Insights.
Your AI that knows.
Ask anything. Get cited answers.
Ask any question of your business in plain English. Get cited, verified answers with charts and narrative, in seconds, not sprints.
No SQL. No dashboard sprawl. The Insights Engine mines patterns from every data source you have and returns grounded answers your team can defend in a meeting or an audit.
What Insights does for you
Four core jobs, done well.
Natural-language queries
Ask complex analytical questions the way you'd ask a senior analyst. No schema knowledge required.
Grounded + cited
Every number, entity, and claim traces back to its source. Verifiers block hallucinations before they reach you.
Automatic visualization
The right chart for the result shape. Switch axes, filters, and comparisons interactively.
Narrative explanation
Every answer ships with a plain-English “why”. The drivers behind the trend, not just the trend.
What's in the box
The Insights Engine, and the Foundation underneath.
Every Quadrazene Engine ships with the same Foundation. Insights adds its own Skills, atoms, and Reactions on top.
The Insights Engine adds
Insights Skills library
Built-in Skills for top-N, trends, segmentation, variance, retention, and ad-hoc queries. Parameterized with {{template}} variables.
NL-to-SQL atom
Domain-aware NL→SQL with per-System schema prompts and multi-section UNION support.
Chart-selector
Picks bar / line / area / scatter / heatmap / big-number to match result shape, and falls back gracefully when degenerate.
Insights-interpreter
Generates the narrative and observation bullets that accompany every answer.
Followup-generator
Suggests next questions with pre-validated SQL so the click skips nl-to-sql entirely.
Agentic SQL retry
A broken query is auto-corrected and re-run. Users see “Adjusting… attempt 2 of 3,” not a red error.
Foundation · included
- ›Reactor - The chat workspace where every question lands.
- ›Compose - Describe what you need. An agent drafts the Skill, Formula, or View.
- ›Designer - Dual-pane visual + NL editor for Formulas and Views.
- ›Recipes - Formulas chain into Chains. Versioned, replayable runs.
- ›Skills - Parameterized prompt templates with {{template}} variables.
- ›Inbox - HITL approvals, Action Items, and Findings in one place.
- ›Mailbox - Email + channel intake. Parse, classify, route.
- ›Records - Immutable provenance timeline for every Reaction.
- ›Connections - Data sources, SAP, MCP servers, Teams, and the API spine.
- ›Trust Layer - Content filters, prompt-injection detection, model allowlist, classification cap.
- ›Risk - Composite 0-100 score per Reaction. Auto-HITL above threshold.
- ›Models - Multi-provider LLM routing with a spend dashboard.
Every Engine ships with the full Foundation. No separate purchase.
Live sample · real data from the app database
Not a mockup. The platform actually produced this.
Which product categories contributed most to the highest revenue months?
Electronics dominates the highest revenue months with consistent leadership across all three peak periods (March, July, October 2024), generating 37-39% of total revenue in each month. The category mix remains remarkably stable across these peak months, with Electronics, Home & Kitchen, and Accessories forming the top tier while Furniture consistently trails at roughly half the revenue of the leading category.
- Electronics leads all three highest revenue months with $1.45M (March), $1.34M (October), and $1.23M (July), showing consistent dominance despite month-to-month fluctuations.
- The revenue hierarchy remains locked across all peak months: Electronics > Home & Kitchen/Accessories (alternating 2nd/3rd) > Furniture, suggesting structural market positioning rather than seasonal shifts.
- Furniture consistently generates the lowest revenue at roughly $500-600K per peak month, representing only 40-42% of Electronics' revenue in the same periods.
- Home & Kitchen shows the most volatility among top categories, ranging from $872K (March) to $1.08M (July), a 23% swing that exceeds other categories' fluctuations.
▸ Inspect rows (12)
| category | revenue month | category revenue |
|---|---|---|
| Electronics | 2024-10 | $1.34M |
| Home & Kitchen | 2024-10 | $971K |
| Accessories | 2024-10 | $858K |
| Furniture | 2024-10 | $608K |
| Electronics | 2024-07 | $1.23M |
| Home & Kitchen | 2024-07 | $1.08M |
| Accessories | 2024-07 | $884K |
| Furniture | 2024-07 | $569K |
| Electronics | 2024-03 | $1.45M |
| Accessories | 2024-03 | $929K |
▸ Inspect SQL · executed against the configured System
SELECT
p.category,
DATE_FORMAT(o.order_date, '%Y-%m') AS revenue_month,
SUM(oi.quantity * oi.price_at_purchase) AS category_revenue
FROM orders o
JOIN order_items oi ON o.order_id = oi.order_id
JOIN products p ON oi.product_id = p.product_id
WHERE DATE_FORMAT(o.order_date, '%Y-%m') IN (
SELECT revenue_month
FROM (
SELECT
DATE_FORMAT(order_date, '%Y-%m') AS revenue_month,
SUM(total_price) AS monthly_revenue,
RANK() OVER (ORDER BY SUM(total_price) DESC) AS revenue_rank
FROM orders
GROUP BY DATE_FORMAT(order_date, '%Y-%m')
) ranked_months
WHERE revenue_rank <= 3
)
GROUP BY p.category, DATE_FORMAT(o.order_date, '%Y-%m')
ORDER BY revenue_month DESC, category_revenue DESCCharts, narrative, findings, and payloads are exactly what the platform produced. Sanitized for display.
Bond more Engines
Each Engine bonded into the Ring catalyzes the others.
Insights stands on its own. Bond another Engine and the loop closes. Here's what each pairing unlocks.
Continuous controls over your real data.
Every Insights query passes through your data-classification ceiling and PII filters. Restricted Systems never leak to a model that isn't cleared for them, and every read is on the record.
Add Governance →Forecasts grounded in real history.
Advisory's recommendations cite the Insights queries they're built on. Forecasts and scenarios are anchored to the same audited rows that drove the answer.
Add Advisory →Turn answers into writes.
From the same Reactor message that surfaced the customer at risk, post the PO, update the order date, or open the QM notification, without leaving the chat.
Add Actions →Where it pays off
Insights in the real world.
Finance & FP&A
Variance analysis, close commentary, cashflow deep-dives that auditors and boards can trust.
Revenue operations
Pipeline diagnostics, churn root-cause, segmentation, without a BI ticket.
Operations leadership
Throughput, quality trends, supply-chain anomalies. See what happened and why.
Sample questions
What users actually ask.
Use it however you want
The Reactor, or your own framework.
Expose Insights Skills as tools to your existing agent framework. Agent Core, Step Functions, n8n, or a homegrown orchestrator can invoke any Skill over REST/SSE. Same auth, same audit, same answer the Reactor would have produced.
Two adoption patterns
Put the Insights Engine to work.
A working session with your own data. Start with Insights. Bond more Engines when you're ready.