LLenny's Podcast
← All episodes
Julie Zhuo (Facebook VP, Sundial CEO, The Making of a Manage21 September 2025

From managing people to managing AI: The leadership skills everyone needs now

8Frameworks
15Insights

Episode overview

Julie Zhuo argues that managing AI agents relies on familiar management fundamentals: a clear outcome, knowledge of each resource’s strengths, and an effective process. She describes how AI enables smaller, flatter teams of generalist “builders,” while managers must provide stability, flexibility, feedback, compassion, and conviction amid accelerating change. The conversation also covers Sundial’s approach to analytics, including the principle “diagnose with data and treat with design,” plus timeless advice on self-awareness, feedback, win-win relationships, and aligning skill development with personal goals.

Key ideas

  • Managing agents resembles managing people: define success clearly, select resources according to their strengths, and design the process that brings them together.
  • AI can dissolve rigid job boundaries and enable smaller teams whose members act as cross-functional builders while retaining access to specialists when context demands it.
  • Managers should be sturdy yet flexible like a willow tree, acknowledging uncertainty while helping teams adapt to accelerating technological change.
  • Data should diagnose reality and identify problems or opportunities; creative design judgment is still required to invent and implement treatments.
  • Fast-growing AI companies may postpone analytics infrastructure, but weak observability becomes costly when growth slows and teams need root-cause explanations.
  • Effective management begins with self-knowledge: strengths and weaknesses are often two sides of the same trait, and growth should be guided by the person’s goals.
  • Feedback works best as a frequent, explicitly agreed practice grounded in a genuine intention to help, with vulnerability stated openly when difficult feedback is delivered.
  • AI can personalize education and accelerate learning, but Julie emphasizes emotional regulation and deliberately choosing meaningful challenges rather than using technology only as a shortcut.

Transcript available · source text is retained privately and is not published

Frameworks in this episode

Productivity4 steps

AI as Your Personalized Just-In-Time Tutor

Feed AI a curriculum tuned to how you learn, then prove understanding by teaching it back.

Read →
Leadership4 steps

Decompose the Strategy You Disagree With Into Hypotheses

Break a plan you doubt into assumptions, find the one you reject, and design the smallest test.

Read →
Communication3 steps

Define Success Before You Prompt

The clearer your definition of success and failure, the better the work you get from people or AI.

Read →
Strategy3 steps

Diagnose With Data, Treat With Design

Data tells you where the problem is; only a creative process tells you how to solve it.

Read →
Self-Mastery4 steps

Dimensionality: Every Strength Is Its Own Weakness

See yourself as infinite dimensions so feedback becomes data, not an identity threat.

Read →
Leadership4 steps

Dissolve the Roles: Build Small Builder Teams

Shrink teams and drop role labels so AI-empowered individuals own the whole problem.

Read →
Communication3 steps

Feedback as a Daily Practice: Opt-In, Check Intention, Name the Difficulty

Make feedback frequent and safe by pre-agreeing to it, checking your motive, and admitting it's hard.

Read →
Leadership4 steps

Goal-Talent-Purpose-Process: Managing People and AI With One Playbook

Treat managing agents like managing people: same four levers, different resources.

Read →

People & resources mentioned

Attributed to the moment in the episode. Timestamps are approximate.

People · 8

  • Julie ZhuoMentionsJulie was a long time design leader at Facebook she's now the founder of a company called Sundial

    “Today my guest is Julie Zoo.”

    ▶ Watch at ~1:00approx.
  • Julie ZhuoCoinedJulie was a long time design leader at Facebook she's now the founder of a company called Sundial

    “She's also the author of the best-selling and very important book, The Making of a Manager.”

    ▶ Watch at ~1:30approx.
  • Lenny RachitskyMentionstoday we've got another very special compilation episode something I've been pulling on more and more with the podcast and the newsletter

    “Thank you, Lenny. I'm so excited to be here.”

    ▶ Watch at ~5:00approx.
  • Marc BenioffMentionsPreviously appeared on the podcast and discussed Agentforce and the future of sales and customer-experience jobs.

    “I had Mark Beni off on the podcast and I asked him just how do you deal with all this change?”

