20Growth: Meta CMO Alex Schultz on How All Founders Have to Change Their Marketing Playbook in a World of AI | Is AI Plateauing, What it Means if China Wins the AI Race and Why Zuck is a Generational Leader

Alex Schultz is the Chief Marketing Officer and VP of Analytics at Meta, where he has spent nearly two decades shaping the company's growth and marketing strategy. He has been instrumental in scaling Facebook, Instagram, and WhatsApp to billions of users worldwide. Alex is is also the author of Click Here: The New Rules of Marketing, the definitive guide to modern growth — available now on Amazon. AGENDA: 00:00 – Is All Marketing Actually Performance Marketing? 04:00 – When Did Facebook Have the Wrong North Star? What Did They Learn? 16:00 – Will AI Create Companies Run by Just ONE Person? 27:00 – Is AI About to Hit the Biggest Plateau Since Self-Driving Cars? 30:00 – Is China Secretly Winning the Global AI Arms Race? 38:00 – Does AI Kill Content or Supercharge It? 44:00 – Why Brand Marketing Is Harder (and More Important) Than You Think 47:00 – Will Glasses Replace Phones Forever? 51:00 – What Would Alex Do If He Were Sundar at Google Today? 59:00 – What is the Greatest Strength and the Greatest Weakness of Zuck?

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:04Speaker 00

  Performance marketing is all marketing. So people make this distinction that you have brand marketing and performance. You have brand and direct response and all marketing is performance marketing. I wouldn't have taken CMO if I thought brand didn't matter.

• 0:18Harry Stebbings?

  This is 20 Growth with me, Harry Stebbings. Now, I am so excited for the show today. I think we probably have the best growth leader I've ever had the chance to have on the show. He's become a dear friend since our first episode and now he has an incredible new book, Click Here by Alex Schultz. Alex is the CMO at Meta and the VP of Analytics. Today is this incredible wide-ranging discussion. We discuss everything from how AI impacts the future of marketing, the future of channel spend, the future of content. We also discuss open versus closed, China, the different worlds that are created by open versus closed models, Chinese models. This is a phenomenal discussion. Alex was incredible in the studio and I'm so excited to hear your thoughts. But before we dive into the show today, SecureFrame empowers businesses to build

• 1:03Harry Stebbings?

  trust with customers by simplifying information security and compliance through AI and automation. Thousands of fast-growing businesses, including Nasdaq, Angel List, Doodle and Coda, trust SecureFrame to expedite their compliance journey for global security and privacy standards such as SOC 2 and ISO 27001, CMMC, MIST standards and more. Backed by top-tier investors and corporations like Google and Kleiner Perkins, the company is among Forbes' list of the top 100 startup employers for 2024, G2's best software awards for highest satisfaction products, and a recipient of the 2024 Cybersecurity Excellence Awards, something I definitely never got in school myself. Learn more today at SecureFrame.com.

• 1:47Harry Stebbings?

  And once SecureFrame locks down compliance, Pendo makes sure your product truly lands with customers. Whether it's the software you build or the software you buy, your tech stack should be creating results, not creating roadblocks. Well, Pendo's no-code software experience management platform makes your software better with tools to see where users get stuck, guide them with in-app messaging, and constantly improving your UI. It's so easy that over 14,000 businesses use Pendo to increase revenue, lower costs and reduce risks. Businesses love the control, engineers love the freedom, everyone wins. Start for free today at pendo.io/20product. And after Pendo shows you what users need, Miro helps your team bring those ideas to life.

• 2:30Harry Stebbings?

  We all know that innovation is the lifeblood of any company, yet more than 80% of innovation projects stall before they even make it to execution. Why is that? Because once teams move from discovery to development, outdated processes, endless contact switching and misalignment slow everything down. Don't you hate it when that happens? Well, that's exactly why Miro built the Innovation Workspace. God, that's a cool name. A faster, smarter way to take ideas to execution. Now, at 20VC, we've worked with countless startup founders who know that speed is everything. And that's where AI-powered tools like Miro's come in. From AI sidekicks that act as a designer, analyst or engineer on demand, to instant AI summaries that condense meeting notes, product briefs and retrospectives in seconds,

• 3:15Harry Stebbings?

  Miro is built for teams that want to move fast. And one feature I love, the two-way sync with Jira and Azure. So when something is updated in Miro, it's automatically reflected in your development tools and vice versa. No more duplication, no more miscommunication, just seamless execution. So whether you're in product, UX, engineering, agile or IT, Miro's Innovation Workspace helps teams go from idea to outcome faster. Try it today at miro.com. That's M-I-R-O dot com. You have now arrived at your destination. Alex, I'm so excited for this, dude. I always learn so much from our chats. So thank you so much for joining me once again.

• 3:52Harry Stebbings?

  You're more than welcome. I always enjoy talking to you. The energy is fantastic.

• 3:56Harry Stebbings?

  And you know what? I never lack energy. And I'm so excited because we can skip the like, how did you get into tech? How did you get into growth? Which honestly, I get so bored of when I hear other shows. And so, I'm like, Jesus, we don't want to hear this. I want to start at the top, which is we all need a sense of direction and clarity. And that most often comes from a north star. You said before, you need one metric to rule them all. And if it's the wrong one, your company will drift.

• 4:20Harry Stebbings?

  Leader within a company today, I'm the founder. How do you advise me on how to select the North Star?

• 4:25Speaker 00

  Well, the most important thing is the North Star is not a metric. So the North Star is your goal. Your goal is like, what are you trying to achieve? When I joined Meta, it was connect the world online, full stop. And then the metric describes the goal. And so the single most important thing is to have absolute clarity on what your goal is. And then do the best you possibly can to describe that goal with a metric. And remember, a metric never perfectly describes a goal. A metric is always broken.

• 4:49Harry Stebbings?

  Can I ask you, I think we learn a lot from mistakes. I remember with 20VC, there was a time when we just solely focused on the numbers. And we went into this phase of like, actually, we can get glitzy people who don't really relate so much to our core, which is startups, venture, growth. And we had kind of famous people on and the North Star misled us and it was the wrong North Star. If I say to you, when do you think you had the wrong North Star or wrong goal structure? And what did you learn from that?

• 5:17Speaker 00

  I'll be honest, I think Mark is extraordinarily consistent. The biggest place which I talk about in the book is before I joined. It felt like the North Star wasn't clear within the leadership of the company. And that was the famous time when Mark nearly sold the company to Yahoo. Because it felt like the North Star to a lot of the executives was it's a startup. We need to make it big and exit and become millionaires. And for Mark, obviously, the North Star was connect the world online, do what he felt would make the world, you know, a better place. And that was at odds. And I think that led to a lot of the executive team saying, I want to sell to Yahoo. And Mark saying, I want to stay independent. And we turned over the entire executive team. And then out of that fire, which I think really forged the company.

• 6:02Speaker 00

  We came to much more clarity on what the North Star was. So everyone knew what they were signing up for.

• 6:07Harry Stebbings?

  When we think about startups, in particular, we have a huge startup audience, huge founder like early stage. Is all of this a bit of fluff? Or actually, is it fundamentally important pre-product market fit? At what stage does this become crucial?

• 6:19Speaker 00

  Setting a North Star is critical pre-product market fit. Metrics, pre-product market fit, I think the only metric you care about in.

• 6:30Speaker 00

  Consumer and SaaS is retention. Like, that's the only metric that you should care about is do you create a product that people keep using? A lot of bad things happen where people, they create a metric where it's MAU. And then they game the metric and they have terrible retention, but they have monster acquisition. And then at some point, gravity catches up with them because the acquisition can't overwhelm the churn. And then their numbers collapse. And you've seen that with a number of companies out there, right? So pre-product market fit, I wouldn't get too complicated. I'd have a clear north star, what your company is about and what you want to achieve. I would have retention, like you need to get to actually a product that's retentive. Otherwise, it doesn't have product market fit.

• 7:09Harry Stebbings?

  The hardest thing always for me in product is how do you build for a broad audience without alienating super users? Do you have any lessons on how to do that effectively? And anything that really stands out to you?

• 7:22Speaker 00

  It's a question of what period of scale you're at. So in the early days, you need to get something that is going to have a passionate retained core user base. And so that's super important. And so actually, there's no delta in the early days, like when you're starting your company, when you're getting going. You need to have that passionate retained core user base. As you get larger to tens of millions or more people using your product. Then I think you actually need to split into a growth team who thinks about the marginal user and thinks about getting people onto the service and the rest of the company who think about power users. If you look at Meta, and again, I described this in the book, it's almost as if you've read it. If you look at Meta, we had this problem where in Instagram, all of the Instagram team were power users of Instagram.

• 8:05Speaker 00

  And so a really good example here of the delta between what the growth team focused on versus the Instagram sort of core team focused on was notifications. So if you're a power Instagram user and you upload something, you're going to get so many notifications. You get a thousand likes or 10,000 likes or whatever it is. You are going to get so many notifications, your phone is unusable with notifications popping up on the phone. So we made the default for notifications notifications off and you had to choose to turn on notifications for likes. But if you're a new user and you've uploaded your first photo ever to Instagram, that first like is really meaningful. And it's kind of heartwarming and touching and probably for the first year or so of using the site, it just means a lot to you that you get likes. And so for them, you want it on.

