The four prompts every AI Insights user should know
Most business questions reduce to one of four shapes. Master these and you can answer 90% of what your team will ever ask.
1. Top N
"Top 10 customers by revenue last quarter"
What it does: ranks one dimension by an aggregate of one measure, with an optional time filter. The agent will SUM, GROUP BY, ORDER BY, LIMIT.
Variations:
- "Bottom 5 products by margin this year"
- "Top 20 referrers by signups last week"
2. Time-series trend
"Daily active users over the last 90 days"
What it does: groups one measure by a time bucket. The agent picks the right bucket (day, week, month) based on the time window.
Variations:
- "Monthly recurring revenue this year"
- "Hourly latency over the last 24 hours"
3. Comparison
"Compare orders this week vs last week"
What it does: two parallel aggregates with a side-by-side. The agent picks the right comparison method (chart or table) based on the dimension cardinality.
Variations:
- "Revenue by region this quarter vs last quarter"
- "Conversion rate by channel: A/B test variant A vs B"
4. Anomaly
"Which products had unusual sales last month?"
What it does: computes a baseline (mean ± stddev or rolling median) and flags rows outside the expected range. The agent will write a CTE with the baseline and surface deviations.
The meta-prompt
When you don't know which shape to use, just ask: "What's interesting about my orders data this month?" The agent surfaces a few candidate angles and lets you pick.
Save the prompts that work into the chat sidebar's Pinned list — they're one-click reusable.