    ▶ Watch at ~40:00approx.
  • Kevin WeilMentionsWe got connected through Kevin Wheel, who is former CPO at OpenAI, now head of science at OpenAI.

    “I forget who it was exactly. Maybe Kevin Wheel, maybe Mike Kger.”

    ▶ Watch at ~41:30approx.
  • Mike KriegerMentionsMike is chief product officer at Anthropic the company behind Claude. He's also the co-founder of Instagram.

    “I forget who it was exactly. Maybe Kevin Wheel, maybe Mike Kger.”

    ▶ Watch at ~41:30approx.
  • Ethan EvansMentionsEthan is a former vice president at Amazon executive coach and course creator focused on helping leaders grow into Executives

    “It reminds me of a guest post by Ethan Evans that I'll link to”

    ▶ Watch at ~56:00approx.
  • Emily OsterMentionsmentioned

    “I love her and I love her stuff. I watch all her videos on Tik Tok and Emily Auster.”

    ▶ Watch at ~1:27:00approx.

Resources · 36

  • Lenny's NewsletterCoinednewsletter · Lenny Rachitsky

    “Julie's newsletter, The Looking Glass, was the inspiration for my newsletter”

    ▶ Watch at ~1:30approx.
  • The Looking GlassCoinednewsletter · Julie Zhuo

    “Julie's newsletter, The Looking Glass, was the inspiration for my newsletter”

    ▶ Watch at ~1:30approx.
  • FacebookMentionscompany · Meta

    “she was the longtime head of design for the Facebook app used by over 3 billion people.”

    ▶ Watch at ~1:30approx.
  • The Making of a ManagerCoinedbook · Julie Zhuo

    “She's also the author of the best-selling and very important book, The Making of a Manager.”

    ▶ Watch at ~1:30approx.
  • SundialUsescompany · Julie Zhuo

    “used by companies like OpenAI, Gamma, and Character AI.”

    ▶ Watch at ~2:00approx.
  • Character.AIUsescompany

    “used by companies like OpenAI, Gamma, and Character AI.”

    ▶ Watch at ~2:00approx.
  • GammaUsessoftware · Gamma

    “used by companies like OpenAI, Gamma, and Character AI.”

    ▶ Watch at ~2:00approx.
  • OpenAIUsescompany

    “used by companies like OpenAI, Gamma, and Character AI.”

    ▶ Watch at ~2:00approx.
  • MetaMentionscompany

    “Having worked at a mega-arge corp like Meta as head of design”

    ▶ Watch at ~2:00approx.
  • SundialCoinedcompany · Julie Zhuo

    “And most recently, she started her own company, Sundial, which is an AI powered analyst”

    ▶ Watch at ~2:00approx.
  • Lenny's NewsletterMentionsnewsletter · Lenny Rachitsky

    “Head on over to lennisnewsletter.com and click product pass.”

    ▶ Watch at ~3:00approx.
  • The Making of a ManagerMentionsbook · Julie Zhuo

    “you're re-releasing your incredible book, The Making of a Manager.”

    ▶ Watch at ~7:00approx.
  • High Output ManagementMentionsbook · Andy Grove

    “I feel like it's the modern-day high output management.”

    ▶ Watch at ~8:00approx.
  • GoogleMentionscompany · Google

    “I also just saw a stat that Google let go of so many of their middle managers of smaller teams.”

    ▶ Watch at ~14:00approx.
  • AirtableMentionssoftware · Airtable

    “I just had the CEO of Air Table on the podcast”

    ▶ Watch at ~14:00approx.
  • ChatGPTUsessoftware · OpenAI

    “I'll just feed it into chat PT and I'll say help me customize a program for me”

    ▶ Watch at ~23:30approx.
  • CursorUsessoftware · Anysphere

    “We can use cursor. It helps us. It autocompletes. It writes a bunch of things.”

    ▶ Watch at ~24:00approx.
  • ChatGPTRecommendssoftware · OpenAI

    “if you just talk to chat GBT it's probably going to give you a much better answer and allow you to go deeper”

    ▶ Watch at ~25:30approx.
  • XMentionswebsite · X Corp.

