Head of Growth (Anthropic): Anthropic is automating its own growth

[00:00] A lot of companies claim to be the fastest growing companies of all time. Anthropic actually is. You guys were at a billion ARR at the start of 2025. The last number I've seen is 19 billion ARR. That's one to $19 billion in 14 months. Historically, we were very much the smallest, least well-funded player in this space. We didn't have the free cash flow or the distribution of a meta or Google. We didn't have the first mover advantage of an open AI. It's a complete miracle that we've gotten to the stage that we have.

[00:30] of what it's like to be leading growth inside of Anthropics. The hardest job I've had in my life to come into Anthropics, you need to understand that 50, 60, 70% of how you operate in the past just throw it out the door. One of the cleverest growth moves you all made was this idea of importing memory from ChatGPT. Activation is a really big challenge in AI. We are starting to look at how do we automate growth. Our growth platform team is driving this effort called [01:00] hypergrowth, how can we use Claw to automate growth experimentation?

And it's delivering results. You're basically living in the future. We always talk about the exponential. The product value that we will deliver in two years time is probably like 1000x what it is today. The funniest thing is I've noticed internally linear charts are just not cool. Everything is log linear. It's just show me a log linear scale. [01:22] Today my guest is Amol Avasari. [01:24] Amol is head of growth at Anthropic, which is on the most unprecedented growth run in history. [01:31] In the past 14 months, they grew from $1 billion to over $19 billion in annual recurring revenue.

[01:37] Just in the past few months, their revenue doubled. They've been growing 10x year over year. This is unheard of at this scale. By the time this episode comes out, their revenue will be even higher. To put this scale in perspective, companies like Atlassian and Palantir and Snowflake, which have been around for 15 to 20 years, each do something like $4.5 to $6 billion in ARR. [02:00] Anthropic is adding this much ARR every few months. [02:04] And if that isn't interesting enough to you, Amol, who leads growth at Anthropic, is an incredible human.

He previously led growth at Mercury and Masterclass. Before that, he was a founder and an investment banker. And most interestingly, something that most people don't know about him, [02:19] is that Amol suffered a severe brain injury. He had to spend 9 months relearning how to walk, and work, and just not be nauseous all the time. [02:27] He shared this story in a guest post in my newsletter a number of years ago, [02:30] We actually chat about this during the conversation. These are my favorite kind of conversations, because Amol and his team are living in the future.

And he's come to tell us where things are heading and what's going to change. And in this episode, Amol shares an unprecedented look at how a company like Anthropik operates and grows, including how they think about growth, what parts of the job they've automated, the future of the product and growth roles, how Amol got the job in the first place by cold emailing Mike Krieger, [02:57] and so much more. Amol is wonderful and just try to count the number of times that he blew my mind during this conversation. [03:04] Before we get into it, don't forget to check out Lenny's Product

com for an incredible set of deals available exclusively to Lenny's newsletter subscribers. [03:12] With that, I bring you Amol Avasari. [03:15] Amal, thank you so much for being here and welcome to the podcast. [03:22] Pleasure to be here. [03:23] Head of Growth at Anthropic, [03:25] No big deal. I've had a lot of people come on this podcast from companies that claim to be the fastest growing companies of all time. [03:33] Anthropic actually is, if you look at the trajectory, I just have some of the numbers here just so people understand how absurd.

[03:39] This is so you guys were at a billion ARR at the start of 2025, then hit something like four billion mid 2025. [03:48] than 9 billion ARR at the end of 2025. [03:52] And the last number I've seen is you guys are at 19 billion ARR. [03:58] which, just to put a couple... [04:00] pieces of context here. One is that's from one to 19 billion dollars in 14 months. [04:05] Um, [04:06] I have so many questions. [04:09] First of all, the story of how you actually landed this role is really interesting.

Talk about how you got this role. [04:14] Yeah, it's a little unorthodox, so... [04:17] It's funny, when I did my onboarding, they walked through what percentage of the cohort came through referrals, what percentage came through applying on the website, what percentage... [04:26] came through sourcing... [04:29] And I was on none of those. And I was like, OK, this is interesting. Basically, the way that I got to Anthropic was that [04:36] I was actually a user of Claude, and I... [04:40] I was using a lot of like, man, these guys like great product, great company, but they really like obviously don't have a growth team.

And what I did was I just sent Mike Krieger a cold email. He was a chief product officer. [04:52] I sent him a cold email saying like, hey, love what you guys do. Love the product. I think you guys badly need a growth team. Want to chat? [04:59] and uh i didn't expect he would respond and uh you know he responds and says hey yeah i'm interested let's let's talk and it's funny i didn't know i mean i wouldn't have known they were not hiring for a growth team there were no growth pm roles listed but they were just at that time starting to think about hiring a growth team so it was very good timing but [05:19] Yeah, one spoke to Mike and one thing led to another.

He said, "I'm the only PM that is hired from cold email and I feel very lucky that he decided to respond to my email." [05:30] I did not know the story. That is another absurd fact. [05:36] Clearly you're good at cold email. What did you do in this cold email to get his attention? I would say like I've basically perfected cold email over the years. So [05:44] When I was a founder, I had to get really, really good at this. So I sent a lot of cold emails out. [05:52] and just honed the...

[05:55] subject line, the message and the tone. And so, [05:59] Basically, I have in the subject line, the first thing is like from a conversion standpoint, someone sees the email, they need to click on it. And so I have a copy that I've tested that is like very, very high open rate. And so wait, what is this copy? Or is this a secret? It's a secret. OK, we'll keep it. We'll keep some secret. It's a secret. [06:16] So that's one, getting them to open. I think the second is then the tactics of [06:21] you need to understand like where are people getting outreach and if you if everyone's getting outreach in one area and then [06:26] you reach out to them there, then you're not going to get as high of a response rate.

