Skip to content
M↓ View as Markdown ↗

Saved Prompts

Saved prompts are reusable slash commands curated by your data team. Each one belongs to the current Workspace and is scoped either to one AI Analyst or to Global (all analysts). Users insert a saved prompt by typing / at the start of an empty message box, in Explorer or in the Studio Test your analyst panel, then picking it by slug or title.

Availability

  • Who can do this: members with the Edit capability in the active Workspace can create, edit and delete saved prompts. Anyone with the Chat capability in the Workspace can use them. Members without Edit see a View only badge next to the AI Analyst's name and can't save changes.

Info

Looking for how to use saved prompts as a member with the Chat capability? See Saved Prompts in Explorer.

Open the Saved prompts tab

  1. In Studio, go to AI Analysts and open the AI Analyst.
  2. Select the Saved prompts tab (next to Overview, Instructions, Models, Metrics, Glossary and Suggestions). If the window is narrow, it may be in the … menu at the end of the tab bar.

The tab lists the prompts scoped to this Analyst and all global prompts in the Workspace. Each row shows the /slug, a (Global (all analysts)) marker for global prompts, the title, and how often the prompt has been used (Used N times).

Add a saved prompt

  1. On the Saved prompts tab, click Add saved prompt.

  2. Fill in the form:

    Field What to enter
    Slug Required. The command name users type after /, for example weekly-revenue. The / is shown for you; if you type one, it's removed.
    Title Required. A descriptive name, for example "Weekly revenue by region". Users can also find the prompt by its title.
    Body The prompt text that is inserted into the message box. Can include placeholders.
    Global (all analysts) Optional. Tick to make the prompt available with every AI Analyst in this Workspace. Leave unticked to scope it to this Analyst.

  3. Click Save changes at the bottom of the tab. To discard the draft, click Cancel.

Info

Only the prompt is saved, not an answer. Answers go stale as your data changes. The value is the proven question that reliably produces a good result.

Slug rules

  • A global prompt's slug must not match any other prompt that is scoped to an Analyst or global in this Workspace.
  • An Analyst-scoped prompt's slug must differ from global prompts and from that Analyst's other prompts. Two different Analysts can each have their own /revenue.
  • If a slug is taken, the form shows This slug is already used by another prompt.
  • The slug of a deleted prompt can be reused.

Placeholders

Prompts can include placeholders written in double curly braces, such as {{region}} or {{timeframe}}. Placeholder names can contain letters, digits, _, - and ., and spaces inside the braces are allowed (for example {{ fiscal.year }}). Square brackets such as [region] are not placeholders.

Example:

Show me revenue by {{region}} for {{timeframe}}

While you type the Body, the form lists the placeholders it found under Placeholders. If there are none, it shows Add {{name}} tokens to the body to create fill-in placeholders.

When a user inserts the prompt with /, the first placeholder is selected so they can type over it, and Tab jumps to the next one.

Global and Analyst-scoped prompts

Scope Where users can insert it
Analyst-scoped In conversations with this AI Analyst only.
Global (all analysts) In every conversation in this Workspace, including a new conversation before an Analyst is chosen.

Saved prompts never appear in other Workspaces. The Steward has its own saved prompts, which are managed separately in the Steward.

Edit or delete saved prompts

  • Edit: click the pencil icon (Edit saved prompt), change the slug, title, body or Global (all analysts) setting, then click Save changes. Unticking Global (all analysts) scopes the prompt to the Analyst you're editing it from.
  • Delete: click the trash icon (Delete saved prompt) and confirm. This removes the prompt from every message box that can insert it.

Usage counts

The Used N times count on each prompt goes up each time a user sends a message that started from that prompt. Sends from the Test your analyst panel count too, so regression testing increases the count. The Saved prompts row on the Analyst's Overview tab shows how many prompts are scoped to this Analyst. Global prompts are not included in that count.

Using saved prompts for regression testing

When you iterate on the semantic layer or the Analyst's instructions, your saved prompts become a set of "golden queries": the canonical questions this AI Analyst must answer correctly.

After changing a model, metric or guidance, open the Test your analyst panel in Studio, type / at the start of an empty message box to insert each saved prompt, and check that the AI Analyst still produces the right results. See Testing your AI Analyst.