AI-Driven Column Insights in Edilitics

The AI-Driven Column Insights feature in Edilitics automatically generates metadata-driven column descriptions that make datasets understandable, documented, and ready for AI-powered analytics.

Whether you're enabling AskEdi, AI-generated chart suggestions, or other intelligent features, accurate column descriptions are mandatory. These descriptions ensure AI models have the right context to produce secure, accurate, and governed outputs - without ever exposing raw data.

By leveraging enterprise-grade AI models under strict privacy controls, Edilitics helps both technical teams and business users eliminate manual documentation work, improve cross-team understanding, and scale enterprise-wide data governance.

Why Column Insights Matter

Without high-quality column descriptions, AI models are essentially guessing what your data represents. This creates real-world problems like:

  • Incorrect analytics and decisions due to misunderstood data.
  • Risk of sensitive data exposure if context isn’t carefully controlled.
  • Slower team adoption because business users can’t trust or interpret results.

AI-Driven Column Insights solve this by automatically creating governed, privacy-safe descriptions based on your metadata, so your datasets become self-describing, trustworthy, and AI-ready.

INFO

If you plan to use AskEdi or AI-driven visualizations, you must have column insights configured for every table you connect. No column descriptions = no AI features.

Privacy & Context Modes

Edilitics uses privacy and context modes to control exactly what information is shared with the AI model during column insights generation. This helps balance accuracy vs. data exposure based on your governance needs.

ModeWhat’s Shared with the LLMBest For
PrivateReal column names + data types only. No table names. No sample data.Maximum privacy with minimal context for basic documentation.
Balanced (Default)Real column names + data types + table name. No sample data.General documentation with richer context for better descriptions.
Full ContextReal column names + data types + table name + 5 random sample rows.Complex datasets where deeper context improves accuracy significantly.
  • Sample data is strictly limited to 5 randomly selected rows to keep sharing minimal and safe.
  • Mode selection is flexible and can be changed per integration at any time.

Choosing an LLM Provider

When generating column insights, you can choose from the LLM providers included in your plan:

  • Anthropic
  • DeepMind
  • OpenAI

This ensures teams can match provider choice to their governance requirements while staying within their licensed plan limits.

How Column Insights Work

  • Metadata Extraction

    Edilitics scans column names, data types, and schema details.

  • Context Mode Enforcement

    Based on the selected privacy mode (Private, Balanced, or Full Context), Edilitics determines what to share with the AI model.

  • AI Description Generation

    The chosen LLM generates natural language descriptions, fully respecting your governance settings.

  • User Review & Finalization

    Users review and edit descriptions for accuracy and alignment with business terminology.

INFO

Once edited, user-provided descriptions become authoritative and are never overwritten by AI, even when schema updates occur.

Tracking Description Status

Every column description clearly shows:

StatusMeaning
AI-GeneratedAutomatically created by the selected LLM with no manual edits.
User-VerifiedA user has edited and saved the description - either partially or completely - making it authoritative.
  • Last Updated Timestamp: Each description displays the last updated date and user or AI attribution. This provides clear auditability for governance and compliance.

Example:


Status: User-Verified
Last Updated: Sep 12, 2025

Configuring AI-Driven Column Insights

Enabling During Integration Setup

You can enable AI-driven column insights as part of your integration setup:

  • Enable

    • Automatically generates descriptions when you save the integration.
    • Dynamically updates when new tables are added or schemas change.
    • Required if you plan to use AskEdi or AI chart suggestions.

  • Disable

    • No descriptions are generated initially.
    • You can generate them later from the Integrate Home Screen at any time.

Important

You cannot manually add descriptions from scratch.
Column insights must first be generated by AI before you can review, edit, and finalize them.

Editing Column Descriptions

  • Review AI Outputs Ensure AI-generated descriptions accurately reflect your business rules and data semantics.

