We're considering rolling out the AI Assistant to more users, but as admins we've no real way of seeing how it's actually being used. Usage Tracking is great for logical and physical queries, but it doesn't tell us what questions people typed, which ones the Assistant couldn't answer, or where it picked the wrong column or subject area.
It would be really useful to have a new Usage Tracking table (and a matching subject area in FDI) that logs:
- the natural language question asked
- the dataset or subject area used
- whether a visual was returned, or it failed or fell back
- whether the user kept, edited, discarded the result
That would let us spot gaps in our indexing and synonyms, tidy up column names users don't recognise, and show real adoption stats to the business. Obviously this would need to respect data security and be admin-only.
Has anyone else been looking for this? Please let me know.