So you can think about LinkedIn, you can think about work email, these are things that [06:34] everyone is emailing. So there's, [06:35] there's ways to get people's personal emails out there. And so like, that's one thing that I did. And so, okay, I've got his personal email. [06:41] I know the copy that works. And then it's just keeping it very short on, [06:45] Here's who I am. [06:46] here's why I'd be a good fit and we should chat. And these things typically don't work and then [06:51] You should always follow up a few times.

I think my rule of thumb is like, if I really care about it, I should just keep [06:57] keep reaching out to them until they tell me, like, please stop. And so I would have kept doing that, but he responded the first time. [07:03] It makes sense that a talented growth person would be very good at cold email and getting people's attention. [07:08] So that's almost like an interview step as just, did I want to read this email? [07:12] This episode is brought to you by our season's presenting sponsor, WorkOS.

What do OpenAI, Anthropic, Cursor, Vercel, Replit, Sierra, Clay, and hundreds of other winning companies all have in common? They are all powered by WorkOS. If you're building a product for the enterprise, you've felt the pain of integrating single sign-on, SCIM, RBAC, audit logs, and other features required by large companies. WorkOS turns those deal blockers into drop-in APIs with a modern developer platform built [07:42] specifically for B2B SaaS. Literally every startup that I'm an investor in that starts to expand up market ends up working with WorkOS. And that's because they are the best.

Whether you are a seed stage startup trying to land your first enterprise customer or a unicorn expanding globally, WorkOS is the fastest path to becoming enterprise ready and unblocking growth. It's essentially Stripe for enterprise features. Visit com to get started or just hit up their [08:12] WorkOS allows you to build faster with delightful APIs, comprehensive docs, and a smooth developer experience. Go to com to make your app enterprise-ready today. [08:22] Okay, so give us just like a glimpse of what it's like to be leading growth inside of Anthropic right now.

[08:30] the most by far fastest growing company in history. Just what is it like? [08:35] yeah i'd say it's very much a company-wide effort right so like yes we we are the growth team we [08:42] have done great i think we've driven a lot of impact but [08:45] Honestly, man, we can't claim too much credit for the success of the company. [08:50] We as Anthropic are really a model company and an intelligence company first and foremost. And so the lion's share of what has driven our success is our research team.

We have the best research team in the world. We have [09:04] great teams on inference and compute and then there's [09:07] many other teams like Claude Code, GoToMarket, etc, who I think deserve much more credit than us. I think just zooming out, going to some of what you said earlier, the growth trajectory has just been insane. That 10x year on year revenue growth trend has been there since the beginning, I think. [09:22] 2023 was 0 to 100 mil, 2024 was 100 to 1. [09:27] Last year was 1 to roughly 10.

And I look back to when I joined in 2024. [09:32] Revenue was in the hundreds of millions and [09:36] Just that trajectory to the end of 2024 and 2025, like week two of when I joined, we're going into 2025 revenue planning. [09:44] And we have these like base case and aggressive case scenarios. And Daria is pushing the aggressive case scenario. And people are freaking out being like, how the hell are we going to get that? [09:52] And Darya's like, I think we can actually go much higher than that. And I'm coming in like this.

This place is crazy. Like there's absolutely no way. [09:59] And that happened, right? And then you get to the end of 2025 and it's like, okay, [10:05] Law of large numbers, there's going to be a pretty big slowdown here based on your baseline rate of 10 billion. How are you going to keep growing at this rate? And like, it just like has not slowed down. And any of those numbers are public. The 19 billion number you quoted is from the end of Feb. So that is also out of date. [10:23] And it's absolutely insane.

The funniest thing is [10:27] something that I've noticed internally is like linear charts are just like not cool. Like no one cares about linear charts. Everything is log linear. So show me a log linear scale. And that's a scale we think. And I think overall, we're just, you know, really hanging on by the seat of our pants. So we're trying to manage the growth and [10:43] do the best that we can for our users. [10:46] I was talking to somebody at Anthropic about you. [10:48] And they said that basically anytime they want something to grow, they ask you to help.

And it works. [10:54] So you talked about just like [10:56] Things are magical and amazing and they like clawed in all the tools you all build are amazing innately. And that's a big part of the reason they grow. [11:03] I think many people listening to this will be like, what do you even do a mole with a magical... [11:09] micro god that just can do anything for you why do we need a growth person what do you even do [11:14] Talk about just the stuff that you focus on and maybe like a couple of the wins that your team has shipped that has helped accelerate growth.

[11:20] I would say they're not fully wrong. We're very, very lucky to have the best models in the world. We're very lucky to have [11:27] products like Cloud Code and Cowork, it certainly makes life a lot easier. [11:32] Having said that, I would say this is like the hardest job I've had in my life, and that's [11:37] Having been a founder, having been an investment banker and other things like that. [11:42] It's tough. And if I look at [11:45] what do we do as a growth team here? I think it's ultimately the same categories of things that you would think about at a normal company.

So we care about acquisition, how are we getting more people in the door, the intent of the people coming through the door. We care about [11:58] activation, the signup flow, funneling people to the right products, making them successful. We care about [12:05] things like monetization, free to pay conversion, pricing and packaging, all of that stuff. But the categories of work is the same. I think then the probably the big differences is I would say that like roughly 70 percent of [12:18] what I spend my time on is what we internally refer to as success disasters.

