20Growth：ElevenLabs背后的66亿美元增长引擎｜ElevenLabs为何没有PM｜用7部分发布打法碾压所有发布——Luke Harries，ElevenLabs增长负责人

Luke, I am so excited for this, dude. Listen, we've been fortunate enough to get to know each other. I so appreciate our relationship, by the way. First, that's a heartfelt start, which is unusual for 20 Growth. But thank you so much for joining me, man.

Yeah, thanks so much for having me. Excited to be here.

I was just chatting to Mati before this on my walk around the park, and he said, “Ask him first: how did we first connect in terms of you and Mati, and how did you first get to know about ElevenLabs?”

1. From Hackathons To ElevenLabs

We met when we were 19, and we did a hackathon at Cambridge. We turned up—it was me, Mati, and this guy called Filip, who organized it. He's Polish, and he basically brought together his smartest friends and me to compete at this hackathon.

My top tip for any young people listening is: do hackathons with your smartest friends. What happened with that hackathon is, one, we ended up winning the Microsoft prize. We each won an Xbox and immediately put that on eBay, so we got about £300. Two, it landed me my first job at Microsoft. Then I also got to meet Mati.

Mati and I stayed in touch. Pretty much every 6 months, we'd catch up and riff on startups. Scroll forward about 7 years—so, this is a year and a half ago—and Mati's like, “Hey, Luke, I've just started a new company. Do you want to go catch up?”

I was angel investing then, and we were swimming in the Hampstead Heath ponds. When I tell Americans this, they're like, “Oh, ponds? You mean a small, dirty puddle of water?” I'm like, “Yeah, but it's actually quite nice.” So we're swimming in these ponds, and he taps the degrees. He's like, “Oh, it's 11 degrees. No way. My company's called ElevenLabs.”

He takes me through this grand plan of what he's going to do with ElevenLabs, which is, step 1, they're going to build the world's best audio AI models, and step 2, they're going to sell it to everyone. For people listening on their way to work, for creators making audiobooks and voiceovers, and for developers making AI voice agents.

I was like, “That is a terrible go-to-market plan. That's absolutely terrible.” I was angel investing then, but I was like, “This is not something I want to invest in. What do you mean, step 1 is you're going to build the world's best AI audio model—something no one else has ever done?”

Anyway, so I didn't invest. And that's the—

Oh, Luke, this is the moment where you're like, “This is a friend check.”

Yeah.

You know, the supportive, “I don't believe in it—”

Yeah. I should have.

“—but I just—I’ll give you $10K or something.”

I really should have. I mean, that $10K would've done very well.

Anyway, we stayed in touch, and then he phoned me up about 6 months later, being like, “Luke, we've just hit 1 million users. It's going incredibly well. We're looking for someone to lead growth.”

That hackathon was a lot of fun. I met Mati, and also Filip, the guy who organized it. He founded Wordware, which had the largest raise from YC ever. I think it's about $40 million. So, yeah, that was a great hackathon to do.

No fucking way.

Yeah.

I didn't know that was the same Filip. I love Filip.

Oh, yeah, Filip's awesome. That's still my LinkedIn cover photo: me, Mati, and Filip hacking over this computer when we were 19. I've not changed it since.

Oh, wow. I met Filip recently, and he is fantastic.

You have to have him on the pod.

Yeah, that's a very good shout.

There's a couple of things I just want to unpack. You very kindly scrolled forward 7 years. Most of my American—

Yes.

—guests don't, and they painfully tell me about their second and third years at Stanford and every class they took in between.

I do want to touch on one thing that Mati also told me to touch on, which is Fella, the company you started—

Yeah.

—before joining ElevenLabs. What are some of the big takeaways from that when you reflect on it, and how did it shape how you think about growth and product building?

2. Fella Taught Patience

Fella was a wild journey. It was me and my friend Richie, and we just wanted to do a startup. We applied to YC. We got into YC with an AI developer tool, and we quickly realized we were trying to do browser-based automation, where you automate tasks on people's behalf. You'd be sending an email, and it would autocomplete it.

But this was back in 2019, before LLMs were really a thing. The Transformer paper had just come out, but GPT-3 didn't even exist. So we'd gotten to YC, we'd tried building a prototype, and we were like, “Oh, this just isn't going to work.”

As we realized this wasn't going to work, COVID hit. We were there trying to brainstorm the next billion-dollar idea while also seeing this impending wave of what everyone thought was a killer virus crossing the world.

We set up a COVID testing clinic in San Francisco. We were some of the first people to realize COVID was going to hit. We were following Balaji on Twitter, so we were like, “Okay, how do we actually set up all this testing?”

We set up a drive-through COVID testing clinic. We were very European. We were like, “Of course, we're going to do it not-for-profit. This is a pandemic.”

We found a company, also in YC, which was doing testing for, I think, sepsis. They decided to pivot into COVID testing, and we distributed all their first tests.

One key lesson, which is probably a bit niche, is: don't do beta tests for pandemic testing. We set up this drive-through clinic and posted in Bookface, saying, “Hey, we're doing the first 100 tests for free. Just come through. We want to check that all our operations work.”

We accidentally ended up with an entire car park full of San Francisco tech billionaires turning up in their Lamborghinis and Ferraris to get the first COVID test.

That was a lot of fun, but we were like, “Okay, not-for-profit.” The pandemic was hitting. It turned out the company we partnered with took a different approach. They were some of the first people to do this testing, did incredibly well, and became one of the fastest companies ever to reach $100 million in revenue.

That company became Curative. Fred Turner is an incredible founder and is now building one of the largest health tech companies in Texas.

Absolutely fucking nuts. I remember we had the Hyrax founders on.

Yes.

You know Hyrax?

Yeah, yeah.

They made money by turning a coffee shop into a COVID testing center—

Wow.

During COVID, because Hyrax was up and running.

Yeah, yeah, yeah.

It was for-profit—they didn’t do it as a not-for-profit—and it was a fantastic business for them.

Yeah. Yeah, yeah.

That is absolutely nuts.

So we did that detour, and then we decided that we did actually want to build a big business. We started exploring different directions, and Richie had really struggled with binge eating and his mental health. I think we actually spoke to you at the time about some of this stuff.

Yeah.

We were going really deep on the men’s binge-eating and men’s weight-loss space. We set up a digital CBT clinic around how to help people who were struggling with these eating issues.

The tricky thing with digital CBT is that it’s great for lots of people, but from a business point of view, it really only lasts about 6 weeks. You can help them get to a great place, but we wanted to have a much bigger impact and build a much bigger business. That’s when we were exploring the health tech business models that could really grow and make a lot of sense.

It seemed that you needed a hardware piece or a medication piece, and we thought, “None of these quite make sense,” until we read the GLP-1 paper the day it came out. This is semaglutide, and we thought, “Wow, this is the biggest thing in medicine since penicillin.”

We raced to set up a clinic. We thought, “How do we help people get access to this fantastic medication, which reduces people’s body weight by 15%?” We managed to get to about $300K in revenue, which was great, and then it flatlined.

This was the first company I’d really started. I was still pretty young, and truthfully, I think I was just too impatient. I was like, “Oh, we’re only on $200K a year of revenue. I want to have a much bigger impact and do a technical product. I think we probably need to pivot to B2B.”

Richie was like, “No, Luke, I think we just need to focus as is. The wave will come.” So I stepped back, joined PostHog, and moved on. But Richie has done an incredible job, and that company is now doing over $30 million a year in revenue, helping thousands of patients, so that’s been really cool.

