I'm a PM on a lean team. Over the last few months I've been iterating on using LLMs for data analysis, starting from basic query help and working up to running queries against our analytics database directly from the codebase.
The setup itself is pretty simple: a schema docs and a library of SQL templates, plus obviously direct access to the analytics code itself.
This is my workflow in progress - it's not perfect. Keen to hear what others have tried, especially around verifying AI-generated queries and keeping the error rate down.
I'm a PM on a lean team. Over the last few months I've been iterating on using LLMs for data analysis, starting from basic query help and working up to running queries against our analytics database directly from the codebase.
The setup itself is pretty simple: a schema docs and a library of SQL templates, plus obviously direct access to the analytics code itself.
This is my workflow in progress - it's not perfect. Keen to hear what others have tried, especially around verifying AI-generated queries and keeping the error rate down.