• 8:49Speaker 00

  The general company, in my experience, at every company, whether you talk Amazon, whether you talk Google, whether you talk Meta, whether you talk Airbnb, they as employees tend to be power users of the product and they by nature optimize for people like them. The thing you need the growth team for is to make sure you remember there are people who are not power users of the product, who need to be acquired and who need to ramp up and they need to have a different experience. And the Instagram notifications example is my, my favorite example of that.

• 9:16Harry Stebbings?

  How does the structure of that growth team versus product team at Instagram work? Were they in the Instagram company? Was that outside of it? How was that structure in the relationship between the two?

• 9:25Speaker 00

  I recently had to testify to this at length in court.

• 9:29Harry Stebbings?

  So you're well prepped.

• 9:31Speaker 00

  Yeah. So, look, I think the way that we do it isn't the way everyone does it. We had a central growth team who would help WhatsApp, who would help Instagram, who would help Facebook, who, you know, now helps Reality Labs. And that central growth team had analytics and marketing with me, before it was analytics and marketing with Chamath. And then it had a product team and an engineering team, which today is with Naomi Gleit. So it's like Naomi does product and engineering and I do marketing and analytics and then we help people around the company. But what's critically important in a company with as many divisions as us and at the scale we are, is that Instagram actually has a product manager and a growth team inside Instagram. Because you don't want some team on the outside of your product, shipping loads of things into your product. You want the growth team to do the, the standard

• 10:16Speaker 00

  platforms that are necessary for the whole company. Things like SEO, optimizing the ad tech for the paid channels, things like email infrastructure. Like, there's a series of very obvious.

• 10:25Harry Stebbings?

  Foundational pillars.

• 10:26Speaker 00

  Foundational pillars that Facebook uses and Instagram uses and Reality Labs uses and everyone uses. And then each team has their own little team working on it. So it was that. We had a central team with expertise who helped everyone, basically led by Naomi and me, today under Havi, previously under Chamath. And then each of the individual products have their own growth team who we partner with where they do product specific things.

• 10:48Harry Stebbings?

  Each of the individual teams then. Within each product, what do those look like? How many people are in them? How do you segment between the different functions within them?

• 10:56Speaker 00

  I mean, these are actually pretty small teams. Even at the scale that we're at at these stages, like across the whole company, outside of the central team, I'd say less than a hundred people in growth. And then in the central team, a few hundred. So across our nearly a hundred thousand employees, we have less than one percent of the employees in the growth team.

• 11:12Harry Stebbings?

  So if you go into Instagram, in specific, their growth team, there's twenty five people?

• 11:17Speaker 00

  I don't know the exact numbers, but it's not huge. It's a small but mighty team. And that, I think, is right. Like, you can get a lot done with small, talent-dense teams. And I think growth is one of those fields where that's always been true. If you're in a very small company and you've got like two hundred, three hundred people, I would argue you shouldn't have a growth team. At that stage, I would argue the entire company should be working on growth. And at that point, no, you shouldn't delegate it away. Growing your company is probably the single most important job of the CEO. And so there's a transition period when you flip over from your entire job is growing the product, growing the company. That is your whole job to, actually, I've got a team that's working on growing the company. I need to be thinking about strategy and what's next and how we build the next product line and how we take things to the next level and

• 12:00Speaker 00

  all of these different things, maybe even sales, like partnerships, things. But that transition period is far later than most people do it. So many companies try and hire a growth team at like fifty employees and it's just silly.

• 12:11Harry Stebbings?

  I interview a lot of the best product and growth leaders and a lot say we're going to see the kind of collapsing of skills into one kind of superhuman, where you have kind of product, growth, eng in one person with kind of democratization of engineering, what you're seeing with kind of your cursus and your cognitions of the world. Do you think we will see this collapsing of individual into a all-round superhuman product and growth? Or do you think actually they very much remain in isolation?

• 12:37Speaker 00

  I mean, the question you've got to ask yourself is how weird is it going to get? Like, if we really are going to see advances in intelligence through artificial intelligence that are being suggested by the leaders of the field, then it's going to get very weird indeed and you're going to see entire companies be one person. But why would you stop there?

• 12:55Harry Stebbings?

  Do you buy that? I'm not being stupid, like entire companies be one person, really? You think Audi are gonna buy your HR tool where you don't have a CS team and you don't have a security team and your Fritz sitting in Munich unable to deal with customer success because the CEO wants to go for lunch with you to see if you're actually a good person. Oh, and by the way, you've got, you know, the SCG, the cricket ground in Australia who want to buy you too, but you've got to fly out to them. One person? This seems like the most ridiculous statement to me. Sorry.

• 13:25Speaker 00

  I don't think it seems like the most ridiculous statement to you. I think you're being cheeky. Yes, like absolutely. Hundred percent. I think that's gonna happen. The question is when, not if. The question is when, not if. Like, so, so I look at the book, Chip War, incredible book. If we follow the progress you're seeing, you see in Chip War, and I love it as an analogy of like, we knew what the next couple of breakthroughs were in chips. You know, from the point when people were drawing on circuit boards with felt tip pens, basically, and acid etching them, to the point where they were projecting onto them with essentially photo negatives at sort of photo scale, to the point where photo lithography became very, very small, to the point where we now are individually placing atoms and not using etching in all cases. Each of those steps forward, we knew two ahead, give or

• 14:08Speaker 00

  take, and the companies that were involved could say, we're gonna have these breakthroughs and we're gonna gain this capacity. AI is really similar. We were talking about agentic behavior two years ago. That's roughly what we were talking about. Engineers that were AI engineers, well, that was about a year and a half ago. Everyone was kind of talking about that. There's a load of stuffs, by the way, that's proposed in the lab. So I'm not saying in the lab's technique, I'm saying in the corporate environment. So like what we, it looks so similar to semiconductors to me that like, we've got a bunch of breakthroughs. We've got mega scaling data centers is gonna be big. You know, Google can now train in multiple data centers. It's something everyone's working on. So continental scale training systems, that's gonna be big. We know there are advancements coming from Nvidia in their

• 14:50Speaker 00

  chips. That's gonna be big. You had the Grok founder on here, who I think is absolutely fantastic, you know, TPUs and everything. He's really cool. And like the stuff he shared in your, in your podcast, you could tell, we know what the next few iterations are. And then in the lab, we've got 50 going. It's gonna get better for the next two years and we're really confident of that. Some areas you will literally have companies where it's one person and they are outsourcing and they're like, the finance is mostly handled by AI and their team, but they have a company and they're like, this is the email to the company. Even on your recent newscast, you were talking to your partners on there and they were pointing out how they were doing ninety percent of their legal themselves. And then they were saying three correct questions to the legal firm that supports them.

• 15:32Speaker 00

  And so they scaled themselves an entire in-house legal department with GPT. So yeah, it's gonna happen. But do we hit a plateau that means it doesn't happen in our lifetime? Maybe. And that's what we don't know beyond sort of two innovations out. Which of the things that are going on in the lab right now are gonna come through? I think no one knows that and anyone who says they think they know that. The OpenAI guys didn't know G P two was gonna be the breakthrough that it was themselves the day they launched it.

• 15:57Harry Stebbings?

  You said about agentic AI engineering being the things two years ago, we kind of knew and it was kind of bubbling up. And now we obviously see it kind of really comes to the fore. And then you said a couple, which I was trying to write down ferociously. But I don't think I did a very good job, but I just have to ask about them. You said about mega scaling data systems?

• 16:14Speaker 00

  Data centers. So like what we're doing in Louisiana, where we're building a data center the size of Manhattan or Sam with Stargate or these massive data centers where we're gonna keep going at scaling laws and we think it's, everyone's confident. It's just gonna release a new wave of capacity.

• 16:31Harry Stebbings?

  And this is like, hey, we are ramming compute down the throat because that will continue to increase and it doesn't plateau.

• 16:37Speaker 00

  The bitter lesson.

• 16:38Harry Stebbings?

  And so what does that mean?

• 16:40Speaker 00

  That means that the intelligence is gonna step up further in the next two years because these large data centers will come online.

• 16:45Harry Stebbings?

  Can data center development move at the pace of leadership demands?

• 16:49Speaker 00

  No, no, we are way lagging right now.

• 16:52Harry Stebbings?

  Why is that?

• 16:53Speaker 00

  Physics, like physics is just, there's a really good essay about the UK called UKfoundations.co. It's an amazing essay. There was a certain supply chain globally that existed for making data centers. People who could make the right steel, people who knew how to wire it, people who knew how to put in the fiber optics, people who knew how to do data centers. And there was also a certain supply chain in terms of energy available. This is why you see Elon Musk and us and others doing gas turbines or Microsoft saying. Bring Three Mile Island back online as a nuclear power station because you can't suddenly spin up a gigawatt scale generation of energy in one place. There's a bunch of just physics of things don't exist that are needed. Even if you can scale up the gas turbines, right now, there is a gap in terms of the generators.

• 17:38Speaker 00

  So you need to wrap the copper wire and you need to like build these massive generators. And there's only so many that are typically ordered a year. But now Elon wants generators himself and we want generators ourself and Microsoft wants generators themselves on our own sites so that you can have capacity that makes up. So there's a bunch of these things where the supply chain globally hasn't been built out to supply what's needed in the data centers. And so there's a bottleneck right now. Can humanity do it? Yes, it will just take humanity a few years to speed up in terms of building the supply chain to fill these things. And then we'll hit another bottleneck. Maybe TSMC can't make enough chips for Nvidia. Maybe we went through this period when the substrate was the bottleneck for Nvidia and TSMC wasn't the bottleneck. Maybe ASML can't make their machines because they're so complicated.