And that is where like, [12:25] Things have gone so well that other things are breaking now. And I think anyone who's worked at companies that have gone through rapid growth, you think like Facebook or Uber or Doordash early on, like they understand this viscerally. Where scaling this much just brings a lot of challenges. So if you think about each of those categories on acquisition, on activation, on monetization. [12:46] There's just a ton of... [12:48] firefighting, jumping from one urgent thing to another. And it's often extremely painful, [12:54] And it's funny because you look at all the charts, all the charts are like green, like fully up into the right.

And everyone's just like, it can be quite tough emotionally still. And so you need to sort of step back and just realize that we're very lucky to have these problems. [13:07] But that says 70% of my time, I'd say, is just these like firefighting of success disasters. And I think the 30% [13:14] remaining is just much more standard bread and butter growth work where it's like more proactive. So you think about [13:21] okay, if we have limited resources, which of the products [13:24] We have many different products. Which of the products do we want to put some juice behind?

[13:28] what is our long-term pricing and packaging look like, especially given that the technology is changing a lot and behavior and engagement trends are shifting. And then, you know, things like we have, [13:37] a lot of new products coming up, like, okay, you ship co-work. Now what? Like, when is the right time that we should lean in as a growth team to start optimizing the core adoption funnel for co-work? So it's probably 70%, just crazy firefighting, 30% more bread and butter stuff. Okay, I'm going to dig into a lot of that stuff.

One of the cleverest growth projects [13:57] moves you all made recently was this idea of importing [14:01] memory from ChatGPT, where he just made it really easy and kind of jumped on this trend of people getting really excited about Anthropik. [14:07] Is there anything you could share about the behind-the-scenes story of that feature? [14:11] who was thinking about [14:12] what can we do to improve the cold start problem and improve the new user experience i think that activation is a really big challenge in in ai and so [14:22] That's one example of something that we shipped that was very specific to a moment in time.

But ultimately, if you zoom out, it's like, okay, how do you really... [14:33] how do you really make it easier for people who are signing up to, [14:37] have Claude understand who they are and understand how Claude can help them and get them to the right place. [14:43] I want to follow that thread, activation, [14:45] There's something that comes up a ton when I talk to [14:47] People leading, driving growth on AI products. [14:50] It's just like [14:52] There is so much stuff trying to get your attention these days for people to get to a place where they, okay, wow, this is really going to be something I want to keep using.

It proves to be really hard. And it's also just unreliable. Sometimes it's not going to be magical. It's AI. It's non-deterministic. [15:07] I guess one is just like [15:08] How important is [15:09] focusing on activation, getting people to that aha moment, [15:12] with AI products and two, what are some things you've learned about how to do that well with, with Cloud or AI in general? [15:18] Yeah, it's a good question. I think that activation, it's critical, right? And defining that as like early activation, call it day zero, day one product experience.

I think that... [15:29] Historically, anyone who's been in growth or been in product understands that that's usually one of the highest levers that you have to actually even increase longer term retention. [15:39] I think that the importance of that has just gotten exponentially higher. Now, zooming out, I feel like one of the biggest... [15:45] problems in the industry is capability overhang, where the models are just getting better so quickly. And the real challenge is on the product side of how do we start to [15:58] diffuse those benefits to people. Even internally, there's new models coming internally and you're sitting there, you're so busy.

And when a new model's available, [16:09] you need to really carve out time to be like, "What can this do? How do I need to update my priors?" And if you think about more broadly for most people, [16:16] You may have a model that is like you may have [16:19] AGI or some model that can do all sorts of crazy things. But if people's instinct is to come there and be like, hey, what's the weather in SF? [16:26] then they're not going to get the most out of the product. [16:31] And so...

[16:31] I think that it's tough because the model capabilities are rising so much. So like if I think about, okay, back when we had, I don't know, say like Opus 4, there's a series of things the model can do at that point and Opus 4.5 unlocked a whole bunch of new things. You think about, okay, we sit there, we've got this new model Opus 4, [16:52] The time to then go and [16:54] run a bunch of tests, figure out, okay, there's the capabilities from this new model, what are the right on-ramps to guide people to those features?

You run tests, you get the learnings, you then ship a new flow. By then, you may already have the next model, which unlocks newer capabilities that makes all these learnings [17:11] irrelevant, right? So it's actually just a really difficult [17:14] problem to stay on top of. I think that many of the same things, same old trends in growth and activation remain, I think, accurate. To me, it's like, [17:25] Ultimately, some of the highest leverage is from [17:29] finding... [17:30] the right product or the right feature for the right user. And I think that one learning, you've seen this time and time again across companies, actually like,

[17:39] the right friction helps and adding more friction usually works if you do it the right way. I think that's something I've consistently seen that we've seen here as well. [17:48] So to me, I think it's really... [17:50] being able to identify which [17:53] what are the characteristics of a user that allows you to then recommend them to the right feature or product? [17:58] And not being shy about adding friction to do that, I think, is probably like the single biggest thing that's important here. [18:05] When I asked Ben Mann, one of the co-founders, former podcast guest, what to ask you about, and this is what he highlighted is your...

[18:12] experience, especially at Mercury, redoing onboarding and making it magical. And basically, he's in the same place as you of just how important it is for people to understand what the AI tool is capable of. [18:23] to help people. [18:24] decide to use it and stick with it. [18:26] Is there an example of something you changed in onboarding that helped significantly improve [18:31] activation. [18:32] Yeah, it's a great, great question. And I like that he brings up Mercury. I talk about their product a lot. So I worked on the growth team at Mercury.