The big lesson for me is to commit for the long term and be patient. Ozempic had a much, much bigger wave afterward. We were actually quite early on.

Fucking nuts.

Yeah.

Also, the importance of market timing.

Yes.

If you’d done Fella 5 years before and there were no Ozempic, there’s no business there.

Oh yeah, it wouldn’t have been a thing.

I totally get you. Fascinating. Dude, I have to skip to another bit that you said there, which is that you were in the lake when he was pitching you: “Mati, this is for ElevenLabs,” and it’s like, “We’re gonna be everything for everyone.”

This sounds bad. I like a tight ICP where we can market to it, know which channels we’re selling through, and craft a product for them. How has a horizontal product offering worked so well when it is completely counter to the narrative that we’re always told?

3. The Horizontal Growth Engine

I 100% agree. When I worked for PostHog, part of the magic was that you have 1 ICP, which is a product engineer, and you’re selling multiple products into it. You’re selling product analytics, session recording, and feature flags all in one. Each new product you add gives you another entry point. It enables you to integrate the products to do completely new use cases that couldn’t be done before.

We don’t have any of that, in some ways. With ElevenLabs, we have the best AI audio models for turning text into speech, speech into text, making sound effects, voice cloning, and voice isolation. This is a very horizontal, foundational layer, which we sell as an API. That’s your first product: the developer product.

Then you can sell that API at scale. You can sell it to enterprises, and you can build workflows around it, doing conversational AI. That’s fantastic. But the team is super ambitious. Mati and Piotr literally want to build a bigger company than Google. They want to build something absolutely massive.

Then they see, “Oh, we could actually build a reader app for your phone to read articles.” So that’s now a product. We’re also building a creator platform. This text box, in which anyone can type text, is now creating whole audiobooks and voiceovers.

The interesting thing is that it’s actually really working for us. I think the reason we’ve managed to get it working is because we started with these incredible audio models. All these people come naturally to us. We’ve then been able to raise money, hire incredible founder-type people who really own these individual products, and grow them.

So yeah, it’s definitely not the norm. I’m not sure I’d recommend it, but for ElevenLabs, it seems to be working really well.

When I listen to you, it sounds fantastic.

Yeah.

I’m also going, “Whoa, whoa, whoa.”

Yeah.

Focus. Reader apps—

Yes.

Communities. How would you advise founders on focus when it comes to product expansions?

If I were to advise people, it would be: choose your 1 ICP and build all the products around it so they compound. However, that’s not actually what we’re doing. Instead, we’re building discrete products, and the way we tackle that is we basically shard the company.

We have a consumer app-focused team, which has an entire growth team with a growth marketer, paid ads, and community. We have creator growth, with its own paid ads, affiliates, and so forth. We have developer marketing, which is standalone.

And all of these sharded teams have their own growth elements to them?

Yes. Yeah, exactly.

So there’s not a horizontal growth team that sits on top of them.

There is. The structure we’ve set up is that we have a horizontal growth team of channel specialists. We have a head of performance marketing who used to lead performance marketing for Shopify. We have an SEO expert who used to do most of this for Canva. We have incredible individual specialists.

But within each team, we have a product growth lead who’s responsible for it. They’re basically the CMO for their product, and they’re thinking about all the metrics, the activation, and the awareness. Then they have their own sub-teams, which are specialized just for them.

Wow.

Yeah, it’s pretty funky. Our enterprise marketing team will be 20 people by the end of this year, and that’s just for enterprise marketing. Separately, our mobile app will be its own 5- to 10-person growth team.

Wow. What was this like when you joined? And if you were to advise founders on scaling that team and structure—

Yes.

What’s the—

Yeah. When I joined, we were 3 people, all very junior and very specialist. I hadn’t actually done a growth role before.

This is the advice that I normally give: start with 1 generalist growth marketer. They should be responsible for everything from messaging and positioning, working with the founder, up to awareness, setting up channels, and testing.

The reason you need them to do all of that is that if you only hire a product marketer, your positioning is lovely and your messaging is lovely, but no one’s heard of it. If you only hire someone who’s very quantitative and channel-focused, they’re only going to be focused on getting out there, but no one’s actually going to convert or resonate.

So hire 1 generalist person. The most important thing is that they really understand your product. If it’s a technical product, they need to be technical. If it’s a consumer product, they should really understand the consumers and the virality. It depends on the person, but start with a growth marketer.

Normally, the second hire I recommend is a growth engineer who’s front-end-leaning. This is a software engineer who’s hacky, who’s really into metrics. They should be setting up landing pages for SEO, doing mini tools, and doing automated outreach. Then you can keep growing from there. Other great hires are motion designers and back-end-focused growth engineers.

What’s a motion designer?

Motion designers. Yeah.

I have not heard of this. Are we reincarnating Titanic? Why do we need motion designers?

Video is hugely underrated as a medium for growth. Every launch video we do—the keystone marketing piece we ship with it—is the video.

There are different types of videos you can do. You can do motion design, which is basically animated graphics. You could do a founder-led video where they’re speaking to the camera. You could do more of a screen-share-style video where they’re recording and interacting, and it’s very product-focused.

But I love motion design because it’s abstract, so you can really focus on what the core value propositions are. You can make it catchy and attention-focused, and mix in the abstract UI elements.

The big mistake I see with most people’s videos is that they put the founder themselves in it. They do a full 5-minute monologue about the company mission.

And unless you have the editing skills of MrBeast, you're just not going to keep people for the launch video. So make it short. Get the key message across in the first 30 seconds, and motion design is fantastic for this.

Just digging in, because a lot of people will love that: when you say short, is it 30 seconds or 3 minutes?

The way I would do it is that nearly all your attention for your video should be on the first 30 seconds. Have a very quick intro, then get straight to the core value props. If you actually want to keep building out the video to make it 5 minutes long, that's absolutely fine, but just recognize that the majority of the drop-off will happen after that. So make sure you make those first 30 seconds truly incredible.

You've mentioned some different types there. We've got the motion design ones—

Yes.

—which are more animated. Is that correct?

Yes. Yeah, exactly.

Then we've got the founder-led kind of—

Yes.

—narrative. Any other types?

Motion design, founder-led, and screen share. Those are the 3.

What are the different use cases we'd use them for?

For any big launch, we lean toward motion design. You can get very complex products across quickly. The founder-led one, personally, I think those are really risky because it can become quite ego-ful. To slightly call out Superhuman, they are an incredible company—I know you're good friends with them. Truthfully, Superhuman are incredible at growth generally, but their most recent launch video is about 5 minutes and very founder-led.

Did you not like the staircase?

The staircase was beautiful.

That was another Titanic scene. I was waiting for Rose to come.

The staircase was beautiful, but you're watching Rahul walk down this beautiful staircase, and by the time he's walked to the bottom, you've moved on. In this attention economy, you really need to make it short and snappy. I think that can work if you have a more boring product, like consumer packaged goods, or if you're doing some sort of lifestyle product, or if you have a prominent creator behind it already. Fantastic. But in general, I'd be quite cautious of those.

The third type is the more screen-share format. I think this really works well if you're selling to a technical audience who basically want to nerd out on the details, the product itself, and the craft. Those are great. They're very quick to make. Screen Studio is a fantastic piece of software to use for that.