• 18:22Speaker 00

  One of the things people don't talk a lot about is because there's so much fiber optic in these things, switches, the availability of switches for the fiber optics is actually a problem. There's a fiber optic shortage globally because we're all buying thousands of kilometers of fiber optics. So there's loads of pieces in the supply chain to fill the building that are lacking not least energy, but all of these other pieces as well.

• 18:43Harry Stebbings?

  What is that duration period of wait? And what I mean by that is like, is it like Adam Smith's invisible hand? Increased demand means people build more very fast and it's fine actually. Or actually, is it an inevitable how we actually have to wait a long amount of time?

• 18:56Speaker 00

  I think on most of this, it's like an 18 month lag till the system catches up with the demand. And this is why we need the continental scale training systems. I mean, this is literally-

• 19:04Harry Stebbings?

  This was your second one, wasn't it?

• 19:06Speaker 00

  Yeah, but this has been discussed on your podcast before. I mean, you've had some incredible people talk about this. I actually feel listening to 20VC has made me better at my job in the space of infrastructure. You've had some really, really great people. Around the time you had Jonathan on from Grok, you had a few people on.

• 19:22Harry Stebbings?

  Yeah.

• 19:23Speaker 00

  I keep coming back to the trillion dollar question that you had with Khan, wasn't it, from Sequoia? That's a brilliant podcast. I've sent that around- So many time, this stuff is in your podcasts. The continental scale training question is, right now, we can put a one gigawatt data center in one place. The largest power generation on earth is like, I think, twenty two gigawatts or twelve gigawatts. It's the Three Gorges Dam. So you can only go up ten X from where we are right now. Conceptually, if you took the entire power of the Three Gorges Dam, which, I mean, maybe China will, I don't know, they're pretty aggressive. So the only way you can actually access more power to work with these data centers is to be able to train across multiple data centers and not do it all in one location. Historically, all of these AI systems had to be trained in one room in one building.

• 20:07Speaker 00

  And gradually, the infrastructure is being built to break up the training into multiple centers. I believe Google has actually was the first one to do it. It was within one data center area, but it was across two buildings, I think. But now everyone's talking about training across regions, so maybe a hundred kilometers between. And then the next question is thousand kilometers between. And then you could maybe, if you're Microsoft, have one thing sitting next to the Diablo Canyon, you know, nuclear plant, and another one sitting next to Three Mile Island. And now you've got two gigawatts, whereas there's no one place on the US where you could have two gigawatts, conceptually. So that's the sort of stuff that I mean when I say continental scale. So if you believe in the scaling laws paper and you believe it's going to continue, which clearly OpenAI is betting on that,

• 20:51Speaker 00

  Anthropic's betting on that, we're betting on that, Google's betting on that. So that's clearly a bet out there from the leading players that scaling laws will stay in place. Then you have to believe that continental scale training is going to come online because otherwise you can't get to the next level of scale.

• 21:05Harry Stebbings?

  You mentioned one final one, which was Nvidia. What was it precisely? I was writing so ferociously.

• 21:11Speaker 00

  Oh, Nvidia is such a brilliant company, although I think it's awesome that you've got companies like Grok coming along. I think that's really cool on the inference front. But Nvidia are getting better. Yeah, there are issues with the cutting edge and the odd chip maybe is slow to deliver, or there are rumors about this, that and the other. But fundamentally, what Nvidia is putting out now versus five years ago is a way better chip. I believe Nvidia is going to keep getting more competent as a company and they are going to keep building better chips. And so that means that for the same amount of power and the same amount of buildings, you're going to be able to do more. With the chips you put in them because Nvidia is going to build better chips. And so that will be another thing that's a really good tailwind. So those are three very obvious tailwinds that are coming. There will be others, there are algorithmic based ones, there will be capabilities.

• 21:52Harry Stebbings?

  What's the biggest algorithmic based one?

• 21:54Speaker 00

  So I think there are two interesting things. There's deep research piece that's going back and forth. The deep research piece is really interesting where you've got the the AI debating with itself. And we have not reached the end of that becoming better. There's a second piece, which is this whole space of reinforcement learning for post-training that has been so material for 2025 on the success of many of the top models. You put that together with all that's been going on about the synthetic data or all that's going on with the very high quality data coming out of professors and so on and so forth with the data labeling companies. There's something really interesting in that space as well. And this is the stuff that is breakthroughs that are out to the public right now. But they haven't reached the maximum of what's going to be extracted from those breakthroughs.

• 22:38Speaker 00

  Like nothing I said right now would be groundbreaking to anyone. In fact, they'd probably be like, he's an empty, he doesn't know what he's talking about. They're really advanced people, but clearly they're they're breakthroughs, but they're not maximized breakthroughs, like everyone isn't doing them maximally. There's a huge upside to the head on those. So when I say there will be entire companies of one person. Instagram was seven people and got to tens of millions of active users. I don't see why there won't be the next Instagram that's one person and gets to tens of millions of active users. In fact, I think it's inevitable.

• 23:08Harry Stebbings?

  When we combine all of these things, largely it leads to increased performance, leads to increased efficiency. What's that mean for us? How weird does it get?

• 23:17Speaker 00

  It's a question of whether we get to AI superintelligence in five years. I mean, genuinely, that's you hope you achieve the culture and you don't achieve AI 2027. I think the three essays that I point everyone to right now are AI 2027, which is the max doom scenario, I think. But he was very smart and he put together, I think it was called AI 2026 and.

• 23:35Harry Stebbings?

  And what does he put out basically?

• 23:36Speaker 00

  In 2026, he AI 2026, he basically lines up what's going to happen in 2021 over the next four or five years. And he gets a lot of stuff right. In AI 2027, he says, we're going to reach breakout superintelligence in the next two, three years. Which I think is aggressive and I personally don't believe, but it's it's interesting. And he says it goes wrong. That's what he says in AI twenty twenty.

• 23:56Harry Stebbings?

  And breakout superintelligence is?

• 23:58Speaker 00

  Self-training AI. So synthetic data, the AI can write its own code, the AI can self-improve. It has full agentic capabilities.

• 24:06Harry Stebbings?

  Why do you not think that that's possible? We see cases of that in different model behavior now. And Rappler is like deleting code for itself.

• 24:13Speaker 00

  I just think it's really aggressive for the curve that we're going to be on, but maybe maybe we're there. So it's.

• 24:17Harry Stebbings?

  AI twenty twenty seven.

• 24:18Speaker 00

  So that's like, in my opinion, that's max doom. The other end of the spectrum, you've got Sam's essay, The Gentle Singularity, which you can tell in it, there's already a little bit of argy-bargy between the companies by how he references us. But I think it's brilliant. And I think it's a really positive maximalist mindset. And it's like abundance, consumer surplus, a better world. Everyone's going to have intelligence too cheap to meter. It's going to be great. And there will be, you know, an age of abundance. That's pretty awesome. And then you've got, I think, the mid case, which is Dario's essay, Machines of Loving Grace. And that one, I think, is the mid case where he definitely says there's some doom chance, but there's also some maximalist positive chance. You know, you sit in the middle there. And and so I think there are three views from those three essays that are really interesting.

• 25:01Speaker 00

  Mark has just added his superintelligence view to the mix, which I'd encourage people to listen to his video or read his his post about that.

• 25:08Harry Stebbings?

  Where does it sit in the scale of Sam, Dario and AI twenty twenty?

• 25:11Speaker 00

  I think it's it's closer to Sam in terms of it's not max doom, it's more max positivity and benefit for the world, which I think humanity is capable of achieving. Otherwise, I'd be a much more sad person and stressed out. But if you achieve that, then in five years, it's all synthetic data. So who knows? I mean, like, it's we're in a very, very crazy time.

• 25:28Harry Stebbings?

  Do you think founders' need for capital dictates their narrative around this? And what I mean by that is if you look at Demis or Zuck, traditionally, they've been a little bit more nuanced, saying, hey, it'll take longer than we think. It always takes longer than we think, and actually more thoughtful in that respect. They don't need the money. They don't need someone to come and give them another twenty billion, another ten billion, another five billion, whatever it is. And then you look at your Sams, who I love, you love, great people, and a lot of the other founders in the foundational model space, they're like, it's coming sooner, it's coming sooner. They need the cash. They need to sell that because that's that's what drives cash.

• 26:02Speaker 00

  One hundred percent. One of the things that has me being a little bit more maximalist on the timeframe is Demis and Mark have shifted. I completely agree. I completely agree. So like you look, the three essays I quoted, those are the extremists. And it's the range of views of the extremists. But then you look at Demis, you look at Mark, you look at the big company folks, you look at Satya, you look at Sundar. These folks are saying it's it's coming. The question for me is, do we hit any plateaus? Because right now, if we don't plateau, things are crazy. I would be shocked if we didn't hit a plateau on the way, but but yeah.

• 26:34Harry Stebbings?