I think it's a fantastic product. It's [18:42] something many people use and the reason they use it is because the better banking experience than yeah i'm a i'm a very happy customer just to put that out there i'd love it great i'd highly recommend it they have personal banking everyone should go go and use it right they just launched that i i and and and so i think that the interesting thing about mercury is like [18:59] the core value is that it's a better experience, right? That's the reason you use it.

You're not getting like [19:04] better other things, it's just it's a better product experience and so [19:08] That ethos is very, very deeply held within the company. It comes from a number of the founders and [19:15] I think that we had a big push one quarter when I was there on the onboarding flow. So onboarding flows for [19:23] banking institutions and regulated entities are extremely complex like [19:27] The amount of time I've spent on the difference between a registered agent address and a legal address and a physical address, these things, [19:35] very, very complex.

And we basically looked at the onboarding flow and we were like, okay, we've invested so much in quality and the rest of the product, [19:44] but we haven't really done it here, and this is the first experience that people have. And so we said, forget metrics, forget growth, forget everything else. [19:52] as the growth team on conversion, we're going to spend a whole quarter fixing quality in this flow. And so that's all we do. Forget the metrics. We're just trying to make this as good of an experience as we can and fix all these like [20:04] You go back from one field to the other field and adding in your beneficial ownership details [20:09] And it actually ended out being like, [20:12] And [20:12] probably until I joined here, like the single most impactful quarter that I've ever had as a growth VM in terms of the impact that it had.

And so we saw a significant uplift from basically our like [20:24] onboarding started to completion from just focusing on quality. And so that to me is like a broader learning around quality drives growth that I think I've tried to bring to Anthropic. I think for us at Anthropic, you know, some of the things that we've done in the onboarding flow, [20:40] is basically like we asked users questions around... [20:45] who they are, what their interest areas are, and we then use that to recommend different products and features and like [20:51] A number of people look at the flow and they're like, you have so much friction, it's such a long flow.

And I'm like, we have the data, we're kind of happy with how that's performing. [20:57] What is your kind of approach [20:58] philosophy on friction, good friction versus bad friction. I've just seen time and time again at every job I've been in and growth that [21:06] adding friction and adding the right steps leads to higher conversion and higher funnel completion. So you want to get rid of [21:15] And especially if you have high volume, you should test the majority of this and just learn and see, like, does this apply to your business as well?

But you want to. [21:25] You want to get rid of annoying friction that doesn't add value, but [21:29] The like... [21:30] I think the most... [21:32] simple understanding people have is like, [21:34] Just solve time to value. Cut all the steps and just get him into the product. [21:40] And that doesn't work most times. I think if you've really thoroughly tested your flows, I look at [21:46] the companies I've been at, Masterclass. If you go through Masterclass's purchase flow right now, you will go through all these steps in this quiz when you land on, when you're trying to buy and it's [21:58] You're like, I came here to buy and it's taking me through all these questions.

What are you here for? Et cetera. [22:02] I think it's easy to look at that and be like, why do they have this? This is like a terrible thing. Just cut it all. And it's like, no, like that's been thoroughly tested. And actually, that was like a significant revenue driver because it helps users feel that the product is for them by understanding what their interests are and recommending the content and classes there. One of the growth PMs on our team left to... [22:24] to join calm calm calm the meditation app if you go to calm's [22:28] their landing page and you go through their purchase flow, their login flow, you'll see a quiz.

It's not a coincidence. At Mercury, we also tested, I think, [22:38] I think you might post it on Twitter, like we broke out some steps in onboarding and just having one screen, if you have like five or six different form inputs and you often break that into two screens and reduce the cognitive load to people, that [22:52] that is something that performs well. We added steps into the flow there that actually performed well. [22:57] So I think that the takeaway to me is like, [23:00] cut friction when it doesn't [23:03] add to the experience of helping a user understand why the product is for them.

But if you can help users understand a product [23:13] why product is for them and how to use it and what's most relevant to them. [23:18] and it's going to add friction. [23:19] Don't shy away from it. Test it. Confirm that it works for you. But I think this is something that most [23:26] growth practitioners deeply understand. [23:28] And for them, it's really important there, what you described is, [23:32] adding friction to better understand who they are so that you know how to recommend the right thing for them. [23:38] correct yes and like that that like done right that just it just flows through right so it helps you with activation but then it helps you with life cycle you know more about those users and why they're here and and like [23:49] Most sophisticated businesses, you can then, even if someone drops off, you can do lookalike targeting and you can get them at the ad layer as well.

So that initial piece of how you understand who the user is, it's a juice that just keeps on giving if you then use it right for the down funnel. [24:05] Everyone's about to go do a bunch of teardowns of clods onboarding, masterclass onboarding. [24:11] Mercury onboarding. [24:13] Kind of as a tangent, I was at this PM dinner recently and I was asking all the PMs, [24:17] How was your... [24:18] role as a PM changed most with AI? Like, where is AI most impacted what you do? And one of the PM's answer that is actually doing competitive analysis, doing a bunch of, like, teardowns of what other people are doing for pricing pages and onboarding.

[24:30] So it's easy to do now. Just, hey, hey, co-work. I don't know. Would you co-work for this or Claude? What would you use for that? Okay, this is good. Help people. [24:38] pick which tool to use if you want to go do a bunch of teardowns of other competitors onboarding flows what would you use you can use you can use co-work for this right um so be a co-work with the chrome extension so [24:48] If you task CoEG with a Chrome extension, go and look for these flows. [24:52] and show me, give me a view of what's working, what's not.

