Ziga Potocnik

The 7-14 day prediction window is the right place to focus - by the time someone hits the cancellation flow, you've already lost the conversation. What signals does Flywheel weight most heavily for churn prediction - usage drop-off, support tickets, billing events, or something else entirely?

Building Genie - an AI analyst inside Databox. Users ask questions in plain English ("Why are leads down this week?") and get instant answers, charts, and dashboards from their connected data sources. Stack for the AI layer: LangGraph for agent orchestration (evaluated several frameworks - LangGraph gave us the control we needed for stateful, multi-step tool-calling without abstracting too much...

1) The tagline has to answer one question instantly: "is this for me?" 2) Not what it does. Not how it works. Just - do I recognize my problem in those 8 words? What makes me skip: anything that could describe 50 other products. "AI-powered insights for teams" tells me nothing. "Ask your data anything, get answers in seconds" at least tells me there's a specific interaction being promised. The...

The trust question comes down to one thing: where is the AI getting its information from? For drafts and summaries - fine. But for business decisions, the real danger isn't that the AI fails. It's that it confidently gives you a plausible-sounding answer based on nothing real. That's worse than no answer. My line: if I can't trace the answer back to a real source, I don't act on it.

Awesome project! Curious how Naoma handles demos for products with complex, role-based experiences - where what a sales leader needs to see is completely different from what a developer or finance person would care about. Does the agent dynamically adapt the demo flow based on who's watching, or is it one flow per product?

Great product! How does it handle longer recordings like customer calls or screen recordings where the "good parts" are buried? Does it let you describe what you're looking for semantically, or is it more keyword-based?