The Library of Minds: Product Expert: The Future is Fully Automated Software Development | Spenser Skates, Amplitude

AI is going to suggest what to build. It'll test different variants, it'll get it out there and it'll say, okay, this variant performs the best. If you're selling to businesses, yes, absolutely charge more. Right now it's like I get the same piece of software as you get the same piece of software as everyone here. In the future, it's going to be like, no, no, no, this is all going to be personalized.

Hi everyone. Super excited to have Spenser Skates here, co-founder and CEO of Amplitude, pioneer of the modern product analytics movement and a CEO who is still innovating 13 years later and pushing the limits of what product analytics looks like today. And I was telling him earlier, it was very hard to scope these questions to one page because there's so many aspects of your life that are interesting. But I will start with actually your first company, Sonolight, which was pretty much Siri before Siri in a way. And I was reading that it had 500,000 downloads and you still decided to pivot. And I feel like most founders would have had ego or thought about opportunity cost and sunk cost fallacy. So I'm curious, what gave you the conviction that moving away from this thing that had a little bit of traction was the right move?

Yeah, so we had, we didn't know what we were doing when we first chose Sonolight to work on. We wanted to work on something on the bleeding edge of technology that was a hard problem and almost seemed impossible to solve, but we thought might be solved. As it turns out, voice recognition is just too hard of a problem to solve, or at least at the time. Now it looks like there's been an explosion with everything in AI and a lot of the breakthroughs in models and transformers. It looks like it actually is a lot better in the last few years, but at least at the time back in 2012, it was just too difficult of a problem. I think there were kind of two things. So we could have, we spent about a year on it and we could have, continued on it and we kind of saw it'd be an okay consumer app and maybe we could get to millions of downloads and have, there was someone using it. It was like, I think we had like somewhere in the 10,000, 20,000 daily active user range. So it's like, okay, someone who's not us is using it. This is great. But one, I think we realized it would've been a real grind. Users were not sticky enough within the product. Like they weren't engaged in retaining. And then the other was that we just weren't that authentically passionate about the space as founders. It was more like, it's not like we used our own product a ton. It was more like, okay, this is a cool idea on the bleeding edge of technology than something that we were personally passionate about. And so after a year of doing Sonolight, we said, okay, let's actually go work on something else. What happened at the same time was there was a lot of— we had built our own analytics like a lot of early-stage teams do, and we had all these insights around how— the first question we wanted to know was how important the accuracy of the voice recognition was for long-term engagement. As it turns out, it's extraordinarily important. If you have a first accurate recognition event, you are twice as likely to retain long-term than if you don't. And that's part of what made us realize like, okay, hey, this is not going to be a viable long-term product. But as we showed that insight to other teams who were much more successful than us, to our surprise, they didn't know, they didn't even know really basic stuff about how people use. I remember PlanGrid, which is one of the standouts in our batch, didn't know anything about how people were using their product. It was just very, very basic stuff like how many active users do they have and that's it. And so them, lots of other startups in our Y Combinator batch were extraordinarily interested in this product, what we had done in analytics and said, hey, can we, can we get this? And so that was our first signal that there was an opportunity there on the data analytics side.

Enjoying the conversation? Got a question? You can ask Spenser's digital mind. Just check the link in the comments or description. Yeah. And this led to your first paying customer. Brett Terrell, who asked you for a price. You almost said $50, but then you ended up saying $1,000.

$1,000. And you said that's the largest number you could think of. Yeah. At the time.

And now, you know, at the time and maybe now the advice would be like, always charge more. Do you think that advice still holds up in this era of commoditization of software? How do you think about charging more versus like trying to be the customer that grabs the brand value and like the customer relationship?