    “what are they saying in the Twitter verse or Xverse I guess is what it's called now”

    ▶ Watch at ~34:00approx.
  • The Magic LoopCoinedother · Ethan Evans

    “a guest post by Ethan Evans that I'll link to that has a really good framework”

    ▶ Watch at ~56:00approx.
  • The Magic LoopMentionsother · Ethan Evans

    “called the magic loop where it it's kind of a framework for figuring out what to work on”

    ▶ Watch at ~56:00approx.
  • MinecraftMentionssoftware

    “Like, he loves video games, so Minecraft, but what he often listens to on his Alexa are these parody songs.”

    ▶ Watch at ~1:18:00approx.
  • ReplitUsessoftware · Replit

    “So, I created an app on Replet”

    ▶ Watch at ~1:18:30approx.
  • Kingdom RushMentionssoftware

    “Lock likes playing Kingdom Rush right now.”

    ▶ Watch at ~1:19:00approx.
  • Zen and the Art of Motorcycle MaintenanceRecommendsbook · Robert M. Pirsig

    “The first is Zen and the Art of Motorcycle Maintenance. I absolutely love that book.”

    ▶ Watch at ~1:24:30approx.
  • Conscious BusinessRecommendsbook · Fred Kofman

    “Second is Conscious Business. It is my favorite management book.”

    ▶ Watch at ~1:25:00approx.
  • GranolaRecommendssoftware · Granola

    “I love granola. I love Repid.”

    ▶ Watch at ~1:27:30approx.
  • ReplitRecommendssoftware · Replit

    “I love granola. I love Repid.”

    ▶ Watch at ~1:27:30approx.
  • CursorRecommendssoftware · Anysphere

    “I've used all of the different coding apps. Cursor is is big on me for for now.”

    ▶ Watch at ~1:27:30approx.
  • La La LandRecommendsfilm

    “I think the only thing I watched this year was a rewatch of La La Land, which I do truly love.”

    ▶ Watch at ~1:27:30approx.
  • MaticUsesproduct · Matic

    “I just got a Madic robot. I think that's been really delightful so far.”

    ▶ Watch at ~1:28:00approx.
  • Limitless PendantRecommendsproduct · Limitless

    “one of the best things that the pendant does is it gives me feedback on parenting.”

    ▶ Watch at ~1:29:00approx.
  • How I AIMentionspodcast · Claire Vo

    “there's a recent episode of the our how I AI podcast or sister podcast”

    ▶ Watch at ~1:30:00approx.
  • The Looking GlassMentionsnewsletter · Julie Zhuo

    “But for the long form stuff, I have my blog, The Looking Glass. It's on Substack.”

    ▶ Watch at ~1:34:30approx.
  • SubstackUseswebsite · Substack Inc.

    “I have my blog, The Looking Glass. It's on Substack.”

    ▶ Watch at ~1:34:30approx.
  • Lenny's PodcastMentionspodcast · Lenny Rachitsky

    “You can find all past episodes or learn more about the show at lennispodcast.com.”

    ▶ Watch at ~1:36:00approx.

Spot an error or want something removed? Request a correction or removal.

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 1

Myth Buster27:30

Most Hot AI Companies Aren't Actually Using Data Well

Zhuo's contrarian take: fast-growing AI companies mostly run on good instincts and good vibes, not data. Historically, hitting 100M users took years - enough time to build logging, hire data teams, and instrument the business. Today's companies hit hundreds of millions in ARR with 2-10 people and none of that observability infrastructure, which bites them when growth eventually stalls.

  • Fast-growing companies get by on instincts and vibes, not rigorous data
  • Scale used to be slow, giving years to build logging and data teams
  • Today's tiny companies hit huge ARR without observability infrastructure
  • Growth always stops eventually; without observability you'll scramble to find the root cause

“I don't actually think a lot of the fast growing companies are using data well at this point.”

Julie Zhuo · 27:30

“these these companies are are totally getting by on just good instincts and good vibes.”

