not use, or use Select all / Deselect all.
Click Save changes.

The Analyst only sees and queries the granted fields. Pre-defined filters always follow their model and are not granted separately. Field narrowing is only available in Choose specific models mode.

Models-per-analyst limit¶

Your plan may cap how many models one AI Analyst can have in Choose specific models mode. When a cap applies, the M in N of M selected is that cap, and the counter turns red when you reach it. If you try to save more models than your plan allows, the save is rejected and the Models-per-analyst limit reached dialog appears with Upgrade and Contact us options.

No models in this Workspace?¶

If you see No models in this environment yet., the current Workspace has no models. Models belong to a Workspace, so models from other Workspaces never appear here. Create models first:

Go to Models in the Studio sidebar.
Create models from your connected data sources. See Creating Models.
Return to the AI Analyst to grant access.

Metrics tab¶

You can't grant metrics directly. As the tab says: "Metrics are derived from model access — granting a model also grants the metrics it needs."

An AI Analyst can use a metric only when it can access:

the metric's own model,
every model the metric joins to, and
every dimension and measure the metric's definition uses. If you narrow a model's fields in Choose specific models mode, metrics that use a removed field also become unavailable.

The Metrics tab is read-only. It lists the metrics the Analyst can use, grouped by model, with a Manage model access link that takes you to the Models tab. If nothing qualifies, it shows No metrics are available with the current model access.

Glossary tab¶

The Glossary tab controls which glossary terms the Analyst can find and use when it interprets questions.

Select the Glossary tab.

On the Access Mode card, click Edit and choose a mode:

Mode	Terms the Analyst can see
All terms	Every glossary term.
Only terms with these tags (default for new Analysts)	Only terms carrying a selected tag. Selecting no tags shows no terms.
All terms except these tags	Every term except those carrying a selected tag. Selecting no tags shows all terms.

In the two tag modes, pick tags under Tags.
Check the list under Analyst can see these terms (N). If no terms match, it shows No terms are visible with the current settings. If the Workspace has no glossary terms yet, the tab shows No glossary terms in this environment yet. instead.
Click Save changes.

To give a new Analyst full glossary visibility, choose All terms.

Note

Glossary access only controls which terms the Analyst can discover. It is not a data-access boundary: glossary terms describe your business language and don't grant access to data. Use the Models tab to control data access.

Saving changes¶

The Models and Glossary tabs save only when you click Save changes.

While you have unsaved edits, the footer shows Unsaved changes.
If you switch to another tab of the same Analyst with unsaved edits, you are asked Discard unsaved changes? Choose Discard changes or Keep editing. Other ways of leaving the page, such as using the Studio sidebar, don't show this prompt, so save first.
When you run a question in the Test your analyst panel, pending edits are saved first, so the test uses your latest configuration. If a save fails, for example because of the models-per-analyst limit, the question isn't sent and stays in the box; fix the error and try again. See Testing your AI Analyst.

You can also ask the Steward to change an AI Analyst's model or glossary access. By default the Steward asks for your approval before applying the change, but this depends on the Write AI Analyst tool permission in the Steward's settings. If the Models or Glossary tab is open, it switches to the new configuration, replacing any unsaved edits on that tab.

Best practices¶

Start with minimal access and expand as needed.
Use tag modes when you want access to follow your tagging, for example, all models tagged for finance. Use Choose specific models when you need field-level control.
For sensitive measures (for example, salary data), either:
- leave them out of general-purpose Analysts and create a separate AI Analyst that you deploy only to Specific users, or
- for stronger separation, keep the data in a separate Workspace that only authorized members can access, using a connection that isn't attached to other Workspaces. Creating Workspaces requires the Manage organization capability. See Workspaces.
Restricting who can use an AI Analyst in Explorer doesn't limit members with the Edit capability in Studio, or members of Slack or Teams channels where the Analyst is deployed.
Review access when you add models or tags to your semantic layer, especially for Analysts that use tag modes.