  I remember sitting at a dinner with you and it was fascinating. You said about like, hey, do we hit a plateau? And then you aligned it to self-driving. And it really stuck with me because I remember ten years ago being in SF. And it was like, oh, my God, self-driving is here. Everyone's going to be jobless. Like, eight million truck drivers are gone. And then actually, it was only two or three years ago, I might be butchering totally, but where kind of Waymos were really being put out in numbers.

• 26:57Speaker 00

  Yeah, no, I mean, I think it's exactly the same. The interesting thing is, I think if you look at China right now, you've got some self-driving stuff where there are trucks actually doing long-distance journeys. That's my that's my understanding. I have not been to China. It's a little bit dangerous for me to go to China.

• 27:10Harry Stebbings?

  Please don't go.

• 27:11Speaker 00

  Yeah, I'm not going to China. But my understanding is you're starting to actually have some self-driving trucks and so on, really doing like long-distance journeys. That's been tested in the US. Waymo is now actually on the road. Davos twenty sixteen. There was all of this conversation about self-driving cars and trucks are going to put all these people out of work. None of it, by the way, about how it's going to save fifty thousand lives a year, which I think is pretty important as well, because we we sacrifice fifty thousand people a year to deaths on the roads. Hopefully, we won't have to in future because self-driving will be so much safer. But now, Waymo just works. I ride it every time I go to San Francisco and it can go on the freeway too. And it has a safe failure mode for the freeway, which before its failure mode was stop, which is why it wasn't doing freeway journeys. Now it plans out enough it can pull over.

• 27:55Speaker 00

  Actually, we all panicked in twenty sixteen. The thing we panicked about is now here and nobody's panicking.

• 28:01Harry Stebbings?

  Well, nobody's panicking because it hasn't had the infiltration slash adoption from the core group of people that it would make unemployed.

• 28:08Speaker 00

  Yeah, the future's here, isn't it? Evenly distributed yet.

• 28:11Harry Stebbings?

  Yeah.

• 28:11Speaker 00

  One of my absolute favourite sayings. You know, you see that with chat G P T and Claude and even Meta A I and Gemini. Like when people are using those, their behaviour in search changes dramatically.

• 28:21Harry Stebbings?

  I thought it was fascinating Sam's description of how different age segments use it. And it's like, you know, young people use it as like the operating system for life. And you'll put screenshots of WhatsApps in, how do I respond to my boyfriend's question? And then others who use it as a replacement for Google when they're in their forties.

• 28:35Speaker 00

  Mm hmm. So true.

• 28:37Harry Stebbings?

  Yeah, sorry.

• 28:37Speaker 00

  Yeah, I'm too old.

• 28:38Harry Stebbings?

  No, no, you're not. You look wonderful.

• 28:40Speaker 00

  I like to think we're, we're hopefully, we're going to be the sort of Google or Apple or Microsoft of this generation of tech companies and not the Yahoo or eBay.

• 28:49Harry Stebbings?

  I think you'll be fine. Zuck is a beast.

• 28:52Speaker 00

  He's very good. He's so good.

• 28:53Harry Stebbings?

  He's so good. What makes him so good? Um. And has your definition of what makes him so good changed over time?

• 28:59Speaker 00

  That one's hard to answer. He's got a superpower of learning. He listens in meetings more than he talks by a lot. He asks smart people to give him input before he makes decisions. He goes out and spends an awful lot of time learning. I think his superpower is learning.

• 29:12Harry Stebbings?

  You mentioned Chinese long haul routes, all of these models that come out of China and the rate of distillation that's happening. I am sitting here definitely in Europe thinking, wow, it takes about four hours to get an espresso. We're fucked. And the U S, I think, is doing an amazing job in the kind of A I arms race, so to speak. When I see China and the race and the distillation quality that they're getting, I get worried. Do you think we radically underestimate Chinese A I capabilities today?

• 29:39Speaker 00

  I don't know who you mean by we. Certainly on the west coast of the U S, we don't underestimate them.

• 29:44Harry Stebbings?

  You don't think so?

• 29:45Speaker 00

  Not at all. We talk about them all the time. In my mind, Europe's in this interesting place where you have two leading groups. Europe's not in the race right now. I mean, maybe it can find a way to be in the race, but I don't think it's in the race. You've got the U S and China as the two groups that are leading in A I. They're both leading and I think, you know, China had a couple of things that jumped ahead. Yes, they used a lot of distillation, but they did jump ahead. They had some fundamental innovation in there and a lot of distillation. This is the Chinese model. That's how they work. It's how they work with Philips. It's how they work with Open A I. It's how China works. We could end up with the Chinese model, which. Is definitely using computers to censor and control the population. And that is the way the world is if China wins. Or we can have the US model, which is freedom with the drawbacks of freedom.

• 30:30Speaker 00

  But people having the freedom to say what they want and not being executed if they talk about Tiananmen Square. So I do actually think there is a major delta between the world states that happen if China wins versus the US wins. And I think there's a really good argument for why Europe should care that the US wins.

• 30:46Harry Stebbings?

  Is distillation wrong, really?

• 30:47Speaker 00

  No, distillation is totally, I mean, in my mind, it's fine. It's it's just part of the game.

• 30:51Harry Stebbings?

  Should Europe be pressing for a place in the race? Or should we accept? And I don't like to be defeatist or doomist. We are so far behind now at this stage. It's too late to catch up and we can win in other segments.

• 31:03Speaker 00

  No, I think Europe could catch up. It does require changes. It really does require changes in. In the UK, it would be planning and energy, probably more than anything. And there are some privacy laws, which the UK has this thing going on with Apple right now. That's pretty, pretty public at this point where they secretly told Apple to put a backdoor into their encryption. The UK has some privacy laws, which are very difficult for big tech companies to build stuff in the UK. If they're willing to tell Apple to put a backdoor into encryption. That's why we came out in support of Apple. That's why other companies have come out in support of Apple in that situation. Despite the fact as companies, we've had a lot of disagreements. We supported Apple in that. So Europe has some of those same issues.

• 31:43Speaker 00

  Because whenever people hear someone like me say privacy, they're like, oh, you hate GDPR, you want to suck up user data. We do actually care about our users' data and there are some issues in Europe around state ability to take people's data that are pretty aggressive. The UK Apple one is in the public right now, so that's one I'll refer to, but there are others. So Europe needs to think about that. The US doesn't have that. The China has it with bells on. But I think Europe could catch up. Look at what Elon Musk has done. Honestly, look what we did. Two years ago, we were nowhere. Last year, we were right at the vanguard. We had a bit of an iffy year this year, but I'm sure we'll get back in there again. So I actually think like it's totally open. You could come from behind and really accelerate. There are big brains in Europe. Look at the Mistral folks and everyone. The right couple of policy decisions and Europe could accelerate very fast.

• 32:28Speaker 00

  It would take a huge amount of will. To do it, but Europe's capable if it had the will.

• 32:32Harry Stebbings?

  When you say you had an iffy year this year, what does that mean to you?

• 32:36Speaker 00

  It's pretty clear Lama was not like Lama four wasn't leading edge where Lama three was. So, you know, our big release this year was not a slam dunk, but our big release last year was, and I'm, I'm pretty confident our next big release will be again. So.

• 32:48Harry Stebbings?

  Can't just start cool like an all hands and be like, this was not good.

• 32:52Speaker 00

  No.

• 32:53Harry Stebbings?

  How's that relayed? Like, how does the team cajole around like, yeah, let's get fired up?

• 32:58Speaker 00

  I mean, I think people are really smart. They look at the evals. They understand what's going on. And so I don't think anyone looked at the evals and said, oh, wow, it went really well. It went, okay, it's a lot better than Lama three was, but it's not as great as it could have been. I don't think it needs an all hands to tell everyone. We, you can look at the data.

• 33:13Harry Stebbings?

  Do you think the dominant models of the future, wherever they are, open or closed?

• 33:17Speaker 00

  I have no idea. It could go either way.

• 33:20Harry Stebbings?

  Do you like open more?

• 33:21Speaker 00

  I mean, we've said that when it gets to the cutting edge, like the absolute cutting edge, the bleeding edge, there we don't think you should be necessarily releasing open source. And we'll think about that. So I think at the most extreme capabilities, I don't know that open source is the right answer. High level, we're a company that's built on open source. We couldn't exist without MySQL, PHP, all of these things, Memcache, Open Compute, all of those things were critical to us existing as a company. So yeah, we like open. We have said at the very bleeding edge, we would need to reevaluate if open is the right strategy because like it is important for us to be sensible about safety.

• 33:56Harry Stebbings?

  We said about like how weird all of this gets. I do want to kind of slightly bring it back to kind of brand as well and also channels. And it's just like, when we think about that and how weird it gets, I create content. One thing that I'm really worried about Alex is the, the decreasing value of content in a world where supply is infinite.