Julie Zhuo · 28:30
#data#analytics#growth#startups

Hot Take· 4

Hot Take09:00

Management Hasn't Changed - The Resources Just Went From People to Models

Julie Zhuo argues management has always been one thing: having a northstar and figuring out how to use your resources to get an outcome. What's changed is that those resources used to be people and are now increasingly AI models, each with different strengths and 'personalities' you have to develop an intuition for.

  • Management is fundamentally about achieving an outcome with the resources you have
  • The three levers are still talent/resources, purpose/goal, and process
  • Resources used to be people; now they are models with different strengths
  • You have to get to know models like you get to know people to use the right tool for the job

“management is just about in my mind having an outcome. So you want to get something done. That's the thing.”

Julie Zhuo · 09:00

“used to be people but now it's basically models and different models have different strengths.”

Julie Zhuo · 10:00
#management#ai#agents#leadership
Hot Take14:30

Drop the Job Titles - We Can All Just Be 'Builders' Now

AI lets each individual do many jobs that used to require a whole team - design, coding, data analysis - getting you from the 10th to the 60-70th percentile fast. Zhuo's call to action is to dissolve traditional role boundaries and simply call everyone a 'builder', a term Lenny says he's increasingly adopting too.

  • AI empowers one person to do jobs that used to need specialists
  • It won't make you top 1%, but it gets you from the 10th to the 60-70th percentile quickly
  • Teams can shrink to two people from any disciplines using AI to cover the gaps
  • Zhuo wants 'builder' to become the actual job title

“we need to dissolve the boundaries of these traditional roles.”

Julie Zhuo · 15:30

“we can drop all of these different role distinctions and call ourselves builders.”

Julie Zhuo · 16:00
#ai#org-design#roles#teams
Hot Take32:30

Diagnose With Data, Treat With Design

Zhuo's core philosophy for designers who resist data: data tells you where the problem or opportunity is, but never what to build. You diagnose with data and then treat with design through a creative process. She also broadens 'data' to include TikTok virality, tweets, and customer interviews - anything that reflects reality.

  • Data reveals if and where you have a problem, not the solution
  • Solving still requires a creative design process with no formula
  • 'Data' includes qualitative signal: what goes viral, what people post, customer interviews
  • Every great designer she's met is obsessed with understanding reality

“what you really want is you want to diagnose with data and treat with design.”

Julie Zhuo · 32:30

“data is not a tool that's going to tell you what you should build or what the solution is”

Julie Zhuo · 33:00
#design#data#product#decision-making
Hot Take1:05:30

Even Firing Someone Should Be a Win-Win

Zhuo thinks constantly about win-win. If you frame management as getting more out of people at their expense, you can't build a real strategy. Even telling someone they shouldn't be on the team can be win-win: prolonging a bad fit just extends their misery, and respecting them enough to say 'this isn't it' frees them to find where they'll actually be valued. The mental model changes how you convey it.

  • If better outcomes come at someone else's expense, that's not win-win thinking
  • The win-win frame makes strategy and difficult conversations easier
  • Keeping someone in a bad-fit role just prolongs their misery
  • The person can push back that it isn't win-win - which is useful information

“my getting better outcomes has to come at the expense of somebody else losing”

Julie Zhuo · 1:06:00

“And right here, what we got, this isn't it.”

Julie Zhuo · 1:08:00
#management#win-win#firing#leadership

Explainer· 2

Explainer42:30

Dimensionality: Every Strength Is Its Own Weakness

Zhuo shares her mental model that a person is made of infinite dimensions - like a fingerprint of skills, some top 1%, some below average. Because no single dimension is your whole self or your worth, you can hear feedback about a weakness without your identity feeling at risk. She notes strengths and weaknesses are often the same thing: being thoughtful and slow-on-your-feet are two sides of one coin.

  • See yourself as infinite dimensions, not one identity
  • No single dimension represents your true worth as a person
  • Every strength is its own weakness and every weakness is a strength
  • Her 'thoughtful' strength was the same trait as being quiet and slow on her feet

“every strength is its own weakness and every weakness is a strength.”

Julie Zhuo · 43:00

“you are all of these infinite dimensions in one.”

