When we left that world we thought we could solve the problem with auto-documentation. Cut to now and we've landed on the modern version of that solution, which is to turn previously executed queries into context for AI. The intuition is roughly this: a smart analyst can read SQL and have a pretty good idea of what's going on, and in fact can infer a lot of institutional knowledge about the domain and how to analyze particular data. LLMs are not great at this (Anthropic and Snowflake have written as much). So we act as the "smart analyst" to pull out institutional knowledge and practices (in the form of SQL) that you can give to an LLM so that it can code like a competent analyst on your team.
Right now we're deploying it as a semantic layer population tool. You want to fill up a Genie or Cortex semantic layer? Run our code on your query history, prune it with your experts (delete this, rename that, etc.), and get moving. That can take hours/days instead of weeks. Plus, everything we export is portable to whatever system you choose.
FYI we ask for emails on the free trial download so we can monitor our traction and build our network but you don't need to include it if you don't want to.
We're hungry for feedback from people who work in the field and are facing these challenges. Major, major thanks in advance for your time.