Fascinating. How do you think about using humor? Is there any world where quirky humor works, or is it actually, “No, no, no, this is a B2B sale; this is a professional tool”?

Yeah.

Don't try to be funny.

I think it depends on the brand of your company. At ElevenLabs, we try to make the core ElevenLabs brand pretty serious and very concrete. We see ourselves among Palantir, OpenAI, and SpaceX: we're mission-driven. We're going all in to build the world's AI audio models.

If instead you have a quirky brand, yes, lean into it. We try to get glimpses of humor. For April Fools, we did Text to Bark, an imaginary product where you type in text and can speak to your dog through barking, or bark to text and translate it. So, yeah, I think it can work, but if you're not funny, don't use it.

Can you do Text to Wife?

When you have a wife, Harry, yes.

This would be a hilarious product, actually, for the next April Fools. Have you ever done one that totally flopped, where you were like, “Ha,” and there was a lesson tied to it?

At the start, we didn't have good motion design contractors. We didn't have anyone in-house, so the main way it flopped was trying to work with contractors. We just didn't have good enough contractors set up.

I actually recommend bringing video in-house really quickly because when you're doing launches, you need to move super quickly. You have about a week from the product getting ready when you're running all this different marketing, so bring it in-house. The main mistakes are when we just don't have the right team in place to execute.

I so agree with you on bringing it in-house, but then founders say to me, “I get it, but you have the resources and I want to test.” Is there a way to test without spending $50,000 to $100,000 on hiring a motion designer?

Most motion designers work with you on a specific project for about $5,000 to $10,000. For most people who've raised money, that should be enough to be able to test it. If you think of every launch we do, most launches get somewhere between 200,000 and 700,000 views. If you had to spend that on Facebook ads or any other medium, the CPMs and the CACs are going to be way higher. So it's absolutely worth doing.

Fucking nuts. Do you worry that the value of video is going down with the explosion of supply? It's something that I do worry about. And we have teams of—

Yes.

—video people, but everyone is fucking doing video. I just worry that, with the increase in supply, it becomes harder and harder to discover.

Maybe. I don't think we're there yet, and I still think it's the most effective medium for getting your message across and keeping people's attention. People don't want to read blog posts now. Instead, have a crisp video. The platforms also love it and boost it.

The way I could potentially see it moving is with Twitter now and its live streams. I'm not sure if you've actually seen that whenever someone does a live stream, their entire audience has a bright red bar on top of the Twitter screen saying, “Blahdy blah is live streaming.” I could see the shift toward live streaming, but I'm unsure whether that makes sense for launches. Maybe a post-launch follow-up is another great way to grab attention.

I love this in terms of video being a core component. It's a core component of, as you said—

Yes.

—launches: 200—

Yeah.

—700,000. You do launches really, really well. When you think about what you do to make them successful, what are your biggest lessons on how to do launches well?

4. Seven Part Launch Playbook

Every time we ship a major product or feature, we run it through a checklist. We have 3 tiers of launches: tier 1, the big new model or massive new product line; tier 2, a major feature that customers will care about; and tier 3, which is basically just put on the changelog. Tier 1 is where we put the majority of our attention for launches, and this is normally led by the growth lead for that particular product.

The first thing they need to work out is: What are we actually launching, for whom, and why do they really care? We distill this down into audience, KPIs, and the core value props you want to get across, and then we turn that into consistent messaging. What are those short lines we want to get across?

To use an example, we launched Speech-to-Text, which is the world's best transcription model. You can say stuff like this, and then it'll transcribe it super accurately. The key messaging we wanted to get across is that it's the most accurate speech-to-text model. It's feature-complete: it does diarization, character-level timestamps, and supports 99 languages. That's actually what we spend a lot of time on: really nailing that messaging.

How many messages can we have?

So, yeah.

If we want the most accurate, fastest, and cheapest—

Yes.

—is that too much?

Yeah. We always have a primary one.

Okay.

The primary one, which you'll hear me, Mati, the entire team, and all our messaging repeat time and time again, is “the most accurate speech-to-text model.” Then there are secondary ones, so the 3 that I've said.

We then take this core messaging and turn it into the core assets. We always start with the tweet thread. What's the tweet thread? What's that core hook? That's what we make sure we nail first. Then we make the video as well, which corresponds with the messaging and the tweet threads.

Sometimes there are multiple videos, but particularly for tier 1, we want a fantastic motion-design video. We normally take that tweet thread. We also create a blog post, particularly for the more technical launches.

Then the next thing we think about is distribution. The way we see it is that we like to cross-post absolutely everywhere. We post on X, LinkedIn, Bluesky, and Threads—every single channel. Product Hunt, Reddit, Hacker News.

BeReal, I mean—

BeReal.

At that rate, you're fucking going into the crowds, aren't you?

You have to. The biggest advice I give to growth people is that you need to be loud. You need to get it across every single channel.

We actually have an amplified channel in our Slack where the entire team is in. When we do a launch, we post all the different links for the different tweet thread, LinkedIn, and everything else. Everyone else then shares, likes, and comments. They then do their own threads. The idea is that we want a surround sound for that day, getting that core messaging across.

And then I'm pretty certain the social algorithms see 30 or 40 likes in the first 5 minutes—

Yes.

—and go, “Wow, this must be good content. We'll promote it to our larger audience.”

100%.

What’s interesting is when I speak to new founders, they’re like, “Oh, Luke, but this is easy for you. You have a 200-person team. You have top investors. You have Harry as a mate who will share your stuff.”

What I’d put back to them is: okay, chances are if you’re getting to launch a product, you’ve raised a bit of money, you have some friends, and you know a few people in the industry. You should be shameless.

I ran this through with someone recently, and we were like, “Let’s take your Gmail. Let’s literally get every single person you’ve sent an email to in the last 5 years. Great, you’ve now got 3,000 people. Then who’s following you on Twitter? Who are you connected with on LinkedIn? Let’s just manually go through and DM all of them to boost your launch.” You need to give your launch a big boost for these algorithms to actually care about it.

I’m amazed by how nervous people are to ask, number 1. I’m amazed by how nervous people are to do things that won’t scale. For the first 50,000 Twitter followers I got, I personally DM’d everyone and suggested that they follow our newsletter as well.

Wow.

I was cross-advertising channels—

Yeah, 100%.

For 50,000.

Wow. I’m going to do that after.

No, seriously—at the end of every day, 15 minutes, you just continuously cross-promote other channels.

Hmm.

Honestly, it’s how we got to several hundred thousand on the newsletter. For founders who are maybe not as versed, what makes a good tweet, Luke? What should they not do? How would you advise me?

Yeah, it’s pretty simple. The first thing is, get the first word right. If you’re doing a launch, use the word “introducing” or “we’re excited to launch.” Make it really clear that you’re launching a new product: “Introducing the world’s best speech-to-text model.” That’s the first line.

Make it simple. One line, get across exactly what you’re launching with that keyword, which flags it. Then have a space, and do either a bullet-point list or a paragraph going into a bit more detail—maybe some of these secondary value props you want to get across. Then attach the launch video, which you spent lots of time working on.

To the first one.

To the first one, yeah. So that’s then your first tweet.

Yeah.

Then you could do a thread going into more information, maybe adding supplementary videos. The way Twitter works is when you do a tweet, what will go out there and what other people will see on their feeds will be the first tweet. The others will actually be minimized, and then the last 2.