Julie Zhuo · 45:30
#self-management#feedback#growth#leadership
Explainer1:01:00

How to Deliver Hard Feedback So It Lands

Zhuo offers three tactics for hard feedback. First, establish an opt-in relationship up front where you both want to help each other grow, before anything goes wrong. Second, check your intention - if you're really trying to validate yourself or punish, it won't go well. Third, name your own nervousness out loud, which humanizes the moment and makes the person hear you.

  • Set up an opt-in 'help each other grow' relationship before problems arise
  • That opt-in does about 60% of the hard work of future feedback
  • Check your intention: feedback to punish or be right will never land
  • Say your nervousness out loud - vulnerability builds connection and makes it heard

“So when you first start working with someone, don't wait until something bad has happened.”

Julie Zhuo · 1:01:30

“Lenny, I'm so nervous right now is I want to give you some feedback and I'm really worried that it's going to impact our relationship”

Julie Zhuo · 1:04:30
#feedback#communication#management#leadership

Story· 2

Story17:00

Why Julie's Startup Deliberately Doesn't Hire PMs

At Sundial, Zhuo eliminated roles they assumed they'd need - like product managers. Her insight: when a PM exists, engineers delegate product thinking away as 'not their job'. Removing the PM forces engineers to own communication and user value themselves, producing better, faster work and happier employees.

  • Having a PM tempts engineers to delegate product definition instead of owning it
  • Smaller teams without dedicated roles push everyone to take responsibility
  • Same logic applied to front-end vs back-end: 'you're a builder, figure it out'
  • Early investment is slower, but well-rounded builders pay off long-term; still bring in a specialist for heavily specialized work

“we thought trai we would need a bunch of product managers.”

Julie Zhuo · 17:00

“communication's up to me. Figuring out how we get greatest value to users is something that is now strictly in my charter.”

Julie Zhuo · 18:00
#product-management#teams#hiring#org-design
Story1:16:00

How Julie Used AI to Build a Talking Raccoon for Her 6-Year-Old

In AI corner, Zhuo shares becoming an IC to make her kids' birthday gifts. Inspired by Eric Antonow - who surgically wired a microphone, speaker, and ChatGPT voice mode into a parrot - she built a talking raccoon for her raccoon-obsessed son. For her middle son who loves video-game parody songs, she built a Replit app that writes personalized parody lyrics she records herself.

  • AI made it fun for her to go back to being an IC and build gifts herself
  • The idea came from Eric Antonow's parrot wired with a mic, speaker, and ChatGPT voice mode
  • She built a talking raccoon for her raccoon-loving six-year-old - a huge hit
  • A Replit app writes personalized video-game parody lyrics she sings and records into an album

“what had happened is that he had hooked up a microphone. He like kind of surgically went into the parrot and added like a microphone…”

Julie Zhuo · 1:17:30

“I was like, I want a raccoon that can talk to him.”

Julie Zhuo · 1:18:00
#ai#parenting#replit#creativity

Q&A· 1

Q&A49:30

Fix Weaknesses or Double Down on Strengths? It Depends on Your Goal

Asked how hard to push on things you're bad at, Zhuo says it depends entirely on your goals. An IC deepening a craft may never need to become a manager. But suffering comes when your goals and your chosen path are misaligned - like wanting a giant mansion while only wanting to perfect your egg omelette. Get clear on the goal, then decide which skills you actually need.

  • How much to work on a weakness depends on what your actual goal is
  • An IC can deepen one craft and never need management skills if that reaches their goal
  • Suffering happens when your desired outcome and your chosen path don't align
  • Decide the goal first, then whether a given skill is required to reach it

“I also just want to spend all my time perfecting my egg omelette and then you're just like in this tension place”

Julie Zhuo · 52:00
#career#goals#self-management#growth

Takeaway· 5

Takeaway12:00

The #1 Manager Skill That Makes You Great at AI: Defining Success

The most valuable skill for managing both teams and AI is getting crystal clear on what success looks like. Zhuo connects this to why evals matter - they force you to make the objective criteria explicit. If you're unclear about success in the prompt, you won't get great work, whether from humans or AIs.

