Natural Language Data Analysis
Ask your business data questions in plain language and get answers in seconds. No SQL, no dashboards, no waiting for the data team.
The problem
Business users need answers from company data every day. Most can’t write SQL. Dashboards only answer questions someone already thought to ask. Every ad hoc question means waiting for the data team.
Sending raw data to AI doesn’t work either — datasets are too large for context windows, costs explode, and there’s no persistent connection to live data.
What we built
A system where you ask questions in plain language. The AI generates the database query, the database executes it, and only the results come back to your conversation. The AI never sees the raw data — just the answer.
Currently running with 70,000 rows across 6 business data tables: budget, transactions, orders, customers, revenue targets, and employees.
Example questions:
- “What were our top 5 customers by margin last quarter?”
- “Show me departments that exceeded budget by more than 10%”
- “Compare actual revenue vs targets for Q1, by product line”
Why it works
The key is a context layer — a structured file that tells the AI what the data means: table relationships, naming conventions, and business logic rules. Without this, AI generates wrong queries. With it, accuracy is high.
Stack
DuckDB · MCP · Claude · Python · CSV/Excel