  • Make Manual Edits Fine-tune or completely rewrite descriptions for critical or sensitive columns - but only after initial AI generation.

  • Authoritative Locking Once a description is saved by a user, its status changes to User-Verified and it becomes the single source of truth. AI updates will never overwrite user-verified content.

Benefits of AI-Driven Column Insights

The real value of this feature lies in how it transforms your data workflow, making your datasets accurate, trustworthy, and AI-ready at scale.

The Foundation for AI Analytics

  • Mandatory for AI features: AskEdi and AI chart suggestions require column insights to function correctly.
  • Eliminates AI guesswork by providing structured, accurate context.
  • Ensures models produce reliable, governed results every time.

Time Savings & Operational Efficiency

  • Automatically documents thousands of columns in seconds.
  • Removes repetitive documentation work for engineers and analysts.
  • Accelerates time-to-insight when onboarding new integrations or datasets.

Better Collaboration Between Teams

  • Creates a shared language for data across business and technical teams.
  • Helps non-technical users understand datasets without needing SQL or technical support.
  • Reduces back-and-forth requests for column clarifications.

Strong Governance & Compliance

  • Privacy modes allow you to balance context accuracy with exposure control.
  • Every generation, edit, and provider choice is audit logged for traceability.
  • Aligns with global privacy standards like GDPR, SOC 2, and ISO 27001.

Data Trust and Accountability

  • Transparent status labels (AI-Generated vs. User-Verified) ensure everyone knows the source of a description.
  • Builds confidence across teams by showing who approved what and when.
  • Encourages proactive data stewardship by surfacing stale or unverified descriptions.

Consistency Across the Organization

  • Standardizes dataset documentation across departments.
  • Reduces reporting conflicts caused by inconsistent definitions.
  • Provides a single source of truth for analytics and reporting.

Future-Proof Scalability

  • Automatically adapts to schema changes and new integrations.
  • Handles data growth without increasing manual workload.
  • Ready for advanced features like Replicate and Transform modules.

Security & Governance

  • No Raw Data Exposure: Even in Full Context mode, only 5 random sample rows are shared, never entire datasets.

  • Encrypted at Every Stage: All metadata, descriptions, and sample data are encrypted in transit and at rest.

  • Strict Read-Only Processing: LLMs can only read governed inputs - they cannot modify source data.

  • Comprehensive Audit Trail: Every generation event, privacy mode selection, edit, and LLM provider choice is logged.

Best Practices

  • Start in Balanced Mode Offers strong context without exposing sample rows.

  • Prioritize User Verification for Critical Columns Ensure sensitive or business-critical fields are reviewed and finalized by humans.

  • Keep Metadata Clean Clear column names and accurate data types lead to higher-quality AI descriptions.

  • Switch to Full Context When Needed For highly complex datasets, provide sample data to improve AI accuracy.

Example Output

Column Name: customer_id Status: AI-Generated Last Updated: Sep 10, 2025 by Edilitics AI

“Unique identifier for a customer record, used to link transactions and profiles across systems.”

Column Name: sales_amount Status: User-Verified Last Updated: Sep 12, 2025 by Jane Doe

“Represents the total value of each sale in USD, used for revenue tracking and forecasting.”

Summary

The AI-Driven Column Insights feature provides the governed foundation for all AI analytics in Edilitics.

By combining privacy-first context modes, per-plan LLM selection, and clear accountability tracking, it ensures that AI features like AskEdi and AI visualizations deliver accurate, secure, and trusted results.

No column insights = no AI analytics. This safeguard guarantees every AI-powered decision starts with documented, verified, and compliant data.

Enterprise Support & Technical Assistance

For technical inquiries, implementation support, or enterprise-level assistance, our dedicated technical support team is available to ensure optimal deployment and utilization of Edilitics solutions. Please contact our enterprise support desk at support@edilitics.com. Our team of specialists will respond promptly to address your requirements.

« Getting Started

Auto Schema Mapper »