Sundial

Julie Zhuo spent years as head of design for the Facebook app, working at massive scale with billions of users. But after leaving Meta and launching her newsletter 'The Looking Glass,' which became foundational to product thinking culture, she noticed a critical gap: companies growing at hyperspeed—with hundreds of millions in ARR and users—were making decisions on "good instincts and good vibes" rather than data. They lacked the infrastructure that took Meta years to build. Most fast-growing startups, she realized, never had time to instrument their businesses properly. That became her north star for Sundial: automate data analysis so any team could understand their reality instead of relying on gut feel.

Julie built Sundial with a radically lean team—mostly engineers—refusing to hire traditional product managers. Instead, she pushed engineers to think like builders, making product decisions themselves rather than delegating to a PM. This constraint forced deeper ownership. The team used AI heavily in their own workflows: ChatGPT for teaching new skills, Cursor for coding acceleration, and prompt engineering for custom learning. Engineers on the team learned data analysis and product thinking by doing, supported by AI as a tutor. This became a living experiment in her thesis that AI empowers individuals to wear multiple hats.

Sundial found early traction with AI-native companies—OpenAI, Gamma, and Character AI—who understood the problem acutely. These companies were growing so fast they couldn't hire data teams quickly enough and didn't have mature analytics infrastructure. They needed to diagnose what was working in conversational interfaces, understand user intent in conversation flows, and measure engagement in entirely new ways. Traditional metrics (clicks, sessions, time-on-page) didn't apply to LLM products. Sundial's automation solved this urgently.

What worked: removing friction on decision-making. Julie's framework—"diagnose with data, treat with design"—resonated because it wasn't about replacing intuition with rigid A/B testing, but empowering teams to see reality more clearly before making creative leaps. Designers initially resisted data, fearing it would constrain their work. But Julie showed that great designers are obsessed with understanding user reality; data is just a tool to make that visible.

What also worked: hiring for builder mentality over roles. By eliminating the PM layer, engineers owned product thinking. It slowed things initially—they had to learn new skills—but compounded long-term as people became more well-rounded and autonomous.

Sundial is working with some of the most ambitious AI companies on the planet, helping them answer questions no traditional analytics could address: What conversational patterns indicate user satisfaction? Which use cases are growing? How do we measure value in agentic systems? Julie has also become a thought leader on the broader implication of AI in management—arguing that managing AI agents uses the exact same skills as managing people (clear goals, understanding strengths, defining process), and predicting that job titles will collapse into simply 'builder.' Her new paperback edition of 'The Making of a Manager' includes chapters on managing in the age of AI.