Introducing Natural Language SQL: Ask Questions, Get Answers, Skip the Query Builder

Connect your database, ask a question in plain English, and Skopx generates optimized SQL, runs it, and shows you the results with interactive charts. No SQL knowledge required. No query builder to learn.

The Problem With Query Builders

Every BI tool has a query builder. It is the interface where you drag fields, apply filters, choose aggregations, and configure grouping. Learning a query builder takes days to weeks. And every tool has a different one.

For the 73% of business users who are not technically trained, query builders are a wall between them and their data.

How Natural Language SQL Works

When you ask Skopx a question like "What were our top 10 customers by revenue last quarter?", the following happens:

1. Schema understanding: The AI reads your database schema (tables, columns, types, relationships) and identifies which tables and columns are relevant.

2. Query generation: The AI generates a SQL query that answers your question. For the example above, it might produce:

SELECT customer_name, SUM(amount) as total_revenue FROM orders WHERE created_at >= '2026-01-01' AND created_at < '2026-04-01' GROUP BY customer_name ORDER BY total_revenue DESC LIMIT 10

3. Execution: The query runs against your database with a 10-second timeout and row limit for safety.

4. Visualization: Results are displayed as an interactive table with automatic chart suggestions. Revenue data gets bar charts. Time series gets line charts. Distributions get histograms.

5. Follow-up: You can drill down with follow-up questions. "Break that down by product category" adds a GROUP BY without starting over.

Accuracy

Our NL2SQL pipeline achieves 94.2% accuracy on production workloads across PostgreSQL and Supabase databases. The remaining 5.8% of queries either produce incorrect results (2.1%) or fail to execute (3.7%).

When a query fails, the AI automatically attempts self-correction up to 3 times, adjusting the SQL based on the error message. This brings the effective success rate to 97.5%.

Safety

Natural language SQL raises legitimate safety concerns. Our guardrails:

• Read-only by default: All queries run with read-only permissions. Write operations require explicit user confirmation.
• Row limits: Results are capped at 10,000 rows to prevent accidental full-table scans.
• Timeout: Queries that run longer than 10 seconds are killed.
• No DDL: The AI cannot create, alter, or drop tables.

Supported Databases

Currently supported: PostgreSQL, Supabase, MySQL, and SQLite. Support for BigQuery, Snowflake, and Redshift is in development.

Getting Started

1. Go to the Data tab in any chat
2. Click "Connect Data Source"
3. Paste your database connection string
4. Start asking questions

The AI automatically discovers your schema and is ready to answer questions within seconds of connecting.