AI insights

The numbers are computed.
The words explain them.

Every card carries an interpretation written from the computed card values. The AI explains what the statistics found — and an independent pass checks the numbers before the report ships.

See it on a live report → Case studies →

A card's interpretationwritten from the values
Interpretation
Sales Representative has the highest attrition rate at 39.8%, while Research Director has the lowest at 2.5%. Across all 9 roles, the median is 16.1%.
from: IBM HR attrition case study — live reportgrounded in the computation

The Problem with Statistical Output

Raw numbers don't tell the business story

Hard to Interpret

p-values, confidence intervals, R² scores—what do they actually mean for your business?

Time-Consuming

Analysts spend hours translating statistical output into executive summaries and recommendations.

Expertise Required

Understanding model diagnostics and assumption checks requires specialized knowledge.

Automatic AI Interpretation

Every report card has a "back" with plain-English insights

Executive Summary

High-level takeaways that anyone can understand. Key findings distilled into bullet points.

Statistical Interpretation

What the numbers actually mean. Model performance explained without jargon.

Actionable Recommendations

Specific next steps based on the analysis. Not just insights—actions you can take.

Interactive Demo

Hover to Flip

Every card has a front (data) and back (AI insight)

OV
Executive Summary
Analysis: Customer Churn
0.892
R² Score
Good
Model Quality
3
Key Predictors
12,450
Sample Size
Classification Decision Tree
AI Insight

Strategic Findings

  • Model explains 89% of churn variance with high confidence
  • Contract type is the strongest predictor—month-to-month customers churn 3.2x more
  • Customers under 12 months tenure are highest risk
  • Recommend: Target month-to-month customers with retention offers
DG
Model Diagnostics
Assumption Checks
Residuals vs Fitted
Diagnostics 4 Tests
AI Insight

Model Validation

  • ✓ Residuals show random scatter—no systematic bias
  • ✓ No multicollinearity detected (VIF < 5)
  • ✓ Sample size adequate for 3 predictors
  • Model assumptions satisfied—results are reliable for business decisions
CF
Model Coefficients
Variable Importance
Variable Impact Sig
Contract Type +0.824 ***
Tenure -0.512 ***
Monthly Charges +0.234 **
Coefficients 3 Variables
AI Insight

Business Impact

  • Contract Type: Month-to-month customers are 82% more likely to churn
  • Tenure: Each additional year reduces churn probability by 51%
  • Monthly Charges: Every $10 increase adds 23% churn risk
  • Focus retention on new, high-paying, month-to-month customers
RC
Model Performance
Prediction Accuracy
Accuracy
87.3% correct predictions
Precision
84.1% of churn predictions correct
Recall
79.2% of actual churners caught
Performance Production Ready
AI Insight

Action Items

  • Deploy: Model passed cross-validation; promote to production
  • Priority: Focus on customers flagged as "high risk" by the model
  • ROI: At $500 per saved customer, model could save $125K/month
  • Next: A/B test retention offers on model-identified at-risk segment

Insights are generated automatically for every analysis you run

How Insights Work

From raw output to actionable intelligence

1

Run Analysis

Execute any statistical analysis through the chat interface or API. Insights are generated automatically.

2

AI Processes Results

Each report card is analyzed by AI to extract key findings, interpret statistics, and identify actions.

3

Get Plain English

Flip any card to see the AI interpretation. Share reports with stakeholders who don't speak statistics.

For Developers

Control insight generation through the API. Enable or disable per-analysis, configure determinism for reproducible outputs.

  • generate_insights=true enables AI interpretation
  • Deterministic outputs with seed parameter
  • Batch processing for multiple insights in parallel
  • Async polling for long-running analyses 
# Enable insights in your analysis
tools_run(
    tool_name="classification_binary",
    taskList={
        "inputs": {
            "dataset": "uuid://your-dataset",
            "target": "churned",
            "predictor_1": "contract_type",
            "predictor_2": "tenure"
        }
    },
    generate_insights=True  # AI insights enabled
)

See Insights in Action

Upload a dataset, run an analysis, and flip the cards to see AI-powered interpretations.

Try It Now How It Works