So I want to distinguish a few different strategies. If you're selling to businesses, Yes, absolutely charge more. When we think of buying or selling things as consumers, we think of it as like, okay, hey, you know, what's expensive? Like, you know, what do we spend money on? Like, okay, you know, we're gonna go get a coffee for $5 or we're gonna, you know, spend $50 on a dinner or $30 an Uber ride. Like that's our frame of reference for what stuff costs and what's expensive and what's not. In comparison for businesses, their frame of reference for what cost is like 10, 100, 1,000 times that. Like the fully burdened carrying cost of an employee for a place like Amplitude is like $250,000, something like that a year. And so, if I wanna spend $1,000 to make that person more productive, that's a no-brainer. Like I'm gonna do that all day as a business. Like even if it gets them a little bit of a benefit, like I would be very willing to do that. Or even $10,000. That's not an issue. And so the reason that there's an advice charge more is because our frame of reference as consumers is much, much smaller than what businesses actually care about. And so for a business, like $10,000 is almost nothing in terms of their ability. If they have a few employees, at least that's not, it's just not a big deal to them if the pain is large enough. And if it is, That means, okay, you probably don't want to spend time then. You want to focus on elsewhere. So we quickly went out from, you know, so it was $1,000 a month and the next contract was $2,000 and $3,000 and $4,000 and $6,000 and then $8,000. And then we got to $10,000 a month like within the first few months of selling this product. And it was like, okay, you know, yeah, every time you successfully sell one that builds your confidence to ask for more. And then that was a sweet spot for a little bit, you know, to ask for $10,000 a month or $120,000 a year. We had a number of customers that came in at that. And then, you know, but you know, then we'd work our way up there to, you know, next contract was $270,000 with this company like 360. And then, you know, we kind of, and for these guys, like, you know, to us it was a huge amount of money 'cause we're like, oh my God, like someone wants to pay us for software. This is the best thing ever. Like it's not worthless, you know?

Yeah. And speaking of bleeding edge, you know, you've talked about the future of software potentially looking like something that's self-improving. And I'm curious, as you think about Amplitude's product roadmap, how do you decide whether you're ultimately building a product whose end user is an agent or a product whose end user is a human? You can invest in dashboards for humans to analyze data and take actions, or you can invest in data that leads to agents taking actions. I'm curious which of those features you're building for. Maybe you're building for both.

Agent thing is very early, to be clear, We're starting to see it in the traffic, but as people build websites and interfaces for these agents, like, yeah, we have to keep up and build stuff for them too. So we're actually coming out with an agent analytics product in the next month or two and already have a few customers trying it out. So that's good to see. I think what we are on a bet on at our core is the interface between between the human world and the digital world. And whether that shows up in a mobile app or a website or a chat interface or VR or AR or a device or whatever, we're a bet between understanding that interaction. And if you look at it, that interface, we're frankly melding with technology in a lot of ways. We have computers that we carry around in our pockets that we go around on all the time. We have all sorts of devices. We have smartwatch, there's going to be more and more of these things. And so we're a bet on the interface between humans and machines increasing. Now people are like, oh, well websites are going to go away and apps are going to go away. And it's like, maybe, maybe not. But the important thing is like, are humans going to use technology more or less in the future? And it's like very clear if you look at the curves with AI as well, it's only going one way. And so, and even so, even if the, interfaces and form factors are very, very different in the future, that's all a good thing for us. To us, it's about how do we make sure to create the best experience for those humans on the other side of the technology.

And when you are analyzing data on humans and interfaces, do you imagine that humans will, on the other side of the Amplitude dashboard, are going to be the ones making the final decision on what to update? Or are we going to delegate the decision of, you know, this user isn't making it through onboarding, What does that mean for a product?

Oh, totally. That's going to get automated. I think this is actually where software is going. So today you have, it's a very manual process. We talk about the whole product development loop. So it's like you ship software, people use that software, you learn on their usage, and then you decide to build the next thing. And so what's happening now is the shipping side is getting more and more automated, right? I mean, just in the last few months, you have these agentic harnesses that like go ahead, can code tickets end to end, and you're just reviewing pull requests. And the best engineers are not even writing any code directly in IDEs themselves. They're just reviewing code that's AI generated. And so that's gonna go faster and faster and faster. And so what that puts pressure on is, okay, how do I know the right thing to build? And while, okay, you start to look at the data, you start to look at analytics, you start to use systems like Amplitude to do that. And that side is going to get more automated too. And that's what we're very focused on is how do we automate and scale that so that, and then we can feed that back. Today you have a product manager who looks at some data and it's like, okay, this feature's good, this feature's bad. Let me double down on this one and cut this other one. They hand that off to an engineer that gets built and it's like this whole long chain of things that need to happen for that to be successful. In the future, it's like those signals are going to be picked up automatically. AI is going to suggest what to build. It'll test different variants, it'll get it out there and it'll say, okay, this variant performs the best. Let me go ahead with that. And then it'll also even one step beyond that, it'll personalize that on a per-customer basis. Right now we can look at user timelines or user sessions and say, hey, we think this user is trying to do X based on its history. And so let me actually personalize the experience to help to make that easier for them. So this is where this is all going, where you're going to get fully software that fully and dynamically responds to us as humans and then evolves. And right now it's like I get the same piece of software as you get the same piece of software as everyone here. In the future, it's going to be like, no, no, no, this is all going to be personalized. So in the same way that Netflix recommendations or your TikTok feed is personalized, these things are going to be personal. Every single, it's not just going to be the feed, it's going to be every single aspect of the software is going to be personalized for each one of us, which is kind of crazy. Easy to think about.