• 34:14Speaker 00

  Yeah, but attention is not infinite. You are going to get to the point where AI is going to transform lots of industries, including marketing, including content. And there will be a bunch of AI content, but in the end, the creative spark is probably the last thing that gets automated. Like right now, at least when I'm hearing it and I might be wrong, but no one's got really great ideas for how you make the AI creative. Everything it does is pretty much derivative. And so. Like two things will end up mattering, I think authenticity and creativity. And so if you look at books like Player of Games and the culture series of novels as a thought experiment around what could that AI future look like, you're in a place where people value human skill and value actually seeing a human authentically playing a game

• 34:59Speaker 00

  is the theory, one of the theories in Player of Games. And when you're doing a deal, you want to shake hands and look the other person in the eyes, like you said earlier. And so there will be places where that's still really, really, really important. And so from my perspective, what's going to be scarce is attention. The content you have with actual experts in their fields, the content that you have will be prized because it's not derivative stuff that's like created out of nowhere. It is real conversations with the people who actually made real decisions. Other things are going to be valued, that would be the authenticity. Other things will be valued for their creativity, where artists do creative work. It's not like you can't get a print of any Monet, but people really love standing in front of the beautiful water lilies in Paris and just looking at them,

• 35:43Speaker 00

  even though they could have a print of them at home or they could look at them on their computer screen. Being in the presence of that authentic piece of art, that creativity plus authenticity means stuff to people, even when you could have infinite watercolors of Monet's lilies today for free.

• 35:58Harry Stebbings?

  I also find there's like always additive hallucinations. And what I mean by that is like in conversations where something is actually really interesting, in a human interaction like this, I jump on it and go, actually, that's super interesting that you said that. And it leads to a great conversation that is additive. In a model that is progressive, you wouldn't have the ambiguity to move off script into separate sideline discussion. Does that make sense?

• 36:20Speaker 00

  Yeah, absolutely. And I think that's creativity that you're providing.

• 36:23Harry Stebbings?

  But I do, I do worry though, that in a world of scarce attention, feeds that are more and more crowded. I like AI content just as much as I like any other content, honestly. I don't attach.

• 36:33Speaker 00

  And that's now.

• 36:34Harry Stebbings?

  And that's now.

• 36:34Speaker 00

  It's going to get better.

• 36:35Harry Stebbings?

  It's going to get better. I don't attach less value to it. And so I'm just concerned that actually ours is an increasingly tough world to stand out.

• 36:42Speaker 00

  Yeah, but I also have faith in you or the next generation of you will come along who finds a way to blend it with AI and use the tools. Like, I really agree with this point that, again, unless things get really weird, you're not going to be replaced by AI, you're going to be replaced by someone doing your job using AI.

• 36:56Harry Stebbings?

  If I'm a CMO today and I'm thinking about like my channel makeup, does AI change how I think about my channel makeup?

• 37:03Speaker 00

  No, because I I think the tenet of the book, the principle is.

• 37:07Harry Stebbings?

  The book?

• 37:08Speaker 00

  The book, yeah, I've said it fifty times.

• 37:09Harry Stebbings?

  What's the book called? Let me just.

• 37:11Speaker 00

  The book is called Click Here by Alex Schultz.

• 37:13Harry Stebbings?

  Oh.

• 37:14Speaker 00

  It is about online marketing.

• 37:15Harry Stebbings?

  I love it.

• 37:16Speaker 00

  Being really good at it. Thank you. Thank you, Eric.

• 37:19Speaker 00

  The principle, this tenet in the book is the principles are constant, even if tools change. So the principle you're going to look at is incrementality. So look, the channels are going to evolve. There's one world where search gets replaced by paid for interfaces. If you're using GPT today, which you said you are, so you're using ChatGPT. You said that your search volume had gone down.

• 37:40Harry Stebbings?

  Died.

• 37:40Speaker 00

  That is true of a lot of people. That is true. And that's paid for. There's no ads next to it. On the other hand, a bunch of people are using Google and using Gemini. And their search volume, according to Google, has gone up. And they're seeing more ads next to it. So does a paid model like OpenAI win? In which case, search advertising could go away. Does a free model like Gemini in search win? In which case, search advertising could grow tremendously. Or is there a freemium in the middle with some paid and some free? Some ad supported and some not, that is how we get to a hybrid. In any of those cases, yes, search will change as a mix of your channel. But the thing you need to look at is volume and incrementality you get from search. And that is constant. And right now, people are flexing their budget between search and social.

• 38:26Speaker 00

  They're flexing it between affiliates and retail ad networks. They're moving their budget around based on results. And so I think as a CMO, you need to look at incremental results. You need to make sure you're leveraging the cutting edge technologies. And then you should go where you're getting incremental results.

• 38:40Harry Stebbings?

  I saw in the latest results, Matt's latest results, that basically AI was increasing conversion within ad spend for buyers by like three or four percent.

• 38:50Speaker 00

  In multiple ways. There's multiple different tools that are doing that. Yes.

• 38:52Harry Stebbings?

  Got you. And so how does AI increase conversion in spend for buyers?

• 38:57Speaker 00

  So one specific example that we've released a paper about is the click target. So the click target that you use to get people to click on your ad in a reel. Companies can use the built-in click target at the bottom that is associated with a product in the reel that you want to buy. But they can also add in stories and I believe in reels now, but if not now, in the future, they can add another click target of a sticker that is a click target. And so they can select from a range of stickers we produce to have a click target in their reel or in their story, I think, and you can click on it and then you go through. What we've done is we've trained a modern multimedia model on all of the click targets that have been used. And then we've looked for what are the learnings for what will make the best click target when you intersect it with an advertiser?

• 39:42Speaker 00

  And we have created custom stickers that advertisers can now pick that have been optimized by AI to be the right sticker for their ad. When advertisers take advantage of that, they get six percent boosting click-through rate and a commensurate increase in conversion rate. That's one of the technologies that is currently percolating through our advertiser base. Everyone hasn't adopted it yet, a few have. So when you look at that three or four percent or whatever we said in the earnings call, well, there's just one example here where it's six percent that isn't fully adopted yet. There are a bunch of other technologies like that that we've built.

• 40:11Harry Stebbings?

  When you build them and sell them to buyers, you're at Cannes. Is everyone just like, oh, my God, this is great. AI, it's going to make me better and I can sell it back to the boardroom. Is that the response you get from enterprise buyers on the Cannes festival side?

• 40:25Speaker 00

  Well, so there's a range of outcomes. So let's talk about two very impressive companies. Both of them, I think, are excellent, which is why I'll talk about them. Coca-Cola, incredible company. Nike did this massive transition to digital first, right? And it went horribly. And so Nike, obviously, you know, have changed CEO. I really like John Donahoe, he was my old boss at eBay. But whatever, they've turned over CEO, they're changing what they're doing, all these things. Coke have changed from being vast majority offline, just, I think, five years ago, just before the pandemic, in terms of their ad buying, to vast majority online now ad buying. And nobody knows, they just did it quietly. But when we present to Coke, hey, there's this better click target for you to do online conversions. Rightfully, I would assume, this is not a conversation I've had with them, so not breaking an NDA.

• 41:10Speaker 00

  But if I was them, if I was their CMO, I'd be like, I sell Coke cans offline. Why would I need a click-through rate? For someone to buy online. It's not the majority of their business. And so those CMOs would be like, ah, that's not really the right thing for my business, but thanks for telling me about it. Another one, Booking.com are incredible. They're the leading travel company globally. They are ruthless. They are efficient. They are brilliant. I respect them deeply. So they're the first travel company to really make social media work, and they've made it to the extent that their CFO has talked about this on their earnings call that they've finally cracked the code. Both of those are brilliant companies and both of those are doing the right thing. But if I present something that optimizes click through to drive conversions online to a company that mostly sells offline, a company that mostly sells online.

• 41:54Speaker 00

  Guess what? The direct to consumer online people will react differently to the mostly through third party and mostly offline people, and that's actually logical. So it isn't like I get a uniform response at can.

• 42:04Harry Stebbings?

  And the Coke moving to online, they still are doing brand marketing with that then, aren't they? Cause they're not like, they don't expect me to go and click, oh, I'm going to buy a 12 pack through this click through. It's still brand.

• 42:14Speaker 00

  And their online is brilliant. There were two campaigns they did that I thought were absolutely fantastic. One was called Coke Creation, where creators were being asked to play with the Coke logo and Coke was co-creating with anyone who wanted to do something cool with the Coke logo. And if you look into it, there are some really cool things that people did. I mean, I'm a sucker for flowers and there was one that was all flowers, but it was incredibly creative. And the other one they did was Share a Coke, which I think if I remember correctly revolves around the Cokes that have people's names on them. And people posting pictures of them with their Coke with their name on, which a year or so ago you saw a lot of online. That was a deliberate move by Coke. And so, yeah, those are brilliant brand campaigns.

• 42:53Speaker 00

  They've moved from the box of television and thinking about the 60 second slot or the 30 second slot and the tyranny thereof, and they've thought outside the box. And it's been a really successful thing, especially in places like Brazil and India, these big growth markets for them.

• 43:08Harry Stebbings?

  One of the best shows that I've done, well, definitely one of the most widely received was that Nick at Revolut. And he said his single biggest mindset shift over the time with Revolut was actually brand, importance of brand. He didn't think it was important. It was performance, performance. If I can't measure it, fucking useless. How do you? Advise CMOs on the importance of brand marketing today and when it becomes important as well. Also think there's a lot of founders who listens to this where it's like, brand marketing, fuck that. I've got VC dollars, I need to spend them efficiently to grow.

• 43:38Speaker 00

  The point I'll make is, performance marketing is all marketing. So people make this distinction that you have brand marketing and performance. You have brand and direct response and all marketing is performance marketing. I wouldn't have taken CMO if I thought brand didn't matter. It is far harder than direct response to do brand marketing. It is far, far harder.