  • Alignment is hard because people hold different pictures of what success looks like
  • Boiling a vague high-level goal into something an agent can judge is most of the game
  • Evals matter because they force you to define objective success criteria
  • Getting clear on success is 'more an art than a science'

“that's why we have to write eval and that's why they're so important because they're helping us understand what is the objective criteria”

Julie Zhuo · 13:00

“if you're really unclear about what success looks like the prompt, you're probably not going to get the most amazing work.”

Julie Zhuo · 13:30
#ai#evals#goals#management
Takeaway22:30

The Underused AI Superpower: It's a Better Teacher Than a Human

Beyond just-in-time tools like Cursor, Zhuo highlights accelerated learning as an underutilized use of AI. She feeds online curricula into ChatGPT to personalize a program to how she learns, and tests understanding by explaining concepts back and letting the model critique her - something you couldn't do by pulling on a busy colleague's time.

  • AI teachers are often better than asking a busy human who has to explain everything
  • Feed a 12-week curriculum into ChatGPT and ask it to customize a program to how you learn
  • Different people are different types of learners (examples vs. abstractions); AI personalizes to each
  • Test real understanding by retelling a concept in your own words and letting the model critique you

“I think these days, you know, TAGBT or these other AI tools are better teachers.”

Julie Zhuo · 23:00

“I'll just feed it into chat PT and I'll say help me customize a program for me, you know, using the ways that I like…”

Julie Zhuo · 23:30
#ai#learning#chatgpt#tools
Takeaway37:30

Be Like the Willow Tree: Sturdy While Being Flexible

With the rate of change accelerating, Zhuo says a manager's job to manage change is harder than ever. Her metaphor is the willow tree - sturdy enough to survive storms precisely because its branches are so flexible. Management today means being sturdy while being flexible, and channeling that willow-like feeling when leading through uncertainty.

  • Managing change has always been the job; the rate of change is what's accelerating
  • People feel real fear and uncertainty about AI and their careers
  • The willow tree survives storms because it is flexible, not rigid
  • Leadership today is being sturdy while being flexible

“the willow tree is a very sturdy tree you know”

Julie Zhuo · 39:30

“today management is really about this idea of like be sturdy while being flexible”

Julie Zhuo · 40:00
#management#change#leadership#uncertainty
Takeaway57:30

Feedback Is a Daily Practice, Not a Twice-a-Year Review

Zhuo is passionate that we undervalue feedback by confining it to biannual performance cycles. What matters long-term is how quickly a team gets better, and feedback is the best tool for that. A team improving 1% a week compounds past a team improving 1% a month, even from a lower baseline. Feedback is a gift that reflects back what you can't see about yourself.

  • Don't reserve feedback for six-month review cycles - make it a daily practice
  • What matters long-term is how fast the team improves
  • 1% better weekly beats 1% better monthly even from a lower baseline
  • Feedback is a gift that reflects back reality you can't see yourself (the 'leaf on your head')

“feedback really in my mind ideally should be like a daily practice”

Julie Zhuo · 58:00

“a team that just gets 1% better every week compared to a team that gets 1% better a month is even if they start off…”

Julie Zhuo · 58:30
#feedback#management#teams#growth
Takeaway1:31:00

What Julie Teaches Her Kids for an AI World: Emotional Regulation

Asked what she's teaching her kids for an AI future, Zhuo says emotional regulation, above all. Humans still run on the same hardware we've had for thousands of years. Her fear is that AI makes everything more comfortable and becomes the ultimate shortcut - to relationships, to hard feelings - so she wants her kids to still choose hard, challenging things, because that's where real freedom and pride come from.

  • Emotional regulation and self-introspection are the top skills for an AI world
  • Our biology hasn't changed even as tools and environment race ahead
  • Her fear: AI becomes the ultimate shortcut around relationships and hard emotions
  • True freedom is choosing hard things and feeling pride in who you become - not using AI as a shortcut

“Emotional regulation is still really, really, really important. That's probably the thing that I think about the most in terms of what I want my…”

Julie Zhuo · 1:31:00

“to me, true freedom is you can pick the things that are hard and you can feel pride in becoming the the thing that you…”

Julie Zhuo · 1:33:00
#ai#parenting#emotional-regulation#future