Yeah, we all have different experiences.

This is the world that we're entering into.

Yeah, I read that in the early Amplitude days, a lot of your, one of your narratives was, we used to be in the Mad Men style intuition-led marketing and now we're in the data-driven era. There's a lot of talk now on Twitter where intuition is back, taste is back because everyone has access to the same data sources. I'm curious what your take is on intuition and taste being the only moat left.

I take the other side of that actually. I think people who learn to leverage AI to use to create things like, you know, figure out what good taste is and figure out what great experiences look like are going to do so much better than, than people who ignore or don't understand or don't look at that data or don't know how to leverage AI to do it. And so I, I'm a technologist where I think in this revolution, people who understand how to harness this the best and the deepest are going to outpace everyone else. Now, maybe, you know, taste as a part of that or what have you. But I, yeah, I'm kind of on the other side of it. Like, you know, if you, again, if you look, if I look at the best engineers at Amplitude and what they're able to get done, it really is how fast and how much can I do with the prompting I'm doing. And if you're able to do 10, 20, 100 times more than someone else, inevitably just by virtue of the quantity that you get out there, you are gonna develop much better taste than anyone else. And so I think it comes back to your ability to leverage whoever figures out the best to leverage AI as a technology is going to do so much better than everyone else out there. One of the really interesting studies we did was with one of our early customers, Calm, is a meditation app. And one of the things they found out was that the most important use case they had was not actually the walking through meditation. The most important use case they had was helping you develop a regular habit around meditating. And so that reframed what the purpose, it wasn't just, hey, let me have the best walkthroughs for meditations. It was like, okay, let me help you figure out how to schedule, set the right time and like remind you then and like, you know, okay, are you a morning person for meditation, evening person? You know, when do you like doing it? How do you like reminders? And that was huge. And that is now why they're the number one meditation app is 'cause that understanding of, of their users. And that came from the data. That actually was not intuitive to the team or anyone else doing it at the time. And so it's like one of these, I can't remember if it was Einstein who said this, but if we were smart, if we were intelligent enough, we would never need experiments. We'd never need experiments. We would've figured out all of physics and science and math and everything else on our own. But because we're not, because we perfectly cannot predict the future and perfectly understand systems, experiment and data is actually extraordinarily useful to us. And so there's always gonna be a frontier where—

To complement your flaws.

Yeah, understanding the world and reacting to what you see is gonna be valuable.

Obviously talking to customers when you're building a new product can sometimes be misleading because they wanna flatter you, they say yes to everything.

Yeah.

I'm curious, what are those tactics that sales coach taught you to get to the truth of what people actually wanted and how you could build a profitable business around it?