That's definitely something that co-work can do. [24:58] Cool. I imagine that's one of the challenges is like you have all these tools now and just which one's for me. [25:03] By the way, the Chrome extension, I use it all the time. [25:06] That's amazing. [25:07] I want to drill a little bit further into the growth org. There was this whole meme on Twitter the other day of just like, you have one growth marketer driving all growth at Anthropic, and it's like, okay, that's crazy. [25:17] How many growth people are there?

What's kind of the rough org structure of the growth team? [25:22] we're roughly maybe 40 people now and so [25:25] We, I think, structured... [25:27] very much like a traditional birth team in that [25:31] We have sort of horizontals of growth platform and monetization who think about [25:37] the sphere of growth across the entirety of our products. And then we have more sort of audience-focused growth pods. You can think about like, [25:46] B2B growth, you can think about, [25:48] called CodeGross [25:50] knowledge worker growth, API growth. So really like these audiences [25:55] to keep a narrow focus, which is the thing you have to do when you have so many different products.

[26:00] And then these horizontals that sort of think about things across the board. And is it across... [26:06] a team of engineers, designers, PMs, what's kind of like the functions within the growth org? Yeah, it's engineers, designers, PMs, data. I think that [26:18] Overall, the shape of the org is quite similar to, I'd say, a traditional growth team. Probably the things that are maybe different is that we... [26:29] I think that we index a lot. [26:31] more towards larger swings as opposed to smaller optimizations. Like if I think about a traditional [26:38] growth team [26:39] I would have probably done maybe 60-70% of my time on small to medium bets, 20-30% on larger swings, and I think that [26:50] for us, we flip it a lot.

Like we do much more the other way where it's sort of, [26:56] 70-30 or more like 50-50 rather than... [27:01] indexing towards smaller bets. That's probably one of the biggest [27:04] changes, I think. Just to highlight that, what's interesting there is at the scale you guys are at, like a 1% win is massive. [27:12] in the scheme of things. So it's interesting that even at the scale you're at, you're not focusing on these micro-optimizations. [27:19] It is easy, like you could easily focus on these small optimizations and then you tally them up at the end of the quarter and like, look how much you made.

And like you could do that. Another billion. No big deal. But the thing is like we're we've been tracking a 10x year on year and we you know, that's like the thing that we sort of keep in mind. And [27:36] I think that like [27:37] ultimately comes down to our fixation of this company about the exponential. I think if you look at anyone who's talking about [27:43] talking from Anthropet about [27:46] basically anything we always talk about the exponential, like it's, [27:49] is effectively as model capabilities continue to grow on an exponential and the tools around them enable a better job of diffusing that into a [28:00] useful use cases [28:01] You basically just keep unlocking new markets where the value of those markets significantly dwarfs.

[28:07] what the value of the previous markets were. And like agentic coding is a great example. Like it didn't exist. [28:13] you know, a year, a year and a half ago. And then now just the value of agentic coding is is bigger than the like, [28:19] previous market of AI coding use. And so... [28:23] And I think that [28:24] is like the core thing here which is that the future product value is an order of magnitude higher than it is today and i think about [28:32] I don't know, like a normal business, like a number of companies that I've maybe mentioned or like you think about like a trading app or like a grocery delivery product, like [28:41] Many of the leading companies are great businesses, but if I think about

[28:45] What is the product value that a company, like a standard, like call it a grocery delivery app, like what is the product value you get as an end user today versus two years from now? I look at it as like, again, two years from now, even if you're shipping all these new products. [29:01] As an end user, the value you get from that product maybe goes up like [29:05] 30% to 50% if the company's done a really good job of shipping new features, [29:10] It's not exponential, though. And so if I think about, OK, you're here today, in two years, you're going to have 30 to 50 percent more product value than as a growth team.

[29:21] the relative differential of the product value two years from now relative to today [29:26] I can actually capture like a decent percentage of that with the small to medium optimizations that typically have higher conviction as opposed to like, [29:34] larger bets [29:35] But [29:36] And for up there, get... [29:38] It's not really that way where because of the exponential, [29:42] And our products being like very, very, very, the product value coming from AI, [29:48] The... [29:50] product value that we will deliver in two years time is probably like a thousand X, a hundred to a thousand X what it is today.

And so if I think about that and it's like, [30:00] there's so much value on offer, you need to shift more towards, okay, we need to take larger bets, and we need to not sort of miss the forest for the trees. And so that's why we do... [30:12] we still do all the small optimizations. It really matters. And like no one else is going to do some of these things. And so we need to do it. The compounding value is not immaterial, but [30:21] we do take on much larger core product type of swings as well.

So you mentioned that the Chrome extension, like that is now... [30:28] The thing that underpins a number of use cases on cowork and, and, uh, and code code as well. [30:36] And that's something that the growth team built that's like a very like, [30:38] AI-peeled product that is a very research-heavy product that [30:42] We were just bullish on it. We had an engineer who was very bullish on it. And we were just like, hey, no one else is doing it. We're going to do it. And so that's the sort of thing that I wouldn't have done at another company.

[30:53] Oh, wow. I did not know that. [30:55] So the takeaway, one takeaway here is, [30:58] Like, you know, there's like stuff to extract from your advice that is like only true philanthropic. And then there's what can other companies learn from this experience? [31:06] that you've had. [31:09] So... [31:10] One is... [31:12] Is your sense that if you're working in AI, shift more of the pie chart towards bigger bets? Because in the future, the opportunity is so large, you want to find those as soon as possible versus micro-optimize? [31:24] To be more specific there, I'd say that if the primary...

