Decisioning

Text-to-SQL Agent

Let nontechnical users query supply performance in plain English instead of waiting for reports.

My role

I designed and built the agent and its self-service reporting workflow.

Collaboration

I worked with the supply and data teams to define the questions it needed to answer.

OpenAI Assistants APISnowflakeLLMsMulti-Agent Systems

Problem

Business users depended on manual reports for recurring supply questions. Every new question went back through the data team.

Solution

Multi-model system that accepts natural-language questions, generates SQL, executes securely in Snowflake, and returns tables or visualizations in real time.

Outcome

Measured

Nontechnical users could answer supply-performance questions in real time, cutting 100+ hours of manual reporting each year.

How it works

💬

Input

Accepts plain English questions via chat interface

🔀

Route

Classifies intent and selects the right agent path

Validate

Checks query safety before touching the database

🔧

Generate

Converts natural language into executable SQL

Execute

Runs query securely against Snowflake

📊

Render

Returns tables or charts ready for decisions