The funny part was of those first, we didn't want to make the same mistake we did with Sonolight, which was not talking to customers and really validating our idea before doing it. What's ironic is of those first 30, none ended up being long-term customers of Amplitude. Some of them tried it out and yeah, so it's like we were really, it was really going to take a few iterations of going through customers before we found a problem. But we knew this problem was going to be a big one and it needed to be solved. The internet was changing, there was the rise of mobile and that would require a whole new set of infrastructure. So we were convinced that this was going to be a valuable problem to solve. One of the things Mitch did that was extraordinarily helpful is he would always ask me, what's the business pain? He had this great way of framing it where he asked, you could ask someone, okay, on the scale of 1 to 10, what's your pain on this problem? And if someone says somewhere between 1 and 5, then it's not painful enough for them to care about. It might be a nice product and everything, but they're not going to go spend money or take the time to try out this company that has no track record and brand and anything else. Whereas if someone's an 8, 9, or 10, then it's like, okay, now they'll have— that's a signal and you want to lock onto that. And so that would be one great question. The other would be just understanding why it matters for their business. And so they want this particular report, but it's like, okay, that's not a business pain. How is it impacting their business and why is it, why do they care so much? Is it they're trying to save money on doing something that's already really cumbersome to do? Is it they're trying to make more money because they want to increase their particular metric? What is the either save money or make more money reason that this connects to? And so eventually after getting beat this into me over the course of a few weeks, I'd go into these meetings and I'd finally be like, okay, I got to ask why. Why is it they want that? And that would help me understand because most, one of the things you come to appreciate is that most customers are not a good fit for your product for whatever reason. It's not the right timing for them or something else is a better fit or they don't You know, it's not painful enough for them to care and spend money and take the time to try out your product. And that's okay. That's okay. Being okay with that is one of the hardest things because in the early days, like most people, you're desperate for any attention, right? Like nobody cares about what you're doing and you're sure you're trying to bootstrap this positive feedback loop and you're just desperate for anyone who's going to pay attention. So someone's taking a meeting with you. It's like, okay, you really desperately want to turn that into something real. But what you quickly realize is that you can waste all your time doing that. And what you want to get to much more quickly is the fast no on something, or it's not that painful, and then just deprioritize them until it's later painful. So that way you can spend time chasing people who you, who you do think this is painful for. And so that was a really great early lesson in terms of going from being an engineer to being a salesperson. That was extraordinarily important. And That was important to me because there were so many great engineers who built amazing products and amazing pieces of code that wouldn't go anywhere because they didn't take the time to understand how to do this process of customer development and sales and getting this product into the hands of customers. And I was like, okay, I want to make sure that we don't suffer the same fate. And then on top of that, one of the things I really get an appreciation for was how deep and complex and important of a problem the sales was to the success of a business as the engineering. It may sound extraordinarily obvious. There's so much information out there these days.

The product doesn't sell itself.

But it really does not. And you have to put in the work to figure out, it's like they say, first-time founders focus on product, second-time founders focus on distribution. The distribution of the product, it's different. Obviously you guys are more of a consumer thing, but it's just as important as the great piece of technology that you're bringing to the market.

Couple more questions. What's one thing happening in startup land right now that you disagree with, if any?

Yeah, I don't know. And this goes to show how old I am at this point. But back to this length of time thing, there's a lot of people chasing heat right now on this stuff. And there always is with any new technology and any new hype cycle. And so most of the companies that you see that are raising these gajillion dollar rounds or whatever in AI, they're not going to make it. And that's because they don't have the right outlook and the right long-term mentality on these things. And so that, I don't know, maybe that's what, I don't know. I feel like, yeah, yeah. My day was much harder. So yeah, the missionary element has gotten more diluted unfortunately. But I don't know, the great founders, I mean, part of the reason Anne wanted to work with you, Dara, As like, he, he, he, she, she saw that, she saw that. So, you know, I think, yeah, I think, I think that matters more than ever in terms of the difference in outcomes for people.

What's one question you wish you were asked more?

What's one question I wish I were asked more? I think people misunderstand Amplitude as a business and kind of what it's like. Yeah, we're SaaS and it's like, you know, we do these analytics data and all of that, but kind of the what are the underlying bets that we're building the company around? That probably means I should just talk about it more. But yeah, I do wish people asked that more.

Alfred Lin recently had a blog about survival and just refusing to die. And you've been at this for 13 years.

Yeah. And 50% of the variance in outcomes can be explained by length of time. Yeah. And it feels like you're kind of going stronger than ever.

Like you've been posting more on Twitter. You're like, acquiring companies, you're working on AI. I'm curious, what's kept you sane 13 years later? Anything? What's giving you the energy to just survive?

