Product knowledge at a fast-moving company lives in too many places at once: warehouses, docs, code, and the heads of whoever shipped the feature. New hires take weeks to get productive, and PMs and support burn hours re-answering the same deep product questions. So I built an internal knowledge bot, a Glean for Stellic, to close that gap.
I built it solo, from scratch, largely in Claude Code (the first thing I open every morning). It connects to all of our data warehouses and sources, retrieves the right context semantically, and answers deep product questions in plain language.
Impact
- Used by ~40% of the company every day.
- 10,000+ agent responses in the first 90 days after launch.
- New hires answer expert-level product questions from day one instead of week three.
Where it's going
It started as a side project and is now embedded in how the team actually works:
- Triage: it triages incoming Linear tickets for the PMs and helps debug bugs.
- Slack translation: it monitors technical channels and turns raw output, like a data-sync crash dumping stack traces and scripts, into a plain-English root cause and next steps, by researching the codebase and querying our observability layer.
Why it matters
A side project became infrastructure the company relies on. It is the clearest example of how I work as a PM: I don't write a spec and wait, I build the thing, put it in people's hands, and let real usage decide what it becomes.