AI data analysis / narrative insights inside Documents

Nobody has claimed this yet.

**What problem will this feature request solve?**
Insights that explain the charts and queries embedded in a Document have to be written entirely by hand. There's no way to have AI read the embedded data and produce a written analysis.

**Describe the solution you'd like.**
Have AI / Metabot read the charts, queries, and data included in a Document and generate insights, trends, highlights, and narrative explanations as written text — automatically, and ideally refreshable as the data changes.

**How does this feature request impact you?**
It would be additional value the customer wants to deliver to their clients on top of manually edited documents.

**Additional information**
- Related: [#76089](https://github.com/metabase/metabase/issues/76089) — *AI generated dashboard summaries* (open). Same idea applied to dashboards; AI-written summary of key insights, refreshable, includable in subscriptions.
- Related: [#16012](https://github.com/metabase/metabase/issues/16012) — *Using natural-language generation to add data stories to Metabase Dashboards* (open). NLG paragraphs tied to query values that describe the data.
- Related: [#74304](https://github.com/metabase/metabase/issues/74304) — *Create automated (non-dashboard) reports using Metabot/AI* (open).

Start by reviewing how Documents currently contain charts, queries, and embedded data, then compare the related issues #76089, #16012, and #74304. Define the supported AI inputs, narrative output, and refresh behavior before identifying implementation entry points. Done means a scoped design and implementation that generates and refreshes written insights in Documents.

Written by the indexing model from the issue text.