AI tells you where to look in your data, never what it means
You set down your decision and your hunch before opening the tool, you read every query before trusting its number, you recompute whatever will drive the decision. You leave with a sourced place to dig, not a solution.
Why
A clean number is not a correct number. Asking an assistant wired into your analytics why users are not coming back hands it the metric, the query and the explanation. All three come out with the same confidence, and nothing tells the wrong query from the right one. I have done this, or watched a team I was supporting do it, four ways: export into an assistant, the assistant built into the analytics tool, AI-generated SQL, a spreadsheet. AI-generated SQL gets believed only after someone has read it. Data tells you where people drop off. The why comes from a user somebody actually listened…
When
You build this kit when a product decision is waiting on a number. Phase: Discovery. Audience: PMs, POs and product designers who read the data themselves. ✅ You know it's your kit when You have an instrumented analytics tool and nobody to write your queries A figure gets quoted in meetings and nobody knows where it came from You want to know where to dig before running interviews ❌ Not for…
Tools included
- Posthog
- Claude