AI Insight Editor
Last updated
Last updated
The AI Insight Editor allows you to create queries using natural language. There is no need to handcraft any SQL, the AI editor will take care of it for you. Think of this as submitting a request your analyst.
KPI pulse: “Daily active users vs. yesterday” → instant line chart for stand-ups.
Growth analytics: “MRR by plan for the last six quarters” → stacked bar for board decks.
Cohort retention: “Week-1 retention for users who signed up in May” → retention curve without manual joins.
Content performance: “Top 10 landing pages by conversion last month” → sortable table ready for marketing.
Ad-hoc finance: “Average order value by region in Q1” → pivoted view for exec reviews.
In the AI Insight Editor you simply write a plain-English question in the large textbox. The editor will then translate the prompt into production ready SQL, executes the query against your warehouse with existing permissions and returns both the data table and an auto generated chart for you to refine or save.
As you type, the system peeks at your schema and surfaces inline event suggestions (e.g., email_opened, page_view) so you can click them for speed or ignore them entirely; the model will still map your words to the right tables and columns.
Once you are satisfied with your query you can then save the query. When you save the query the query will then be available for later use across the platform.
Similarly you can save the charts that have been generated.
Save changes
Saves the current chart so it is available to view against this query.
New
Create a new chart based on this query.
Add to
Add this chart to an existing report.
Folder
See the existing charts and edit them from this report.
Once generated you can also convert the AI generated query back into SQL. Simply click on the SQL tab and you will be given the option to switch or to convert the query back into SQL.
In most circumstances the default model will be sufficient for most use cases. However we do offer the ability to choose different models and fine tune their settings.
Temperature (0 – 2)
Governs randomness. Lower = more deterministic; higher = more creative.
Top P (0 – 1)
Top P trims off the least-likely words before the model chooses, so a lower value (e.g., 0.5) forces it to stick to the safest, most obvious options, while 1.0 lets it pick from the entire vocabulary.
Maximum Length
Hard cap on how many tokens the model can return (approx. 4 chars per token).