I spent a lot of time before starting Amplitude thinking about why I'm doing this. So that was the most important. I was always going to be an ambitious and hardworking person. And the question is how to channel that energy and One of the things I took away was that if you can attach yourself to a mission that's greater than yourself, you can have that last, you know, forever. Whether that be in business or whether that be in, you know, if you're an academic or in science or, you know, if you're a sports or an entertainer or whatever, like if you can attach yourself to a mission that's greater than yourself, you know, that can last forever. And so I've always been a big believer in the power of technology to improve people's lives. And this is the best way I have to do that. And what's been interesting is just looking at and studying the stories of other entrepreneurs. Most don't last, there's about 50% drop-off after 2 years. So pretty steep curve. And then after 10 years, even if you're successful, most founders at that point quit because the the game and the problem changes again. You have to go from a founder to being a large company executive, and it's very different and disorienting. And so I've seen so many founders at that point leave. If you look at the companies of our competitors from the early days, I mean, they've all left by this point. And so that was a whole thing where it's like, look, I believe in this thing and I believe in what I'm doing, and how can I set myself up lifewise to be able to do that? Now, The funny part, I was actually too extreme in a few different ways, and I've had to moderate. You know, once you're— you can be like a missionary zealot a few years in, but you know, once you get past 10 years, it's like, okay, you know, there's other parts to life. And you, like, very lucky to be with Anne, for example, and you have to figure out, you know, you have to figure out a little more balance on some of that stuff. But it's, you know, so I've done, been doing that over the last few years in terms of thinking like, how do I, How do I set the rest of my life up well too? Not just what I'm doing here as a founder, but yeah, it's like that, like being convicted. Paul Graham has this question, which I think is incredible, which is, is this your life's work? And either answer is okay. Like it doesn't have to be, or it can be. And so for me it's like, yeah, like this is what I want to be doing my life. Like this is great. Like this is a great mission for me to be on and I can put myself into it and I can, if it goes well for decades, the more the better. But it's okay if the answer's no. There's a lot of founders where the answer's no. Hey, there's other parts or other things you want to do. But getting clear on the answer to that question I think is a really good one.

How do you know whenever you start to see a blip that it's just a blip versus you're going in the wrong direction and know that you need to do something to correct it? And do that as early as possible before it's too late. And then the second part is whenever you do make those kind of things and you end up at a tough time, how do you reverse the momentum in terms of team, how they feel, the motivation, as well as being able to make sure you made the right decision to come out of it? You've probably been through many of these. Where do patterns—

Yeah. I mean, the first is entropy is an incredible force. By default, you're not going to be successful. And there's hundreds of reasons why, like product isn't good or you aren't distributing customers in the right way, or there's some critical bugs or you don't have the right talent or whatever. Breakage is actually the default. These things being successful is actually very rare. And so you're always fighting against the force of entropy to make stuff work and be good. And so in some ways, I'd actually flip it, which is the inverse. It's like, This idea that you have this kind of clear end destination and it's kind of a linear path too, and you just stick to it. It's actually the opposite. It's actually stuff is chaotically kind of throwing you off course all the time and breaking down what you do, and you're actually actively having to fight against that entropy. And the great companies, I think, have that mentality where they are always looking to reinvent and rebuild what they do and who they are, and they're not sitting in place. And you can see so many examples of extraordinarily successful companies that did that at one point, but then ended up stopping on that innovation, resting on their laurels and didn't change with the times or with technology or with other things that were happening and ended up being unsuccessful as a result. So I think it's important to have a long-term vision for where you're going and then you'll update that in response to to specific facts. And so yeah, actually in my mind, the right way is to flip it and to realize entropy is actually like one of the lessons that I think I've learned is if there's part of the organization that you haven't looked at, chances are it's probably bad. There's a saying, expect what you inspect. And so if you're not inspecting something, it's probably bad. And that's happened countless times to me at Amplitude where it's like, no, the default state is actually not one of goodness. The default state of one of breakage and badness and all sorts of issues. And so the people that can recognize that as leaders and are not spooked by it, but actually incorporate what they're learning from it and adjust, they will do so much better than everyone else. And I think that's what the great teams, the founders, and that's what the great teams out there out there are doing. And then that actually, you want people who embrace that reality, the people who are there for chasing the good story and the momentum, they're going to wash out real quick because as soon as they encounter that entropy, it's not going to go well. And so you actually want to select on stuff that's where people have that resilience and that long-term view of of what it is that you're doing.

Last question, it's 2026, big year ahead of you. Are there any people, ideas, opportunities that you're looking to attract or find this year?

Oh, so much. I mean, I think analytics and data is going to go through the same thing in the next year or two that coding went through in the last two, where it's like coding today looks extraordinarily different. I mean, two years ago it was like Cursor doing autocomplete and it was like, all right, you might get 10, 30%, 20, 30% more productive. Today it's like 5, 10x, maybe even more for some of the best engineers. And so I think data and analytics is going to go through the same thing. And so for us it's like, okay, we're trying to build that future and we're trying to work with other people who are doing it. So we've been acquiring a lot of companies that are working on similar problems and we are looking to work with people who are really excited about solving that.

Spenser, thank you so much. This is awesome.

Dara, thank you so much. Really appreciate it. Thank you guys.

Thanks for tuning in. If you enjoyed today's episode and want personalized advice from Spenser Skates, head to the link in the comments or description to ask his digital mind on Delphi.