• 43:57Harry Stebbings?

  Why?

• 43:58Speaker 00

  Harder to measure, you don't have a conversion action, you need to move what people feel about you. Longer term, it's not an instant payoff. It was Meta and come on, I was catching a falling knife when I became CMO. It was not a, and now we're, we've actually moved up tremendously in terms of brand sentiment. You think where we were five years ago and think where we are today. People don't look at this stuff, obviously I do, and we were literally a falling brand and now we're a rising brand.

• 44:23Harry Stebbings?

  When we last did the show, you were like, and this is now Fran, it's like, you were like, I think it's actually been pretty successful with Meta and the transition. And at the time, no offence, to the wider populace, it was still in a more like questionable chasm. And now it's like, yeah, totally. It's amazing how that changes. Question then, does name actually matter? Or is it just consumer patience to adjust?

• 44:46Speaker 00

  Oh, both matter. I mean, name matters because the Facebook name was blocking people from understanding we did anything but Facebook. Meta, the company produces WhatsApp, Meta, the company produces Instagram, Meta, the company produces Oakley Meta glasses and Ray-Ban Meta glasses and Quest headsets. And if we said Facebook, all anyone thought was legacy social media. That was all they thought. And so it was necessary for us to be a, when we sell to advertisers, we were in a place where we were selling them Facebook ads, which is super confusing for them. Am I buying Facebook or am I buying Facebook and Instagram? And now Facebook and Instagram and WhatsApp. And so there was something really important about the name, blocking people from understanding we did more.

• 45:28Speaker 00

  So now most advertisers are saying they buy from Meta and they buy on this, that and the other property by Meta. That's been a huge. Shift for us. Ray Ban Stories were called Ray Ban Stories because if we called them Ray Ban Facebooks, I don't think they'd have sold. And they weren't Ray Ban Instagrams because they wanted to be bigger than just the app. They wanted to be company level, but the company was called the same as one of the apps. That's why we called them Ray Ban Stories. Second, we had Meta, we could call them Ray Ban Metas, and that made a difference. So all of those different things are really unlocked by removing the blocker of the Facebook name and then saying we have something to run towards, which is technology beyond Meta, beyond where it is today. Starting with, we believe in the future, the metaverse will be there. And I know people don't talk about it all the time

• 46:13Speaker 00

  anymore, which honestly annoys me because like we are getting those breakthroughs as we speak. The Orion Glasses last year allow you to augment your daily life exactly as we put in our launch video for the Meta brand, being the future of what the metaverse will be. Yes, they're not available to consumers yet. Hold this space and see where we go with that. It really mattered for us to talk about beyond social media, which we wanted to do. So that was super important.

• 46:40Harry Stebbings?

  Are the glasses not the best and most efficient data acquisition machine for real world data capture and labeling? It's one of the biggest problems, which is real world, obviously, data capture and labeling. And if you gave me five pounds a month or gave people five pounds a month, I bet they would be willing to submit all of their data for training, labeling, you name it. Is that not the best way to do a tester and capture millions and millions and millions of inputs of data?

• 47:07Speaker 00

  That's a good idea. It isn't currently something we're doing, but it's a, it's a perfectly reasonable idea. Yeah, I mean, there's stuff like that that's a hundred percent possible. Once you have glasses and if you look at what Sam's talking about with Johnny Ives or what even Google's talking about with their deal with Warby Parker, you can see we're in a place where people are like, aha, this is a really interesting form factor opportunity. And if AI is always going to be with you, you're going to need something, whether it's AirPods, whether it's Meta Glasses, whether it's a pendant hung around your neck. I'm not so sure about, we'll see, but like, you're going to need something.

• 47:39Harry Stebbings?

  Do you think we will have phones in five years' time?

• 47:41Speaker 00

  You're going to need something lightweight, whether it is AirPods, whether it's glasses, whether it's. A pendant, a brooch, like humane tried, like you'll need something lightweight. And then you'll need a puck to offload compute to. And I don't see why that puck shouldn't have a touchscreen. So I would be surprised if phones go away. Because you're not going to put huge batteries into glasses or AirPods or a pendant or a brooch. And you're not going to do all the compute in something with a small battery. So you're going to want to have a lightweight connection to some kind of puck that you offload things to. So yeah, I think something like that will exist. Will it be a phone as we understand a phone today? That I don't know. If you look at what the Orion glasses can do. There's fewer and fewer reasons to take the phone out of your pocket.

• 48:24Speaker 00

  If the cost can be brought down on the wave guides and the brightness can be brought up on the lasers. Which over five years, you've got to believe is going to happen.

• 48:31Harry Stebbings?

  How does SEO change in a world of foundation model?

• 48:35Speaker 00

  People are already starting to do it. It is AI engine optimization or AI optimization. There are companies working at this, which I'm not allowed to fund.

• 48:44Harry Stebbings?

  Which I did fund, I invested in one called peak.ai.

• 48:48Speaker 00

  Exactly.

• 48:48Harry Stebbings?

  Yeah. Amazing.

• 48:50Speaker 00

  Amazing, and it's clearly going to happen and it's some of the same people. And yeah, actually I got on the phone with them, didn't I? Yeah, it's some of the same people and there's really smart ideas.

• 48:59Harry Stebbings?

  Is the TAM for that bigger than the TAM for SEO? Because one of the reasons why, honestly, I said no to the founder before we invested. But I said, listen, like, you know, the biggest SEO tool is what, like Semrush? Like, what, a billion five after like seventeen years? For a venture fund, that's just not very interesting. Is it a much bigger market?

• 49:17Speaker 00

  It should be a much bigger market because the agents are going to take actions. I mean, the agents are already taking actions for people. And so in a world where there are actually actions being taken at the endpoint, it's not just presenting a link to somebody. Who's basically done a, you know, navigational search. It's really, do I buy this from Amazon? Do I buy it from Target? Do I buy it from Walmart in the US? That's a very big part of the conversion funnel that now needs to be optimized. So the question I'd say is what's the ambition of these firms? If their ambition is, I just want to make sure what's ever said in ChatGPT, Meta AI, Gemini, Claude, whatever. Is the name of my client. Well, that's a pretty small ambition and that will be a pretty small outcome.

• 49:58Speaker 00

  If the ambition is, oh my God, AI is going to change everything in terms of AI agents will actually take the actions for you. And I want to impact what actions the AI agents take themselves. That is a big total addressable market. So is the team thinking big or thinking small? And is the team quality enough to expand as the capabilities of AI expand? This comes back to what Sam said on your podcast, which I have used again and again and again, which is if you skate to where the puck is today on AI, we will bulldoze you. That was what he said.

• 50:30Harry Stebbings?

  Steamroll.

• 50:31Speaker 00

  Steamroll.

• 50:31Harry Stebbings?

  Steamroll.

• 50:32Speaker 00

  Steamroll.

• 50:32Harry Stebbings?

  I have been sent by so many friends a picture of Sam in a 20 VC jumper and I'm like, oh, so proud. And it goes startups, we will steamroll you. And I'm like, ah.

• 50:41Speaker 00

  Ah, so it's burned in your mind, not in mine. And I've been misquoting it. But yeah, we'll steamroll you. Whereas if you skate to where the puck is going to, you could do something incredible. This is exactly an example of that. If you're skating to a puck of like AI agents with full agentic behavior, even controlling computers, making full purchase decisions for consumers, that's a huge TAM. If you play for where we are today, tiny TAM. And also probably you'll get steamrolled.

• 51:06Harry Stebbings?

  If you were Sundar today, what would you do? You've got the golden goose that produces billions and billions and billions. What would you do?

• 51:13Speaker 00

  I think he's, I mean, he's done some stuff that's really good. So like, they are willing to disrupt themselves, which is awesome. Gemini at the top of search is a scary thing to do. And they've done it. Building out the cloud business and being in the position they are in cloud as well, pretty amazing. It's just going to be fascinating. Can they make the transition or not? They've got all the components and they seem willing to cannibalize themselves. So it seems like they're doing a lot of the stuff right, but I think it has to be be willing to cannibalize yourself. Don't, don't get caught in the innovator's dilemma. That is the single most important thing for them to think about because, I mean, look at us, right? So five years ago, we were 100% content shown to you based on people you follow or accounts you follow.

• 51:53Speaker 00

  100%. People don't, don't understand, like TikTok was tiny. It was the beginning of the pandemic. Like it was very early on. So let's say five and a half years ago. Today, we are 70% of engagement is watching stuff you don't follow. 70%. And by the way, it's video and back then it was photo. We have been completely disrupted by AI as a company. Who we are is totally different, totally different. And I don't think anyone quite realizes. How drastically Meta has been disrupted by AI already. The majority of what we are today is not what we were five and a half years ago. And we went through that massive dip in our stock price, if you remember, which was a deeply painful experience. I don't want to be self centered about this. It was more painful for the people who got laid off, but

• 52:36Speaker 00

  trust me, it wasn't the job that I thought I signed up for at the little startup, you know, 20 years ago. Very large layoffs, very painful experience, really changed the company as part of a moment that changed the industry. Find some of that was either hiring in the pandemic, but a ton of it was also like we hit this air gap where we were doing exactly the right thing of bringing in this new field of reels or shorts as YouTube call it or whatever, but we didn't have ads on them that worked at all. We put this huge amount of money into infrastructure. Everyone kind of blames RL. We were building out massive infrastructure at that time for the AI ranking that we needed. And honestly, you got to give it to Mark. I'd like to think we're a great leadership team.