[31:29] value that your product delivers is underpinned by AI, [31:34] as a central element of it, then I think you should operate this way. [31:39] I don't know, companies like [31:41] yeah lovable cursor you know all these like great businesses that are like it's [31:44] as the exponential rises, their value crops are also going to continue to rise significantly. Like if you're building a product where it's like it's an AI first product, [31:53] then I would definitely operate in this way. [31:55] I think if you're building a product where it's not necessarily an AI-first product and you have some AI features, [32:02] that are on the side, but it's not the core of your value.

[32:06] I don't know that I would operate this way. It would depend on how is the rest of the product all staffed and how is the growth staffed in relation to that. [32:14] Okay, awesome. [32:16] And then in terms of the way you're structured, I thought that was really interesting. It's like a combination of different sorts of things. So there's like [32:22] the API growth, there's cloud code growth, but then there's also like personas, like a vertical of like, [32:28] knowledge workers and B2B [32:30] Is that intentional, like some specific bets and one just kind of like broad market opportunity?

When you have like one product, it's easier to have a growth team that's more like purely on the funnel, right? It's like you have the conversion, you have activation, you have monetization. But as you start to get into having multiple products, I think that's harder because [32:48] If you do that, then if you just have one activation team, for example... [32:52] but then you have code code, you have co-work, you have all the other things. They're very different audiences and they're very different sets of cross-functional stakeholders internally. So we're kind of looking at, [33:02] what is the thing, and all structures are not perfect and they're right for a point in time, but we're looking for what is the structure that allows us to [33:11] have as much focus, like focus is a really big thing and on audience and [33:16] and problems and also the tie-ins to cross-functional partners is really, really important.

Like the folks in our Cloud Code growth team, like they work extremely closely with Kat and Boris and the others on that team. And so that tie-in is really important as well. [33:34] So you've done growth at a lot of different companies, a lot of basically, let's call it traditional growth before Henthropic. [33:40] How else is growth as a function and as a skill set changing with the rise of AI, AI products, AI startups? I think if your core product value is very backed by AI, then [33:51] It is shifting where you're skewing more towards...

[33:55] larger bets as opposed to smaller medium experiments. [34:00] I think in other things, [34:03] Maybe a big thing related to this, [34:06] that I think will accelerate this, which I am really interested to see how it will play out is [34:11] we are starting to look at how do we automate growth, which I think is a really interesting area. [34:17] Our growth platform team, we're very lucky we have like [34:21] Alexei Komisarok, who teaches growth engineering at Reforge, and he's just like the guy on the team. And so he is...

[34:28] driving this effort. [34:29] Um, [34:30] that it's it's the name is it's a little cringy i didn't come up with it and it's like it's called [34:35] CASH, which is Claude Accelerate Sustainable Hypergrowth. I did not come up with that. But really, it's an effort to look at [34:44] how can we use Claude to automate growth experimentation? And, [34:48] It is still very, very small. It's still very, very early. We kicked it off only, I think, a couple of months ago, I think before Opus 4.5. [34:58] it wasn't really possible.

We were just like not seeing the results and [35:02] more recently with Opus 4.6, we're like, okay, this is like headed in the right direction. And so... [35:07] This, I think, is... [35:09] It will happen more and more across the industry where... [35:13] Basically, if you think about, okay, I think this can happen all across product, but growth teams in particular, because [35:19] there's this whole body of work that is very small optimizations. I think I just more inherently... [35:24] suited to tackle this earlier. [35:28] I think if you think about the life cycle of shipping, there's this sort of four parts to it.

One is [35:34] is identifying opportunities like how good is Claude at actually identifying opportunities based on different trends, based on previous trends that Claude has seen in the past, [35:44] Second is then building the actual feature and getting it ready to ship. [35:50] Third is testing and ensuring that it meets your quality bar and your brand bar. [35:56] And then fourth is then once you've actually shipped the thing, analyzing the data, gathering the learnings. If you think about that as like the loop of, okay, these are four things that you can eval and hill climb on each of these areas and understand how good is a model doing for you there.

[36:13] And, [36:14] We basically think about it in that four ways. [36:19] and we are scoring how good is Claw doing in each of those areas. And so we've been testing this. [36:25] pretty small scale right now it's been a lot of copy changes and some like very minor ui tweaks [36:31] it's delivering results, right? And it's like you can push it, press play with it, and it ultimately prints money where I'd say that the win rate [36:40] is like i would expect a senior pm to do better like i i would say like this is like a junior pm like [36:46] two, three years in, I would say this is like the win rate that I would expect from a junior PM, but it's not quite at the senior PM level.

[36:54] Although, I think, like, you look at the exponential, this wasn't available at all a couple of months ago. [36:59] It's getting better rapidly and I think that [37:04] the the it's going to change where you'll be able to do this for larger and larger types of experiments [37:11] But then, you know, when you think about the largest types of experiments, I think the [37:15] I mentioned the four pieces around identifying opportunities, building, testing, and shipping. The one I didn't mention there, Lenny, is... [37:24] cross-functional stakeholder management. There's still a need for human brains.

Yes, there it is. And I think that that one is... [37:34] is going to mean that, in my eyes, the work of PMs is not going away, actually, anytime soon. And that piece [37:42] especially for larger projects, you don't need to do as much of it for smaller stuff, right? You can skip it, but for larger stuff... [37:48] that piece is not going away. Until the other stakeholders are their own little agents running around. I think that's right. I think that would be the point where it would change. It's funny, we had a difficult meeting a couple of weeks ago, [38:00] Me and our head of design, Joel, we were debriefing afterwards, and he pings me just like, [38:05] Amol, we will have AGI and it will still be impossible to get six people in a room to get to a line.

