Lenny's PodcastJulie Zhuo: How AI is flattening orgs into broader builders
At Sundial, no dedicated PMs: engineers cover design and analysis; Zhuo diagnoses with data and treats with design, leading like a sturdy, flexible willow.
CHAPTERS
- 1:25 – 7:00
Julie Zhuo returns: book rerelease, Sundial, and the new management context
Lenny welcomes back Julie Zhuo—his first-ever podcast guest—and frames her evolution from Meta design leader to founder of an AI-powered analytics company. They set up the episode’s core themes: what management means in an AI era, and why Julie’s updated edition of The Making of a Manager is timely.
- •Julie’s background: Meta head of design → founder of Sundial (AI analyst)
- •Why this conversation matters now: AI reshaping roles, teams, and leadership
- •Julie’s influence on Lenny’s newsletter/podcast trajectory
- •Preview of episode topics: agents, team structure, data, and timeless management skills
- 7:00 – 8:42
The surprising success of The Making of a Manager—and why it resonated so widely
Julie reflects on why she wrote her book—primarily to become a better manager herself—and why it exceeded her expectations. They discuss how the book normalized common management insecurities and became a modern staple alongside classics like High Output Management.
- •Writing as a tool for self-improvement and clarity
- •Readers’ biggest takeaway: feeling less alone in manager struggles
- •Why the book traveled beyond Silicon Valley management culture
- •Comparison to older management classics and why newer guidance is needed
- 8:42 – 11:39
Why AI will make everyone a manager: managing outcomes, models, and process
They explore the idea that the core skills of people management transfer directly to managing AI agents. Julie reframes management as outcome-setting, resource allocation, and process design—now applied to models and tools rather than only humans.
- •Management = outcomes + resources + process (still true with agents)
- •Different models have different strengths (“assemble the Avengers”)
- •Clarity and context become essential inputs to agentic systems
- •AI changes the “resource” mix but not the fundamentals of coordination
- 11:39 – 14:01
The key skill for managing agents: defining success with precision (and evals/KPIs)
Julie argues the most important transferable skill is being crystal clear about what success looks like. She links this to alignment problems in companies, the need for evals in AI, and why metrics exist: to create objective criteria for progress.
- •Defining success is hard for humans—and critical for agents
- •Misalignment often stems from different mental pictures of success
- •Evals and KPIs operationalize objective criteria
- •Prompt quality depends on specificity of desired outcomes
- 14:01 – 16:30
Flattening orgs and fewer managers: do we still need management?
Lenny raises the trend of flatter orgs, fewer middle managers, and leaders becoming ICs again. Julie explains AI’s promise: individuals are more empowered to do formerly specialized work, pushing teams toward fewer role boundaries without eliminating the need for coordination and vision.
- •AI enables one person to do more disciplines at a competent level
- •Traditional role boundaries blur; “builder” becomes a better identity
- •Managers may shrink in number, but management as a function persists
- •Empowerment shifts how work gets done, not the need for direction
- 16:30 – 21:31
How Sundial structures teams: fewer roles, no default PMs, and skill-based staffing
Julie shares how her startup experiments with smaller teams and fewer role “lanes,” including not hiring many product managers. The goal is to reduce delegation-by-title and increase ownership—then fill gaps based on skills needed in context rather than a standard org template.
- •Removing default roles increases ownership and initiative
- •Hiring/staffing based on needed skills, not fixed archetypes
- •Engineers pushed to engage with product thinking and communication
- •Generalists + specialists-as-reviewers model for leverage
- 21:31 – 26:45
Which roles AI accelerates most—and the underrated superpower: accelerated learning
They discuss where AI creates the biggest productivity gains (especially engineering and prototyping) and how non-engineers increasingly build. Julie highlights an underused benefit: using AI as a personalized tutor to rapidly learn new domains and validate understanding interactively.
- •Engineering speedups plus broader prototyping across functions
- •Hybrid roles like “Product Science” blending analysis, CS, and product
- •AI as an adaptive teacher tailored to learning styles
- •Using AI to test understanding by explaining concepts back
- 26:45 – 29:39
Data analysis in AI companies: fast growth, poor instrumentation, and scrambling later
Julie offers a contrarian observation: many fast-growing AI companies aren’t yet good at data because they grew too quickly to build logging and infrastructure. She explains why observability becomes critical when growth slows and root-cause questions suddenly matter.
