Hot AI startups are getting by on 'vibes' — but need data as a reality check, says former Facebook VP (www.businessinsider.com)

Former Facebook design VP Julie Zhuo, now cofounder of AI analytics startup Sundial, warned that many high‑flying AI startups are scaling on "good instincts and good vibes" without a solid data foundation. She says companies with tiny teams are amassing hundreds of millions in ARR and users but often lack basic logging, observability, and dedicated data teams to answer core questions about churn, feature value, and revenue drivers. Zhuo predicts that when hypergrowth slows, these firms will be scrambling to retrofit analytics and infrastructure to understand what’s actually happening. The warning matters because product measurement and instrumentation need to evolve for AI-first experiences. Traditional metrics like clicks and pageviews don’t capture conversational intent, so Zhuo recommends using LLMs or machine‑learning models to bucket user intent and build new analytical pipelines. The caution comes amid a broader funding boom — more than $35B raised in 2024 — and investor concern about overheating; even OpenAI’s CEO has called some valuations “not rational.” The takeaway for the ML community: invest early in logging, observability, and intent‑level analytics to avoid technical debt and strategic blind spots as AI products scale.