[38:11] And I'm like, yeah, I think that's I can see that. [38:14] That's a, like, what's the harder alignment problem? [38:19] No. [38:20] Okay, this is so interesting and this is exactly where I feel like things are going. So just to be clear what you shared here, there's basically this tool that comes up with [38:30] experiments to run to help grow Claude and all the tools. So it comes up with the idea, [38:37] Somebody looks at them, proves cool, let's do these things, builds it. [38:42] Ships it, tests the results, see how it's doing, and then comes back like, here's things that are working.

Is that roughly right? It's roughly right. And like we right now we have human in the loop approving, but... [38:53] like the amount of time I kind of think about [38:55] scale in this way of just like week on week, are things getting better in each of the areas? Are people spending less time on each of the areas? Are the results getting better into the areas? And as long as like, [39:05] week on week, that's getting better than you're like, okay, this whole initiative is like scaling. And so that's roughly right.

But you can think about it as [39:13] I think a lot of this can be automated, whereas human review is not needed. So we care a lot about brand, right? [39:19] That is something that we do look at right now. We don't want to be shipping something that goes against the brand. [39:24] But then, [39:26] You can have a skill that [39:29] contains your brand guidelines and very clear yet do's and don'ts on brand. And so all of these types of like [39:39] Accompaniments, I think, are going to get better.

The model is going to get better at understanding how to use them. And so over time, I think that the need to like [39:46] human review, this really, really decreases significantly. Yeah. And you could always unship it if it's like, oh, yeah, that was actually not a great idea. And that's such a good point that we like we think we need people to do these things. [39:57] forever and it turns out okay a skill could do this really well here's our brand guidelines here's our vision here's our mission [40:02] Here's our goals.

Here's what matters to us. [40:04] Okay, let's not ship that thing. [40:06] So the reason I think this is so interesting, this is like I've just been watching the expansion of AI doing more of the product development process. In this case, the growth process of just, OK, it went from helping you write code to like writing all your codes, reviewing your code. Now it. [40:22] It feels like what are the other ends of this, the two ends around this? It's kind of going from the middle out. [40:27] The top of that is coming up with what to do.

[40:29] And then there's like the alignment stuff, still very hard. And then on the other end, it's reviewing the code and then shipping it and then get distribution. That's like a whole other thing I want to talk to you about. [40:39] So what I'm hearing here, and this is exactly what I thought was going to start happening, is AI is now getting really good at telling us. [40:45] what to do, not just [40:47] taking our orders and building it. [40:49] and it feels like the growth [40:51] version of this is where it starts because it's so much [40:54] simpler, just not that growth is easy, but just like it's data driven.

There's this loop that you talk about. So I think this is such an interesting sign of things to come across just generally product. [41:04] Yeah, just putting that out there. Okay. [41:07] So [41:08] Something that I'm constantly thinking about along these lines is just the future of [41:12] Product PM Engineering [41:14] how those roles shift over time based on the stuff we're talking about. [41:18] How are you working together? Is it? [41:21] as a triad and where do you see these roles going what's most going to change do you think across these three roles this is like something that we [41:29] talk about and think about frequently and the [41:32] picture changes rapidly.

[41:35] Sometimes when things break, execution like the bottlenecks break, [41:39] historically in the past it's like okay now you need to hire more engineers you need to hire more designers etc and it's more just like a [41:45] a life cycle thing of like where is the life cycle of your team and that specific pod to identify where the bottleneck is but now when when things break you need to still look at it of like okay is it like the actual ratio or is there also underlying technological shifts that are also causing this to break and so that's like a [42:02] An interesting thing, I think that smaller companies, if you find like a 15 or 20 person company, I'm like the only PM there working with some designers and engineers.

[42:11] I think [42:12] I think in the smaller companies. [42:14] you'll see probably like the biggest blend where like the PM will be doing all sorts of, they'll be, you know, designing, shipping, etc. And I think you just have, [42:22] extreme bifurcation [42:24] You know, larger organizations, more scaled organizations, I think the jury is still very much out. Like you speak to people even internally here and different people in different teams have different views of like, OK, how much are these roles coming together versus how much are these roles coming together?

[42:39] going to be separate. [42:41] I think that [42:42] It's not to say like even the PMs, like number PMs are shipping and themselves shipping and pushing PRs, etc. But if I look at, okay, across... [42:53] What I'm seeing... [42:55] I think that it's clear that [42:58] while PMs and designers are getting more leverage from AI, engineering is getting the most leverage right now. I look at tools like Cloud Code and the amount of leverage engineers are getting from them is higher than I think the amount of leverage that designers and PMs are getting from them today.

Now, this is rapidly also changing, but that to me is my view today. [43:20] If you think about, okay, a default team... [43:24] which is, say, five engineers, one designer, 1PM. [43:28] with Chord Code, [43:29] that five engineers is like two to three xed, right? And the PMs and designers have also increased, but now they're managing what is effectively a much larger group of engineers. And so even though like the head count and the org structure hasn't changed, you're now just dealing with a situation of maybe [43:49] 15 to 20 engineers [43:51] in the old world, [43:52] one and a half to two PMs and like maybe one and a half to two designers.