- •Hypergrowth outpaces analytics foundations (logging, instrumentation)
- •Companies often run on instincts until growth plateaus
- •Observability enables earlier detection and better diagnosis
- •Data investment usually spikes during slowdowns and uncertainty
- 29:39 – 32:02
New analytics for conversational products: intent, quality, and methodology shifts
Julie explains that each technology shift demands new analytic methods, and LLM products are no exception. With conversational interfaces, clicks and tabs don’t reveal intent; teams must infer use cases, measure experience quality, and develop new metrics for “good” conversations.
- •Conversational analytics differs from traditional clickstream analysis
- •Intent bucketing may require ML/LLMs to classify conversations
- •Ambiguity: long vs short conversations aren’t inherently better
- •Need to invent new methodologies to assess value and flow quality
- 32:02 – 37:03
Data vs design: “diagnose with data, treat with design” (and why designers should care)
Julie addresses designers’ skepticism toward data-driven decisions, clarifying that data reveals reality and problems, while design creates solutions. She broadens “data” beyond A/B tests to include qualitative signals (TikTok, interviews) and warns against false precision and ‘A/B testing your way to greatness.’
- •Framework: diagnose with data; treat with design
- •Data is broader than metrics—includes qualitative cultural signals
- •Numbers require interpretation; choosing what to measure is an art
- •Great designers obsess over understanding reality, not ignoring it
- 37:03 – 42:13
The manager’s job in the AI era: managing change, fear, and ‘sturdy yet flexible’ leadership
They return to management and ask what’s most different now: the accelerating rate of change and heightened uncertainty. Julie introduces the willow tree metaphor—leaders must be sturdy (grounded) while flexible (adaptive), acknowledging fear without letting it dominate.
- •Change management has always mattered; now it’s constant and faster
- •Career anxiety rises as AI threatens traditional identities
- •Leadership requires communication, compassion, and realism
- •Willow tree model: resilient through flexibility
- 42:13 – 49:29
Timeless manager lesson: manage yourself—strengths, weaknesses, and ‘dimensionality’
Julie’s foundational advice is self-management: understanding your unique profile across many dimensions and not equating feedback with identity. She explains how strengths and weaknesses are often two sides of the same trait and how context determines what to lean into or develop.
- •‘Dimensionality’ separates skill feedback from personal worth
- •Strengths create predictable blind spots; weaknesses can enable strengths
- •Mastery = reading context and dialing traits up or down
- •Practical tactic: narrate uncertainty while contributing in real time
- 49:29 – 57:58
Career growth decisions: when to build weaknesses vs double down on strengths
Julie answers how far people should push on weaknesses: it depends on your goals and the path you want (IC craft mastery vs leadership trajectory). Misalignment—wanting outcomes that your chosen path won’t deliver—creates dissatisfaction, so clarity of goals comes first.
- •IC path = deep craft; management path = broader capability set
- •Work on weaknesses only when required by your goals
- •Misalignment between ambitions and chosen focus causes suffering
- •Be explicit with managers about goals to make tailored roles possible
- 57:58 – 1:15:43
Feedback culture, win-win thinking, and navigating disagreement with leadership
Julie emphasizes feedback as a daily practice and outlines tactics for giving/receiving it: opt into feedback early, check your intent, and voice vulnerability when delivering hard notes. She adds ‘win-win’ as a core management lens, and shares how to handle top-down decisions you disagree with by decomposing assumptions and designing small tests.
- •Feedback accelerates improvement; perception is often biased
- •Tactics: opt-in agreements, intent checks, vulnerability statements
- •Win-win framing reduces adversarial manager/employee dynamics
- •Disagree-and-commit works best when assumptions are isolated and testable
- 1:15:43 – 1:36:23
AI corner, contrarian corner, and lightning round: creative parenting, infinity mindset, and rapid-fire favorites
In AI corner, Julie shares playful, hands-on uses of AI for parenting—building talking toys and creating personalized parody songs. In contrarian corner, she offers a philosophy of “infinity in every direction” as a cure for boredom and victimhood. The lightning round covers book recommendations, favorite products, mottos, and what she wants her kids to learn in an AI-first future.
- •AI as a creative amplifier for personalized gifts and learning
- •Building a voice-enabled talking raccoon and parody-song ‘album’ workflow
- •Contrarian belief: ‘infinity in every direction’ and cultivating attention
- •Lightning round: books, tools (Cursor/Replit/Granola/Limitless), motto, emotional regulation