You should put the call to action and the link in the very last tweet. You should make that second-to-last tweet great, too. Don’t put the link in the first tweet, because Elon Musk has even said they actively downrank people when they do that.

We use them for every show.

Yes.

And it’s a real science in that respect. On the blog-post side, people are going, “Hey, no one reads blogs.” Is there still a point to blog posts?

Yeah.

Would you advise that?

100%. So, 2 main points. The first one is for technical audiences. You need to go into the details. You need to show the benchmarks. You need to give a little bit of the secret sauce of how you put it together. Go into those technical details.

The second one is SEO. Lots of people think SEO is dead, but it’s really not. Every time you do a launch, you want to have a key blog post or landing page that you push and get all the backlinks to, whether that’s across the press, whether other people are linking to it, or whether it’s from socials.

SEO is not dead.

No.

Will it be in 2 to 3 years? When we see the shift in terms of users and where they come from, more and more—I think Guillermo from Vercel said it was 1.5% 6 months ago, and now it’s 4.5%. In 2 to 3 years, do you think we will see a much-reduced SEO presence?

With SEO, there are different types of SEO content. The first one is blog-style content. This is long-form articles. That, I think, will increasingly die over time. But actually, it’s not dead yet. If you look at Zapier, over 70% of their SEO traffic still goes to their blog. That’s definitely not dead yet, but I think it will die because you’ll be able to have ChatGPT give you a really great response. It will have scraped it and ingested it.

The part which I am confident will stick around for at least 5 years is tool pages. To give you an example, if you search “text-to-speech Spanish,” the first result will be ElevenLabs’ text-to-speech Spanish, which is a text box where you can type in any Spanish you want, choose a voice, and click play.

That requires some actual engineering work to make this text box. We do the same with speech-to-text. You create these mini tools, often with proprietary data sources or some actual engineering work, and it’s going to be a while before these LLMs are spinning up whole dynamic pages. By that point, maybe you’re even better off, because they’re kind of replacing your product on the fly anyway.

How do you advise founders on mini tools to show the value of a product when it’s actually an enterprise product? It’s really difficult. If you have an enterprise product, people honestly won’t buy without some sort of trial.

Yeah.

But it’s pretty hard to try and commit the resources to building that mini tool. How do you think about that?

This is why I think right at the start you should hire a growth engineer who’s front-end-focused and can independently focus on building these and exposing these small bits of value, so that people can experience the value as quickly as possible.

If you go to the ElevenLabs homepage today, we have 1 text box where you can type in text and play speech. But we also have all our products in these little text boxes. These are outside of the login, and people can try them.

The challenge I give to you is: don’t expose your whole product. Don’t give it all away for free. But what are these small bits of value that you could give away for free so people can experience that “wow” moment as quickly as possible?

You said something about distribution and cross-posting everywhere.

Yeah.

Bluesky—wow. Threads. I mean, I’m glad someone posts there. My question to you on that is: should we not choose 1 or 2 channels to absolutely crush—

Yes.

—and focus on, versus spray and pray?

Do both. First of all, choose the channels where your audience is. For us, that’s X and LinkedIn. Make sure those are really great. But then afterward, just repost it on the others, because there are people on the other platforms who, for some reason, prefer Threads or Bluesky. You want to reach them, too.

Also, because other people are overlooking those channels, you have a much greater chance that you’ll be able to own them and build an audience there.

Is it the same audience on X and LinkedIn, or different?

Largely different. We see X as mainly for our creators and developers. We see LinkedIn mainly for our potential future employees, potential partners, and enterprises. But we write the content in a way that tries to cater toward both.

Do you have TikTok, and do you think of TikTok as a growth channel?

Yes. We’ve actually just hired our first in-house creator, whose only job is to create TikToks, YouTube videos, Instagram Reels, and YouTube Shorts for ElevenLabs.

The way we actually hired this guy is that if you typed in “ElevenLabs” on YouTube, he had the most-viewed video above us. We were like, “Okay, he’s creating fantastic content. Let’s actually hire him and supercharge him.”

Particularly for this creator audience, having someone who can speak to them as a creator will really supercharge growth.

When you look back at the distribution and the channel strategy, what mistake do you think you made?

The mistake I think we made is not hiring early enough. We had super-strong PMF right from the start. We had channels working right from the start, but we didn’t actually have anyone focused on certain channels.

For example, affiliates: we set that up over 1.5 years ago, and that now is bringing in tens of thousands of dollars of MRR per month. But this was set up 1.5 years ago by 1 engineer in 1 week, and then no one’s touched it since.

What we should have done is, once we’d seen signs of life—given that we already had PMF overall—staff 1 person who could focus on growing that 1 product and that 1 set of KPIs and nothing else. Keep them laser-focused.

Overall, I wish we had staffed up more channels with more dedicated people sooner.

When we think about a channel working, what does that mean? I had this conversation with our team because we have, I think, 300,000 followers on TikTok—

Yeah.

—and millions of views a month. Honestly, the conversion back to the main podcast isn’t clearly attributable.

Yeah.

Maybe it’s there, but it’s not clearly attributable. Is that working?

I think social channels are tougher. For core performance-marketing channels, it’s very clear: is the CAC less than the predicted LTV of the person? For affiliates and different paid-marketing channels, that’s super clear.

With organic, and particularly enterprise marketing, trying to gauge that relationship is much harder overall.

Instead, the way I do it is you have to look over a longer time horizon and allow for more fuzzy attribution. For example, we're going to be doing a big enterprise marketing push in San Francisco. We're doing billboards, podcasts, newsletters, and events on the ground. We're not going to measure any one channel in isolation. Instead, we'll look over the campaign: what was the relative lift in the number of leads in that geo compared to a comparable city, like New York or Seattle?

So it's converted dollars. It's site visits in that geo.

Yeah.

How do you think about the North Star, which actually determines value?

What we've landed on for marketing—this is the enterprise marketing team with the enterprise part of our business—is our North Star metric: the number of marketing-sourced sales-qualified leads. Marketing has identified a lead. Someone signed up for a webinar or an email, or even signed up from the product. We've managed to get them to book a call with an SDR, and then that SDR has said, "Yes, this person's great." That's our North Star metric, and the entire marketing team is focused on how we drive that number up.

Webinars.

Yes.

That's still a thing?

It is a thing.

It makes me think of those phones for customer support, with the microphone coming down: "Hello, this is Greg. You've reached your webinar."

Yeah. Webinars definitely work. The key issue with webinars is the name "webinars."

Mm.

I would argue that what we're doing now is basically a webinar on growth. The only things that we're probably missing are, first, we're not livestreaming it, and second, we're not doing email capture. But then you can take this podcast, repurpose it, and push it on all the different platforms.

I think OpenAI did something quite smart recently when they called it OpenAI Academy, which is basically a webinar platform. They're putting on videos for businesses or educators, and they're doing email capture. They're effectively doing webinars, but because you're calling it an academy and it's focused on giving value, I think it goes down a lot better. So, yes, I think we'll be rebranding ElevenLabs webinars to something else soon.

I'm going to throw out some bold statements on topics we mentioned. You said CAC to LTV.

Yes.

Does CAC to LTV matter because it's so inaccurate and transient? The reason being, your CACs go up and down very quickly over time, and your LTVs, as you expand product lines, can significantly expand. I come in as an entry product, and suddenly I'm using 4 or 5. Does it really matter, and can it not mislead you?