[43:57] And so we're seeing that that's putting like a lot of strain on PM and design and it's [44:03] It's not everywhere. Like I look at teams like Claude Code and I think, [44:06] that org, because that product is so technical, it's probably just the right thing. The PMs are all basically engineers themselves anyway. But we had a PM lead on site the other week and we were all just talking about this. Across the board we're feeling this, where PM and design is just squeezed. It's just absolutely squeezed. [44:27] is the right thing here, we just need to actually hire a ton more PMs.

And that could actually be where we land [44:35] You know, on growth, how I think about it is like, [44:38] One, we are hiring a number of Growth PMs. We desperately need people who are very, very good, and we are hiring. [44:47] If you are excited by what we're doing and you know growth, please feel free to apply. We'd love to chat. [44:55] And, um, [44:57] Maybe craft an amazing cold email to you. Yeah, yeah, yeah. Feel free to craft that email. [45:04] So that's one is I think we are going to be hiring a number of PMs.

But then the second thing that we do is we very much hire... [45:13] Product minded engineers, I think this has always been the best thing to do in growth. Like you always want to have the engineers coming up with ideas, etc. And so we especially especially people who can really like step in as mini PMs if people are if the PM is absent. And so we're like basically more formally leveraging that right now because we are so stretched. [45:34] So the frame that we have is that if a project is two weeks of engineering time or less, [45:42] then the engineer is on the hook to effectively be the PM for that.

And so that means things like, [45:50] talking to security, talking to legal, talking to cross-functional stakeholders. And the engineer is very much driving that. The PM will get looped in and they'll advise if needed. [45:59] And if something is like like wildly going off track, then they'll step in. But they're much more in an advisory capacity versus execution. If a project is more than two engineering weeks, then the default is that the PM should continue to to be on the hook for for making that go well. And this will delegate more to end, but they're like squarely accountable.

It's not like fully clean cut. It's like use your head like if this is a one week thing, but it's extremely, [46:25] controversial like the pn should probably still drive it but that that i think is like the the approach that [46:31] I expect more companies will start to do, which is just deputize the engineers to be [46:36] Now, not everyone can do it, right? So the PMs, the engineers who are more product minded, suddenly their value goes up significantly, like an order of magnitude. And then I think we will probably still be hiring a lot of PMs.

[46:52] There is so much interesting stuff I want to follow up on here. Okay, so one is this idea of two weeks. Just briefly, I always joke as a PM, [47:00] You can go on vacation and be away for like a couple weeks from your team and things are going to be all right. Like, you know, there's like a momentum, there's a plan, people keep operating. [47:09] And it feels like that's kind of this rule of thumb you use of just, OK, if it's a two-week project, you'll be all right with that PM.

You can handle it. [47:16] I love that those two connect. Okay, the other here is so interesting. So you're saying here that because engineers are so accelerated in the salt makes sense. [47:23] PMs in design are kind of just like, holy shit, it's hard to keep up with the space of engineering. And what you're saying is you need more and more PMs. [47:32] to keep up. That's one route. Or it's engineers that can PM, essentially. [47:36] Which is so funny. It's just like, okay, great use for product managers until more of the PME stuff can be done by AI.

[47:43] But that's a really interesting trend. I don't know, is there anything else there just like, "Oh wow, we actually may need more PMs." The ratio of more PMs to your engineers might be the future. Yeah, I think this is like, it just really depends on the industry, the size of company. [47:59] Like any company where you're building something that's like much more developer focused, like you're going to rely on the engineers a lot more. [48:06] Earlier companies, you don't have as much of this cross-functional coordination stakeholder alignment nonsense that you need all these plans for.

So you can get by with less. [48:17] But then as a company scales, [48:19] And. [48:21] If I think about, okay, now you have this ratio where [48:24] Maybe the 1pm became 2pm from productivity. [48:29] that like five engineers became like 20 engineers. [48:33] The one designer maybe became like three designers. [48:36] If you think about what is the best use of time for that PM, I think this is a really interesting thing of how much should PMs be actually shipping things themselves versus [48:45] everything else. [48:46] I think like in the world where you're like limited on engineering, the PM should definitely be shipping things.

I think in today's world. [48:54] It's a good way to get... [48:57] an understanding of the tools, which is really important. So PMs should be shipping for that reason. But if I'm 1 PM or 2 PMs and there's 20 engineers... [49:06] I think about what is the incremental value I can add with my time, and is it actually shipping like, [49:12] the 21st PM feature? Or is it saying, how am I getting a little bit better at guiding the team on what the right opportunities are? And so that's where I think like, [49:22] In this world, you may have all these engineers who are like mini PMs, [49:26] And the better that happens, that's where I would love to be doing more of.

But still, if you then get a really good PM who can come in and can like [49:35] improve that. [49:36] the why and the what particularly by 5%, that is such a high leverage hire. [49:44] This is such an interesting insight you're making. It's like so [49:48] counter to how a lot of people are thinking PM is evolving. Like what I'm hearing here is [49:52] Because PMs are so behind, because engineers are just getting so much done, there's... [49:57] like many people here, okay, you need to be prototyping, you need to be shipping PRs as a PM, [50:01] What you're saying, which I completely agree with, is your time is much better spent helping [50:06] PM basically and helping the engineers, uh, [50:10] become better PMs themselves and the leverage there is a lot higher than you spending time coding, shipping PRs in most cases.

I think that's true in certain circumstances. I think that as a smaller company, I don't know that that's the case. In a small company where it's all hands on deck, I think you probably need to be shipping. I think [50:28] If you're in a company where like budgets are very tight and engineers are very tight and [50:33] you're not able to just hire because money is unlimited, then you need to do what is needed to accelerate the impact your team is going to have. So there's going to be a number of cases where like,