• 53:14Harry Stebbings?

  And he was pummeled for it.

• 53:16Speaker 00

  And he was pummeled for it and he stood up and did it. And that's what Sundar needs to do, in my opinion, is Google might be going through its TikTok moment right now. And do they do what we do, which has come out stronger? And I think they can. I mean, my career is based on Google and I'm a big believer in them as a company. It's a hard transition to make, but they've got a lot of the right decisions happening. So I'm hopeful for them, but it's it's hard.

• 53:38Harry Stebbings?

  What disruption did you most doubt and what did you not see?

• 53:43Speaker 00

  I don't know if it's a disruption, but I never got crypto. I still don't get crypto. And then mobile.

• 53:47Harry Stebbings?

  Why don't you get crypto?

• 53:48Speaker 00

  I just don't really understand it. It still feels like a Ponzi scheme to me, which I know it makes me an old fogey, but I don't get it.

• 53:54Harry Stebbings?

  I just feel it as like an alternative store of value, which is most interesting actually for large companies now. You have one of the Fortune 500 which stores crypto. I think you'll have all of them store crypto in the next two years.

• 54:05Speaker 00

  Maybe I just don't understand finance. So that's probably the big one. I don't, I don't, I still don't get it. I feel silly.

• 54:11Harry Stebbings?

  In Facebook's issue, was there one where you're like, oh, I don't know about that decision. It doesn't mean you disagree with it, but you're just like, oh, I'm not sure.

• 54:18Speaker 00

  I didn't agree with us doing the audio rooms thing. The same time we were sticking reels at the top of everything, we also stuck audio rooms at the top of everything. And audio rooms failed.

• 54:26Harry Stebbings?

  Oh, it was Clubhouse competitor.

• 54:28Speaker 00

  They weren't the only ones, but there was a lot, but a lot of the tech companies did it. Twitter did it as well. Like everyone was trying audio rooms. So that one I was like, I don't see it.

• 54:35Harry Stebbings?

  Are you guys quick to kill things? When audio rooms comes out, are you guys like, uh-uh, month one, done?

• 54:40Speaker 00

  Compared to other companies, yes. Compared to our value of Move Fast, probably no.

• 54:46Speaker 00

  But I mean, we move far, look, yeah, we move faster than pretty much anyone.

• 54:49Harry Stebbings?

  What did you most want to do, but you couldn't do?

• 54:51Speaker 00

  Make every button green, big green buttons. People click on big green. The cover of my book is a big green button. My entire career is about big green buttons. And I'm forced to use blue buttons at Meta.

• 55:01Harry Stebbings?

  Do you like this AI integration into WhatsApp?

• 55:04Speaker 00

  Oh, the WhatsApp one's really good.

• 55:05Harry Stebbings?

  You like him?

• 55:06Speaker 00

  It's really good and it's successful around the world.

• 55:08Harry Stebbings?

  It is?

• 55:09Speaker 00

  Yeah, yeah, it's the most successful integration we've got.

• 55:10Harry Stebbings?

  What do people use it for?

• 55:12Speaker 00

  Same things they use AI for everywhere. I mean, it really is just the range of the gamut. People use it for productivity. They use it for help on their homework. But they also use it for just like chatting through, just thinking through what things they want to do, or if they're bored and asking a question like, you know, oh, why is that thing called Osnabruck Street? Why do we have that in London? And then I just like talk to it on the glasses or message on WhatsApp and it replies. So it's kind of what you see people, the range of what you see people use AI for. And yeah, WhatsApp is a very successful integration for us.

• 55:40Harry Stebbings?

  You ready for a quick fire, my friend?

• 55:42Speaker 00

  Yes, as long as everyone buys my book.

• 55:44Harry Stebbings?

  As long, yeah, it is a big green button. They'll all buy it for that.

• 55:47Speaker 00

  Click here.

• 55:49Harry Stebbings?

  Of the hyperscalers, which are you most impressed by?

• 55:52Speaker 00

  Microsoft, the way they have revolutionized themselves in the last 15 years, the prescience of Satya. When he became CEO, he said, three things I think are really interesting in tech, XR, quantum and AI. And he still says, three things are really interesting in tech, XR, quantum and AI. And he positioned himself so brilliantly to take advantage of those. I mean, they're leading in quantum. We don't talk about quantum very much, but Microsoft is leading of the big players in quantum. It's very impressive how they've done it and how they've stayed relevant over 60 years as a software company. I was just like, it's incredible.

• 56:27Harry Stebbings?

  Is quantum not more scary than AI when you look at its ability to have serious security infringements?

• 56:34Speaker 00

  I mean, quantum has a lot of things to be scared of. As I said earlier, started my career in the semiconductor physics department at Cambridge. Herman Hauser was knocking about the place and was investing in the leader in quantum then, and he still is today. It's kind of crazy to see, see what he's done and become amazing man. So yeah, so I've thought about quantum a lot. I think the difference with AI is if you do the P doom maximal, so the max doom kind of view of the world is AI is intelligent and can make decisions and those decisions can be bad and harmful and we could be fighting it. Quantum is just a change in the world. It's a big change in the world, but it breaks encryption. There are ways for us to handle encryption being broken, at least quantum encryption. And so it's a much more solvable problem, whereas if a malign AI superintelligence actually

• 57:18Speaker 00

  existed, that would be less solvable. So of the two, I'm more scared of AI.

• 57:21Harry Stebbings?

  Which part of Matt's business is most underappreciated? There are so many elements.

• 57:28Speaker 00

  I think business messaging. So click to messaging is north of a $10 billion business.

• 57:33Harry Stebbings?

  So what is this? Click to messaging?

• 57:35Speaker 00

  Exactly. It's underappreciated. You can buy an ad on Facebook or Instagram and now on WhatsApp in the status tab where you click on the ad and a messaging thread opens and the messaging thread can be Instagram Direct, Messenger or WhatsApp. It is a huge business outside the United States. Massive.

• 57:51Harry Stebbings?

  So what do people use it for?

• 57:53Speaker 00

  People don't understand what's going on in Southeast Asia, Latin America, India. Like we're in this place where Brazil and India are two of our very most important countries for revenue. Like they're, they're huge and four years from now they will be very, very important. And in those countries, what do people use as their messaging app? Mostly WhatsApp. And what do they do on there? Well, they do commerce on there. In many ways, they skip the in the desktop, they skip mobile and they went straight to a messaging first world. Business messaging and what people are doing to transact on WhatsApp.

• 58:22Harry Stebbings?

  How do you capture the value there though? Cause I would argue back to you there. I do most of my deals via WhatsApp as a venture capitalist in London.

• 58:31Speaker 00

  Which is great.

• 58:32Harry Stebbings?

  Which is great, but there's no value capture for you.

• 58:34Speaker 00

  That's fair. I don't think Twitter captures much value of sort of VCs posting on Twitter about deals, but they get engagement from it.

• 58:40Harry Stebbings?

  No, they pay me now.

• 58:41Speaker 00

  There you go.

• 58:42Harry Stebbings?

  Nuts.

• 58:43Speaker 00

  So I think like the world that you're talking about is like actually quite normal for a very high end thing like what you're doing. What I'm talking about is the local hairdresser or you want to order some stuff from the local shop or whatever. And for the protection that you would get from having like a verified account and this, that and the other, you'll pay to you. Use WhatsApp potentially to do it or give a chunk to WhatsApp. That's one concept of stuff that actually happens in things like WeChat and so on. We don't do that right now. Way we make money right now is, one, businesses want to start that thread where the commerce happens. Businesses pay us for an ad to start the thread. So that's, that's one. We make money off ads and it makes the ads convert at a higher rate and it pushes up the CPMs of the ads. So that makes us over $10 billion a year. Yeah, yeah, yeah. It's a, it's actually a very meaningful percentage of our business.

• 59:27Speaker 00

  Very meaningful. And the last number we set out was over $10 billion a year, and I think that was a year ago. So you can imagine it's a fast growing business and this is a fast growing bit of a fast growing business.

• 59:36Harry Stebbings?

  What do you think that will be in three years?

• 59:38Speaker 00

  Tens of billions. And then two, we make money because, say, I now make the first attempt when someone wants to recover their password on Instagram. The first thing I do is I text them on WhatsApp. I don't text them on SMS. So for example, there's that. And so businesses also pay us for that service.

• 59:57Harry Stebbings?

  Ah, the verification.

• 59:58Speaker 00

  Hmm. And that's a north of a billion dollar business. But, but yeah, no, I mean, it's a really big and it's a growing business. And in fact, we sell it in partnership with a bunch of carriers around the world. They are actually our partners.

• 1:00:07Harry Stebbings?

  What has been the most, you can choose, fun or challenging brand exercise to work on? But you've done some cool shit, dude. You've done like race car, F1 car stuff. You've done some amazing stuff. What's been the most fun or challenging you can choose?

