If you are a developer, getting an answer from the data is easy.

You just connect to your SQL database, write a query, and a few minutes later, you have your answers.

Marketing and product on the other hand are always hungry for numbers and insights into user behavior and metrics, but they don't have access.

And if they have access, they usually can't write their own SQL queries.

For years, this information disparity has been a cause for constant friction. The dev team never has time for doing analytics properly, especially in smaller teams.

What's better than giving a non-technical team access to the raw data?

Configure AI with (read) access to your SQL server and give it to your insights-hungry colleagues.

Claude 3.7 with MCP can interact with a SQL server and the accuracy is quite good, no problem for getting basic insights.

You have to steer it in the right direction if you're dealing with complex queries that use window functions and multiple joins.

But I've found that even if a non-technical person can get most of the answers themselves, the amount of work to fact-check the queries and help AI rewrite the complex ones still saves a lot of time overall.

I wouldn't blindly trust the numbers for any business-critical decisions, and it's not yet a replacement for a dedicated data analyst, but it's useful.

I wrote my own SQL MCP server (using AI), but here's a better implementation if you want to hook up your PostgreSQL to Claude: https://github.com/stuzero/pg-mcp​

Enjoy,

Taj

​