Actually, the day-to-day metric and ratio we're really thinking about is CAC to payback period rather than CAC to LTV. We set different payback periods depending on which product line we have and how aggressive we want to be, but these vary between 12 months and 24, maybe even 36 months for that payback period.

And the 36 would be for the heavy enterprise.

Yes, exactly, where you know the customer is going to retain. They may even be signing multi-year deals. The way we set it up is that each different channel has a marketing lead, and they're thinking about this ratio. If it's positive, that basically means they should be putting their foot on the gas as quickly as possible.

You're exactly right, though, that over time you may be able to increase your payback period sooner. For example, you may layer in an enterprise product, and then you can sell your creators onto your enterprises. The way I would think about this for other companies, as well as for ElevenLabs, is that you set this goal. If you're above that ratio, if the ratio is looking great, you should absolutely be putting your foot on the pedal and trying to grow as quickly as possible.

Some people go, "We'll just raise our performance marketing spend by 20% a week." Well, no. Take advantage of the opportunity. Grow it as quickly as you can. On the other side, if it's actually below, you could give yourself more freedom because maybe you've got a bet on the thesis of how it will play out, where you're still like, "Okay, yes, we are losing money over the first 2 years for this product, but that's a bet we want to make."

But actually, if your marketing is going very well—which, for lots of these great AI companies, it is—then it's more just clear permission to grow as quickly as you can.

Do CACs go up or down over time when you think about it from the different purviews that you've had? Brand marketing becomes a thing. Word of mouth. You can see virality in communities. One would think it goes down. But then you also saturate your core audience and move to a less defined ICP. Maybe it goes up. What have been your observations on CAC going up or down over time?

Yeah.

For specific channels, they do tend to go up. But when you look at it broadly, and as you're adding new channels, increasing virality, improving your activation rate, and improving your conversion to paid, you can get that overall blended CAC to hopefully even just stay flat.

But exactly as you said before, because you're able to layer in, say, enterprise products that you can upsell to—or we've actually done this funky move where we've layered in a consumer product, which enables us to get the CAC incredibly cheap—people may move up to Creator or even Enterprise.

Is the future of enterprise consumer entry? What I mean by that is, to your point, at the end of the day, we all forget that everyone is a consumer. Every big enterprise user is a consumer at some point. Is the future of enterprise consumer entry?

I actually really hate the word "enterprise" in general.

Bold.

For a while within ElevenLabs, there was this phrase of, "We need to sell to enterprises." But who are the enterprises? I was like, "Let's work together. Let's refine who exactly they are." We basically worked out that they're normally engineering managers or product leads at these larger companies. Sometimes the CIOs or CTOs are involved, too.

But the first step is, okay, let's actually define who we're selling to. Then you realize, okay, wait, they have so much overlap with our developer audience and our developer marketing. Maybe it's not about doing long white papers and executive dinners. Maybe it's actually more about this bottoms-up approach.

But zooming out, the bit we've actually landed on at ElevenLabs is that we want to grow super quickly. We want to try and get and work with all the best companies in the world. Why not do both? What we've done is spin up separate teams.

We have an enterprise marketing team, which is just focused on going in from the top. They're asking, "How do I get these SQLs which are marketing-sourced?" They're doing executive dinners, events, and the webinars, which you love.

So they're doing ABM—personalized outreach, very tailored.

Yes.

That sort of thing.

Very tailored. Then, on the other end, we have a developer team made up of developer advocates who are focused on much broader-scale awareness. We're doing hackathons and different events from the bottom up. They have completely different goals, but work in tandem with these roughly overlapping audiences.

You mentioned a push in billboards and podcasts and all these. How do you think about when's the right time to do brand marketing?

5. Build The Brand Early

Ideally, you're doing it right from the start. If you're trying to build a massive company, you want people to join you for your mission. How that may look over time may switch, but right at the start I'd really be thinking about engaging with your community.

Like you said earlier, you messaging every single person on Twitter who follows you—that's you building your brand. That's you as a person being relatable. Ideally, you have that right at the start, and it comes from the founder putting themselves out there, being authentic, and building it.

Then, yes, over time you can scale that into different brand-awareness campaigns. I think you just need to choose the ratio of spend that you're happy with. Maybe that's 20%; maybe that's 70%. I think it depends on how well your other channels are working.

You said there about the founder putting themselves out there.

Yes.

How important is founder brand today?

I think it's incredibly important. It's also, though, super tough and very distracting.

It's a commitment.

It's a commitment.

And that's what people don't recognize. It's not like, "I'll do a post."

No.

It's like you post every day.

You post every day. You're getting the dopamine of likes on your LinkedIn posts and likes on your X threads. I think it's quite risky because you can spend all this time optimizing for the wrong thing.

The way we think about it at ElevenLabs is: what channels really naturally resonate with the founders, and what do they get really excited by? For us, Mati is fantastic in person, at large speaking events, and at fireside chats, so that's what he really leans into.

Equally, he's not that excited about being active on Twitter or X. He naturally isn't that active on those platforms, and that's fine.

Yeah.

The way that we've done it is, basically, we have other people within the company who are instead active and pushing. I'm quite active, and the developer team is quite active. Instead, we're pushing the company mission through channels we get really excited by.

I also always think that just because you focus on one channel—speaking, say, at events—doesn't mean you can't go multichannel with that. I would be like, "Hey, how do we get video editors to go to every single thing that he does?"

Yes.

Get him speaking for, I don't know, 10 a month, say.

Yes.

And just fucking pummel TikTok, Instagram, and YouTube Shorts with Mati speaking at different events.

You could do that, but then you constantly see yourself going viral on these channels. Maybe you're getting negative comments and largely positive comments, but you're still putting yourself out there. If you're not naturally that excited by it, it's a big distraction. The way I view startups is that they're long-term games. They're 10-year, 20-year-long games, and so you need to stay in it and recognize what you personally get really excited by.

Do you believe that there's no such thing as bad press? Essentially, relevance is everything.

I think there is such a thing as bad press, particularly if you're trying to sell to enterprises.

Deel would be bad.

Yeah, Deel. They're in a tough spot. Another good example is this company that raised $4 million for—

Oh my God, the—yeah, I saw it literally—

Cluely—

Roy Lee raised $5.3 million from Susa Ventures.

Yeah. What's it called?

It's the "Cheat on Anything" company.

Yes.

Yeah.

Yeah. They've gone for "Cheat on Anything." Their founder-led marketing is him getting offers from Amazon, Palantir, and Google, and then telling them, "Oh, wait, I cheated on it using my tool." That gets fantastic marketing, but he's also putting a big black X on himself if he ever actually wants to sell to enterprise sales teams, where you're like, "Do I really trust this person?" I think it's a risky move.

Do you think so, or do you think it's just a good entry point to build that brand? Because, actually, do you know what? In 6 to 12 months' time, when he's got experienced sales reps in there who are actually selling real product, going, "Hey, we work with Rippling and we work with ElevenLabs," you can get away from it.

Maybe. But I think the core of your brand needs to be authentic, and it needs to come from a place you really believe in. Everyone is going to remember that company as the one that helps you cheat on different tests, which is fine. If he leans into, "Okay, I want to sell to the more consumer type. I help people in their day-to-day life," maybe that's absolutely fine. But I think if you're trying to do large-scale enterprise sales, where you want to be seen as the trusted partner—which you often do, and that's the differentiator—then I think it's a more risky play.

People often forget how difficult it is to shift a tag.

Yes.

Calm.