• 1:00:23Speaker 00

  The most challenging was the rebrand. I rebranded with a great team. Andrew Sturck was my right hand on this and then my Mayumi was my left hand on this. But rebranding the company in six, seven months as a new CMO was a drastic. And, and a lot of people have the skepticism you had about the meta brand internally and externally. And it was my call and I, I, first thing I said to Mark when I got the job is we have to change the brand of the company. It's so confusing in so many areas where it's Facebook and Facebook. And nobody understands we do more than just the Facebook app when we talk about the company, like it's blocking us from our ambitions. And it took a few months, but Mark came back to me and said that he wanted to rebrand the company. And then I was like, oh no.

• 1:01:03Speaker 00

  The consequences of my actions, they've caught up with me. And then it was really hard. So that was the most challenging one.

• 1:01:09Harry Stebbings?

  If I could give you a magic wand and change one aspect of your job, what would it be?

• 1:01:13Speaker 00

  Right now, I have a great team. I would probably say I would like a larger brand budget for the company. If you look at Meta and you look at our 10K and you look at Google's 10K and the other 10Ks, apart from Apple, whose marketing is paid for by the carriers, we have the lowest percentage of operating income that we spend on the stuff, the media that we disclose in the 10K. And I think we're probably a bit under indexed on that. So I think I would expand the brand budget because there are certain things you can't do with just putting a banner ad in WhatsApp. How big is the budget? I don't feel that's it. I don't think I can say the precise number, but what's in the 10K is low billions. What would you most like to do that you haven't done in terms of a brand creative?

• 1:01:58Speaker 00

  Wrap the statue in Rio de Janeiro green with WhatsApp. No, we've done that. We've literally done that. It was threads, actually, but we we literally wrapped with projectors, but we did. We turned Christ the Redeemer into threads. No, you didn't. We absolutely did. A hundred percent. That was a joke. You know, we did it. That's gonna be a great visual. I have done. I have done a number of things like that. I mean, we wrapped the Empire State Building green, as I think you know. That was WhatsApp. So Threads was was Christ the Redeemer. Empire State Building was WhatsApp. You know, I would love us to do a million AI creatives or sort of do a prompt and then try and do a million AI creatives for a million segments. Try and like actually genuinely go one to one on creative. I've got to ask, me and JC have got a bet on this. Is Threads a success?

• 1:02:43Speaker 00

  Massive slam dunk. And I was skeptical. Seriously? I was hugely skeptical. I actually spent a bunch of time yesterday with Adam and Cox talking about the numbers on threads and also with Jenny, who leads Instagram Analytics. There's some details I'd like to see, but that thing is blowout success. I'm so sorry. Why does no one in our circles use it then? Yeah, I mean, it's just, it's a bit more normie. Well, firstly, it's a huge success in like Japan and Taiwan. Let's start with that. Japan, interestingly, is a text-first country. That's why haikus come from Japan and all that stuff. It's a text-first pseudonymous society, and so it's a really great product market fit for Japan. And we've always struggled. Actually, Twitter is bigger than I think Facebook and Instagram in Japan. It's insane. So Threads is now really thriving in Japan. So firstly, segments that really care about this is different.

• 1:03:28Speaker 00

  Secondly, you know, in general, I think VCs don't tend to use Facebook very much either. And they all kind of act like Twitter is Facebook when they're talking about Facebook in public. You can kind of hear it. You listen to anyone talk about Facebook who's like a VC or a politician or policy elite, we'd call them, or a highly engaged news audience, we'd also call them. They don't really use Facebook or Instagram. They use Twitter. And they think Twitter is Facebook, which is not true. It is the most engaged community on Twitter, I think, is VCs and tech people. Yeah, the normies aren't the VCs and tech people and they use Threads, but VCs and tech people use Twitter. What's been the biggest product fail in the last three to five years? I mean, there's been tons. That's the great thing about the company. The company tries a lot of stuff.

• 1:04:13Speaker 00

  We talked about Audio Rooms. Audio Rooms was a pretty big one. Instagram Shops didn't go desperately well. That wasn't a huge success. Why did that not go well? Well, I think we did it focused on the US and TikTok Shops are successful outside of the US and UK. But they're actually not very successful in the US and UK, huge in Southeast Asia. But in Southeast Asia, click to messaging is huge. And actually, our live video shopping is huge. Live video shopping combined with click to messaging as a combination is a massive business in Southeast Asia. I think it was just designed and launched for the wrong area. The first place we put reels was in stories. That didn't work. That was a failure, but then we moved it. So the great thing about Meta is we try a lot of things and we faceplant a few times. But we move fast and we move on and we learn and then we do something better.

• 1:04:57Speaker 00

  And I'm actually like super proud. We have a bunch of bumps because it means we're a company that's actually pushing hard and trying things. We're in a field where the number of runs you get for hitting a home run in baseball isn't one. If you slug it out of the stadium, you can make a million runs, not just one run. And so at bats really matter. At bats really matter and you don't want base hits in our industry. You want to be knocking out the park. And so we get at bat very often and that means we will have failures. But that's actually okay because in our industry, it matters to have lots of attempts. And that's why you need to.

• 1:05:33Harry Stebbings?

  Increase the speed as fast as possible so you can have more and more at bats.

• 1:05:37Speaker 00

  Which is why Move Fast is such a core part of our company and why it's important to come back to the beginning. You have a clear North Star when you get to be a larger company so that people can run without escalating things and having to resolve things with the CEO all the time. You set a clear North Star, a clear metric and some clear guardrails and teams can run without bothering you.

• 1:05:58Harry Stebbings?

  Penultimate one, is Europe in a stronger or a weaker position in five years' time?

• 1:06:02Speaker 00

  Stronger on absolute basis, weaker on a relative basis.

• 1:06:05Harry Stebbings?

  Bringing it back to North Star and the book, what is your North Star with the book and what would you most deem a success?

• 1:06:11Speaker 00

  About every forty years, there has been a great marketing book. You look at Clodsey Hopkins with scientific advertising, Rossa Reeves with reality and advertising, David Ogilvie with Ogilvie on advertising. My North Star is I would like to become the textbook for this era of advertising, which is, I know hugely, it can be seen as hugely arrogant. But I think you don't do one of these things without having an aspiration like my North Star is this becomes the textbook for modern advertising for the next period. I really want people to see it as if you want to understand the channels that grew up since the brilliant Ogilvie on advertising was released. Use my book. That's what I want and I want people to see it as that.

• 1:06:49Harry Stebbings?

  Alex, this has been so much fun, Stu. Thank you so much for putting up with my very wayward questions. I had a huge amount of fun and you've been amazing.

• 1:06:57Speaker 00

  Thank you very much for having me and buy my book.

• 1:07:01Harry Stebbings?

  I do just want to take a minute to say Alex has become a dear friend since we did our first episode. The book is incredible. Seriously, check it out. It was my favorite read this summer. You can find it easily on Amazon and Click Here by Alex Schultz. If you want to see this episode, you can find it on Spotify by searching for 20VC. But before we leave you today, Secure Frame empowers businesses to build trust with customers by simplifying information security and compliance through AI and automation. Thousands of fast growing businesses, including NASDAQ, Angel List, Doodle and Coder, trust Secure Frame to expedite their compliance journey for global security and privacy standards such as SOC two and ISO twenty seven thousand and one, CMMC, NIST standards and more.

• 1:07:43Harry Stebbings?

  Backed by top-tier investors and corporations like Google and Kleiner Perkins, the company is among Forbes' list of the top 100 startup employers for 2024, G2's best software awards for higher satisfaction products, and a recipient of the 2024 Cyber Security Excellence Awards, something I definitely never got in school myself. Learn more today at SecureFrame.com. And once SecureFrame locks down compliance, Pendo makes sure your product truly lands with customers. Whether it's the software you build or the software you buy, your tech stack should be creating results, not creating roadblocks. Well, Pendo's no-code software experience management platform makes your software better with tools to see where users get stuck, guide them with in-app messaging, and constantly improving your UI.

• 1:08:28Harry Stebbings?

  It's so easy that over 14,000 businesses use Pendo to increase revenue, lower costs, and reduce risks. Businesses love the control, engineers love the freedom, everyone wins. Start for free today at pendo.io/20product. And after Pendo shows you what users need, Miro helps your team bring those ideas to life. We all know that innovation is the lifeblood of any company, yet more than 80% of innovation projects stall before they even make it to execution. Why is that? Because once teams move from discovery to development, outdated processes, endless contact switching, and misalignment slow everything down. Don't you hate it when that happens? Well, that's exactly why Miro built the Innovation Workspace. God, that's a cool name.

• 1:09:11Harry Stebbings?

  A faster, smarter way to take ideas to execution. Now, at 20VC, we've worked with countless startup founders who know that speed is everything. And that's where AI-powered tools like Miro's come in. From AI sidekicks that act as a designer, analyst, or engineer on demand, to instant AI summaries that condense meeting notes, product briefs, and retrospectives in seconds, Miro is built for teams that want to move fast. And one feature I love? The two-way sync with Jira and Azure. So when something is updated in Miro, it's automatically reflected in your development tools and vice versa. No more duplication, no more miscommunication, just seamless execution. So whether you're in product, UX, engineering, agile, or IT,

• 1:09:52Harry Stebbings?

  Miro's Innovation Workspace helps teams go from idea to outcome faster. Try it today at Miro.com. That's M-I-R-O.com. As always, I so appreciate all the support and stay tuned for an incredible episode coming on Monday with the one and only Matty at 11 Labs.