← sources

If PMs can do their own analysis with AI, does the data team just become infra?

source: reddit (reddit:ProductManagement)published 2026-10-27 21:42ingested 2026-10-28 02:36

retained text

I built a prototype for analysis with AI against a local DuckDB file I set up. The tables, views and macros are heavily commented, and there's a simple way for the agent to explore the db a piece at a time, the same way it would look over a big codebase. Duckdb runs efficiently locally and a bad query can never hit a shared resource.

It works so well that I'm wondering about how this changes the analyst role and product. I'm thinking it means something like:

The data team owns and governs the file and stops doing end analysis. *Need* to be able to trust the lineage and the local instance for accuracy.

Users (product, risk, ops) do their own analysis with AI.

Anything novel someone builds gets submitted like a PR for the data team to review and merge.

The catch is that business context, meaning what a field actually means in the real world, moves out of the data team and into the users. In my fintech world that's product, risk or ops, depending on the slice of data.

In my world I still need to understand 100% of the analysis behind any decision I make. It's why, for me, I don't think it's ok to let a model go from db->deck. I have to be able to walk through the logic start to finish when we're talking about a policy change that shifts millions in a regulated space.

Does this hold up, or am I missing why analysts exist?

What breaks first?

No repo, not selling anything. Just trying to figure out if this is where things are headed and whether I should push to go past my prototype. If you want to know more about what I built specifically, send me a DM. 

document fetched 2026-10-28 00:28 · canonical https://reddit.com/r/ProductManagement/comments/1xh5q98/if_pms_can_do_their_own_analysis_with_ai_does_the