Yes.

Everyone goes, "The meditation company."

Yes.

And it's like, that is not what they are. They really try to move away from that, and it's just very difficult when you become very well-known for something to lose that original tagging.

100%.

So I think you need to be really intentional right from the start. What's the messaging? Who do you want to resonate with? Who do you want to stand for? For us at ElevenLabs, we want to be the super-trusted player in the space, both for voice actors and large enterprises.

How do you think about competition, and how do you advise founders on competition?

I think it's a great way to counter-position against other brands. Pretty much every space you go into will have existing large players. There are a couple of different options that work. The YC approach, which often works really well, is basically just to ignore them, put on your blinkers, speak to your customers, and make an incredible product. I think that's really smart for product development. But I think what Ramp has done with its counter-positioning against Brex is just genius marketing, which—

Sorry, you've said "counter-positioning" twice. I love 7 Powers, and so, for me, it's incredibly exciting. But what is counter-positioning first?

Counter-positioning is where you take someone else's core messaging or core positioning in the market, take the exact opposite, and use that as your strength rather than your weakness. The example with Brex is that Brex really stood for, "Hey, spend as much money on your card as possible. We give you all these fantastic points." It was kind of like, "Look how many extra points and rewards and stuff we're giving you."

Ramp took the exact opposite approach, which was, "Hey, we're not going to give you any points. Points are a waste of money. You end up spending them on stupid things. Instead, we're going to pump that all back into software." As soon as they made points an actual bad thing and got the whole market to think about points as being bad—let's use that money to invest in great software, with the focus instead on saving money and almost reducing the number of these imaginary points you're getting—they were really well counter-positioned. Basically, every time Brex would do something, they would be able to riff off it in the other direction.

I'm just thinking about how we could use counter-positioning effectively.

Yes.

I'm thinking about a world of really long podcasts, where three and a half hours of Joe Rogan is pontificating about planetary—

Yeah.

—evolution. Give you a 3-minute episode. Do you know what I mean? Just the best bits: doof, doof, doof, doof.

Yeah, well, I think you did that, right? You were 20VC; you were short and concise. You were counter-positioned at the start. The interesting thing is that you're now kind of the incumbent, and people have moved on. Or not moved on, but—

Dude, no, totally.

The amount of people who message me and say, "We're going to start the 19-Minute VC"—

Yes. Yeah, yeah. I saw one great example recently: TBPN.

TBPN, yeah.

Terrible name, but great content.

Yeah.

Well, okay, so this is interesting.

I'm going to get in trouble for this, but fuck it. I think it's Clubhouse for our era. They have brilliant channel-market fit, which is that they crush it on Twitter with very specific clips. When you look at any of their view rates, very little. When you look at their retention and engagement rates, fuck all.

But they have brilliant channel-market fit. This is not a criticism of them. They have brilliant channel-market fit for Twitter and certain types of videos. But they also do volume. It's a spray game. Clubhouse.

We will see. I think they're still evolving what their actual product is, but they found this, as you said, channel fit. To jump back to the counter-positioning, they've done the really funky thing where they're like, "Okay, they've counter-positioned against All-In." All-In was like, "No ads"; they're like, "We're going to be all ads."

Traditional media is going more in the direction of being anti-tech—or maybe not TechCrunch, but traditional media. They're like, "No, we're going to be completely pro-tech." Others are going, "We're only polished, buttoned-up, clean cuts." They're like, "No, we're going to be streaming live content, the raw stuff."

And they're smartly dressed.

Smartly—yeah, yeah, yeah.

No tech-bro hoodies and stuff. They're wearing shirts, and they look really formal. Listen, I think they've been really smart. I didn't mean that as a takedown, but I think it's interesting. Dude, we were breaking down the teams earlier.

Yes.

You have a weird element, which is—and you said this to me before—you don't have PMs.

Yes.

Well, why? Explain this to me. This is like the—

Yeah.

—holy grail of product teams, no?

And for context, Harry asked me this in front of an audience of about 19 people.

Of only PMs. Do you remember? You were like, "Raise your hand." It was like, "Oh God."

Of only PMs, yeah.

6. Why ElevenLabs Has No PMs

The reason we don't have PMs is that our overall thesis is that the engineers are building the product, and they should be responsible for the product.

And with AI now, we're seeing much more, like, different traditional roles merging. And so engineers are going towards product engineers or engineers towards growth engineers. And what's fantastic is when you have product engineers and the engineers who are actually building the product, they own the roadmap. They can go from idea. And chances are their ideas are better and more right because they're the ones speaking to users and understand the context. So they can have better ideas. They then ship the product. They then get the results and analyze them. And they do that entire loop without needing anyone else's permission or anyone to slow them down. So the ideal setup is engineers own the product, and they're responsible for it. Now, instead of having PMs or a separate marketing function, which is kind of like three separate functions—engineering, product, marketing—instead, we've merged a lot of product with a lot of marketing and called that growth. And most of the team actually is ex-PMs who are now doing growth. And so they're responsible for the marketing as well. And so now each different product has an engineering lead which is responsible for how great the product overall is, and they partner with the growth lead, which is responsible particularly for the awareness and acquisition, and then they collaborate on the activation and retention pieces.

How do you prevent the growth role, which is a lot of former PMs, just slipping into a PM role again and trying to actively manage product and kind of over-engineer with the engineers?

We keep the growth teams pretty small. And so frankly, there's just not enough time to be meddling in the different engineers' work. You're not going to be doing wireframes or large numbers of user interviews because you're too busy running the launches or running the channels for that particular product. So that's one: we keep it pretty busy and lean. And the other one is, if you hire great engineers, they just don't need it. And so the best engineers we hire, we put everyone through a product challenge, which is where we give them an imaginary product and we ask them, step one, work out what features you should build. Imagine I'm a customer, imagine you're doing research. They will actually go through on Google and find competitors and use that to shape the features. So step one is work out what features to build. Step two, they're then going to turn that into the actual wireframes, and they'll be sketching out in Figma, like, okay, this is what I think the product should look like. And then step three is system architecture. So they're like, okay, this is the different back end, the API design. And if you hire an engineer who, one, you know is really strong at coding because you've screened for that before, and two, you've then tested them through this full product loop, you have confidence that they're able to take those products and those features from idea to launch.

Do you think we will see PMs reduce as a role in future tech companies with AI becoming more and more prominent?

Yeah. I think PMs move to either growth, so you combine the PM role with marketing. And depending on your size and how strong your engineers are, you do more or less of the PM. Or many PMs will actually move towards product engineering, which is they're somewhat technical already. They can actually upskill themselves using tools like Cursor or Lovable and actually go and ship full end-to-end products.

Can I ask you today, when you think about Eleven Labs code, how much of it do you think is AI-generated?

So I still write code probably 20% of the time. The only code I write is through Cursor, where I'm describing in the chat what I want done, and then it executes on my behalf. And I'm probably more the extreme because I'm often doing growth features which we're adding on top of the core of the product. In terms of core engineering, I would say probably 60, 70% of code now is AI-written, so still the majority. Where we don't do any AI-written or AI-generated code is our research engineering, where it's just research engineering and research, where it's a very different style of work. And also these code bases are so sensitive. When you think about the initial Eleven Labs launch and how quickly we got to a billion dollars of revenue, that's pretty much Peter's models being world-class, and so we just don't want to go through any sort of LLM.

Was that a million dollars of revenue?

I said billion dollars of valuation.

Ah.

And I think truthfully at that point it was probably half the enterprise value of the company was Peter's model.

Nuts.

Yeah.

Can I ask you—we mentioned the enterprise discussions that you had earlier. The big concern with a lot of these gen AI companies today is their sugar-high revenues. They're unsustainable. We've seen them hit $50 million ARR, wow, wow, wow. But is it real revenue? How do you think about whether we have a generation of companies where it's pretty synthetic revenue or not?

You know, it probably varies by product to product.

Yeah.

But if they're solving a real problem and the retention is strong, I see no reason why these aren't real revenues. And the thing, I've spoken to a few VC friends, I don't understand why they're not upping the rate that they're investing. They're seeing companies which are now, they're like, oh, it's already on $6 million MRR in one year, and I'm like, why aren't you investing in it? And they're like, well, there's this other one which is now on $10 million of MRR. And you're like, surely now is the time. The wrong time to be the Tiger Global play of investing quickly was back in the COVID boom. The exact right time is now when I think you have this unlock of capabilities with the AI models, which has meant new products which just couldn't exist before.

I think the hardest thing is before there were two to three competitors always. Now there are literally 10 to 15 on everything. And so discerning winners is much harder when all of them have relatively impressive traction and revenue.

Yes. But as long as they're solving real problems and scaling quickly and the customers are retaining, particularly if you're doing B2B or enterprise, you're going to get that cohort of customers for a while. And so if you're already on $10 million of revenue, if we think that it used to be the threshold for IPO was, oh, let's get to $100 million of ARR, you actually have entire waves of companies which are now reaching those thresholds.

What is good retention? This is a hard point, which is, in traditional enterprise it was very, very high. And you'd look at a Lovable with, I think, 86%, I think Anton said on the show. In traditional SaaS, that's still shit. You're churning 14% a month.

I think people largely think about retention for consumer and prosumer wrong. So for enterprise it's very clear, which is you sign one contract, say you sign a large enterprise, hypothetically Slack, and you land them one seat and then you expand, and then you can look through and be like, oh, our NRR after one year is we're now on 150% NRR. Great, that contract's grown. If you think about the way it's actually grown underneath, it's seat expansion within Slack. I'm sure lots of those individuals have churned. On an individual user they're probably not growing that individual user's revenue more than 100%, instead you're growing within that account base. And I think when you look at consumer and prosumer, yes, for that individual user you may only be on 87% revenue retention, but what then happens is you build a product on Lovable, you then share it with a bunch of your friends, a handful more people sign up to Lovable. And so when you take into account the natural virality rates, I think your overall NRR of that one account you've grown may actually be a lot higher. You should still look at the individual seat retention, but I think it's fine to have it below 100% on the individual level. And then when you zoom out and take into account the organic growth which comes because of it, you do want to be over 100%.

You know what was nuts is actually the fact that the revenues are growing so fast that by the time the rounds are done, I mean, when we did Lovable, I think it trebled its revenue by the time the round had completed. And I was like, I would want to retrade on price. Even if she was only a 10x revenue multiple when we did it.

Yeah, yeah. We both invested in Lovable and at the end of Q1 on Eleven Labs we did very well. We had our best quarter ever. We had more enterprise using it, more consumers, more prosumers. I was tapping myself on the back. You know, growth team, I was like, we did pretty well. And then we got the email from Lovable with their latest results. I was like, okay, we need to up our game a bit.

Back to work.

Yeah.

No one leaves the office. No one leaves the—

Exactly. Exactly.

Don't let Mati see this email. Don't let Mati see this email. What do you actually think happens there? Does Lovable become a $10 billion company and change the way that we create websites? Does it do what Squarespace and Wix do—adopt similar processes? What happens?

I think there are a few ways they could go. Overall, though, I'm just very bullish. I don't see a reason that the level of intelligence in these AI models will slow down. Given that it will keep rising, I think you're going to see more and more engineering.

We're already at, say, 60% of all our code being AI-generated. Then you have the broader life cycle of all the testing and deployment infrastructure. There's no reason why any of that couldn't be done entirely by AI as well. I think Lovable is in a fantastic place to own that core stack.

The tricky bit, I think, at the start is: How do you make it super secure? How do you get enterprises to build it? How do you do full apps? I think Retool is kind of asleep at the wheel at the moment. I don't hear about the recent things they've been launching, and so I could imagine Lovable going in that direction. You're cringing.

No, I'm with you totally. I remember when they were the hottest company. Actually, I haven't heard anything.

Yes. Yeah.

Dude, I want to dive into a quick-fire. I say a short statement and you give me your immediate thoughts. Does that sound okay?

Great.

Number 1: What's the most common or expensive mistake that you see early founders making?

7. Avoid Paid Marketing Too Early

The biggest mistake I see early founders making, particularly for B2B, is doing paid marketing too early. If you don't have PMF yet, you're not going to get paid marketing working, and you're going to spend all your time and all your engineering time optimizing funnels and conversion, creating these different creatives and ads, and caring about your rankings and your Meta metrics. None of that actually matters apart from just building an incredible product.

I think the core loop of doing incredible launches and building an incredible product is enough for any B2B or prosumer product to get that initial product-market fit, and then you can start testing out different paid channels. The caveat, though, I'd say is if you are a consumer. With B2B, you basically know that if you build an exceptional product, you can staff a sales team for it and they'd be able to scale it.

With consumer, it's very different, where distribution is a key part. You could create an incredible product, but if no one hears about it, then you can't scale it. Instead, you may want to consider performance marketing earlier. But again, I would try and lead with other channels, which also increase brand.

I totally agree with you. The one thing I find paid really helpful for is if you're strategic around how you use it for testing.

Mm-hmm.

I've never seen that work. Have you?

I mean it—

I've seen people waste a lot of time when they—

It runs our media business on, like—

Okay.

...different title optimization...

True.

...and different color optimization, different font optimization.

Yes. Yes. But when you were making your first episodes, should you have spent more time A/B testing different podcasts? Maybe, Harry, instead of 20VC, you could have leaned more into product or growth earlier?

Yeah.

No, it's like—

No. Just—

Yes, exactly. Build a better product first—

Yeah.

...and then lean into paid.

Totally. That was a shit suggestion, Harry. What's the most underappreciated growth channel today? What do people not do that you're like—

The most underappreciated growth channel I see is organic LinkedIn. On Twitter, you're competing against the best writers in the world. You're competing against Harry Stebbings running Shorts.

Bill Ackman. I mean, fuck.

Yeah. Bill Ackman. It's really tough. On LinkedIn, you're competing against Freddy from Deloitte, who got a promotion. The bar to creating great content, standing out, and being authentic is so much lower.

That is so true. LinkedIn is unbelievable as a content engine today. I think people—

100%.

...really. What's the most polluted channel that you're like, “Ugh, stay away”?

Actually, LinkedIn ads. LinkedIn organic is fantastic. The bar is so low, you can get stuff super viral. LinkedIn ads have such high CPMs. Yes, you can do great targeting, but I would argue that you should invest a lot more time into creating excellent content that you distribute organically.

Maybe you want to layer on ads on top, but particularly for the kind of people that are then running LinkedIn ads—you're doing enterprise, you're doing heavy products with a long sales cycle—it's really tough to get that working. Instead, I would just focus on the organic side.

What's your advice to someone starting a new role in growth tomorrow?

The biggest advice I would give is: get great at copywriting. The actual writing of copy—writing words, making something engaging, getting your message across—is so fundamental to any bit of growth you do.

Whether you're running LinkedIn ads, doing organic LinkedIn, writing a blog post, a tweet thread, or copy for your landing pages, get great at copy. That's the foundation for all other growth.

It's a fucking fantastic podcast. You know why you're actually a brilliant podcast guest as well? This would be a big piece of advice that I should actually tweet, and I probably will do later. You repeat the question at the start of your statement, which is so helpful for Shorts. Do you see what I mean? So you say the—

Yeah, I have distribution in mind.

Totally. You say, “The biggest problem people make with X.”

Yes.

'Cause the worst thing is when people just go, “Oh, they do this.”

Yes.

That's really hard—

Mm.

...to short-formify. Do you see what I mean?

Yeah, yeah, yeah.

Many reasons why I'm single. It's 'cause I get excited by that. What would you most like to change about the world of growth?

You've made me aware—

Fucking come on.

...I'm now doing it. No, I'm joking. The thing I'd most like to change about the world of growth is that you just should not launch products which aren't actually ready. The biggest villain I see doing this at the moment is Apple.

We've always held Steve Jobs in our mind as this beacon of marketing, product, and engineering. Now you have Apple pushing the idea that the reason you should upgrade to the newest iPhone or laptop is because of Apple Intelligence, when no one actually knows what Apple Intelligence is or how to use it, or gets any value from it.

And it's not coming out until 2026.

It's still not out.

No.

So why are they pushing it in their marketing? Instead, make sure your product is actually ready, that it performs great, and then ship it and push it hard.

They did kind of change their whole AI team, I think, as a result of that.

Right.

What have you changed your mind on in the last 12 months?

The biggest thing I've changed my mind on in the last 12 months is whether voice AI agents or conversational AI agents actually work. When I joined ElevenLabs, we had this fantastic box where you could type in text and generate lifelike speech.

Mati was saying, “Soon you'll be able to have full AI agents that you can have natural conversations with, and it'll sound really human-like. You could use that for sales, customer support, and relationship coaching.” I was like, “That will never work. The latency will just be too high. You won't be able to have these sub-200-millisecond interactions.” I was completely wrong.

After helping about 20 customers build these conversational AI agents, we were like, “Okay, let's actually turn this into a proper platform and build conversational AI as a feature.” Now, I actually prefer speaking to an AI customer support agent over a real person because they're more knowledgeable. They know exactly how to solve your problems, they can escalate, and they know all the policies.

You also don't have to be fake polite.

Yeah.

You know what I mean? You're like, “I know that you don't have feelings.”

Yeah.

So it's like, “What's this? What's that?” Yeah, yeah, I totally agree.

100%.

When does the infrastructure layer become the application layer? What I mean by that specifically is, you could be Synthesia.

Mm.

You could be HeyGen.

Yeah.

Why does the infrastructure layer not become the application layer?

I think it does, but you as the infrastructure player really need to partner with companies and be very explicit around which areas you want to go deep on. We want to build the best platform to make these end-to-end conversational AI agents, but we explicitly don't want to be building full digital avatar platforms.

That way, we're able to be trusted partners to companies like HeyGen, Synthesia, and Captions, who are all customers, but we're able to build these fantastic infrastructure layers on top.

You know, I invested in Captions.

Really?

Yeah. I invested in Captions. This shows how theory-driven I am as an investor. I invested when it was basically a BeReal competitor.

A year later, a friend of mine texted me, saying, “We've got to invest in this company, Captions.” So I found the founder on LinkedIn and messaged him: “Oh my God, would love to find a way to invest. We love Captions.”

And he's like, “Dude, you already invested.” I'm like, “Say what?” And he's like, “We're a pivot. You remember me? You remember me? I'm Gaurav.”

Wow.

And I was like, “Oh, shit.” Anyway—

Damn, but I'm sure you really helped him through that pivot, and it was very useful having your advice.

Listen, I don't want to sound like a co-founder, but—

Yeah.

It was up there. Why is inbound SDR as a role dead?

If you think about the actual function of most inbound SDRs, with an enterprise sales cycle, someone fills in a form. They then need to speak to an inbound SDR who basically goes through BANT screening. So they work out: do you have the budget? Do you have the authority? Do you have the need? Do you have the timing?

If yes, they then book you into another call with the AE, which actually closes the deal. And so the function of an inbound SDR—I'm sure the best ones are persuasive and friendly—but really, the function is data collection.

Instead, why don't you just add that to the form in the first place? The reason is people don't really want to go through this back-and-forth and answer more detailed questions, or try to design these complex flowcharts or forms.

So instead, you can switch to a conversational AI agent, which can completely replace that role and get the people who should be booked in with AEs much quicker. You could even clone the voice of the AE, so it's like them almost pre-screening for their own deal flow.

I love that as an idea. That'd be very funny to do. You could do a lot of very funny things with your promotional videos, by the way.

Yes. What should we do?

Well, I think you might also get into the realm of copyright problems.

Ah, yes.

I don't know if you're allowed to use Trump's voice and Obama's voice.

No. And actually, we do a whole bunch of different safety work to make sure it's explicitly used where people give permission for their voice on the platform.

Don't do that, kids.

Yes.

Yeah, I remember when I met Victor from Synthesia for the first time—

Yes.

He was like, “Imagine we could have Trump declaring this and doing this.” And I remember being like, “No. That's a bad idea.” And I didn't invest in the seed.

Oh, wow.

Oh, I know.

No. Well, he's done great, and I know they take safety very seriously as well.

No, they do. Absolutely. Sorry, that sounds terrible. He's so good. I like him and Steffen a lot.

Final one: what's the best growth strategy that you've just been really impressed by in the last 12 to 18 months?

Yeah. The best growth strategy I've seen is what Bryan Johnson is doing. His marketing is so good that they made a Netflix show about his marketing. And I do think Bryan Johnson's actually being authentic. He really does care about longevity and about creating great products.

But he's put himself—one, it's just super controversial, what he's doing. If you look at the way he dresses, it's definitely not normal clothes. And the things he's doing—he's doing bloodletting and blood transfusions with his son.

First, he's generating a lot of controversy, such that we're speaking about him now on this podcast. Second, they've really dialed in the messaging. “Don't Die” is super catchy. They've built a brand around it. They've built a community. Their brand building is just fantastic.

And then their distribution: he's constantly on Twitter. Anytime someone talks about doing a late night, eating junk food, or pulling an all-nighter to ship a feature, he's there telling them off. He's there consistently.

In the comments, like, “Wouldn't have done that if I were you.”

Yes. 100%.

Yeah, I saw that. Or he's quite sarcastically funny. Jason Calacanis tweeted a picture of donuts.

Yes.

And he was like, “Sign me up. I'll take the box.”

Yeah.

And it's just really funny to see him engaging in that way.

Yeah, 100%. And you feel it is really authentic. He genuinely does care. I think he's genuinely doing it in a good way.

I have that longevity shake in the morning.

Is it good?

I tell you what, it's fucking brilliant. And I'm on a subscription. It's £60 a month, so I'm £720 a year. No way I'm fucking sharing.

I take it when I travel. I look like a drug dealer. I've got cellophane bags of his stuff. But it's amazing. I'm totally with you.

Dude, this has been so much fun. Thank you so much for doing this, and it's been so good to have you in the studio.

Cool. Thank you so much, Harry.