The Exchanges

Q the next lesson, which is how do you know whether it's a sustainable fundamental shift that you're fighting against or a unsustainable one? Uh, point being that a lot of, like, consumer behaviors were changed in COVID that have now come back and actually we've gone back to original behavior patterns. How much data do you need to know whether something is ongoing and continuous versus actually time-bounded and temporary?

A It's a great question. And I think COVID is the perfect example because you, you knew, or you couldn't be certain, but you're quite certain that this would actually end. The question was, was this a, a fundamental shift in behavior or a temporary shift? And we had exactly this discussion when COVID started. It was much easier with something like the shift to mobile. There was no real model where you saw people going back to desktop computers and other stuff. Macro shift is, you know, from curation to recommendation. Or as Mike Mignano puts it from social media to recommendation media. That's not going to reverse. So those are quite easy, but COVID was hard. So we literally had the discussion. Should we model this as a temporary anomaly? Keep investing, for example, in the car, even though we saw car usage dip completely during COVID, or should we just say no one is ever going to drive again? Let's reposition for, for the home use case. We chose the former. We said that we're going to model this as an anomaly, and we kept investing in The world going back to normal, and I can tell you that that was incredibly painful, because we thought it would be over in six months, then in 12 months, and like 18 months later, it didn't look like it was going to go back. It's actually only now, in the last few months, that we clearly see it going back to the normal behaviors. You know, home us…

AI assessment note: “We said that we're going to model this as an anomaly”

Q Okay. So, okay. I get you. I'm a startup founder listening. How do I set up these debates for success? Do you, you know, head of product and all around OG in life say, Harry, you know, this is the topic, lead the discussion for an hour. Is there multiple discussion topics? How do you like to structure them to get the best outcomes?

A So I think if I were a startup founder again, uh, first of all, I would obviously try to hire people that are very smart. So I would debate a lot of my team and that's what I've done in the, in the two startups that I've founded. And that has been very fruitful. We managed to discuss our way to insights that, that were actually quite novel, I think. That saves us a lot of time and manage to create good outcomes just through talking through, you know, what if, what if, what if with smart people. The other trick if you're a smaller company is you use other people. You call Harry and you say, like, I want to debate with you. What do you think of this? And I think, you know, you can time box it. I actually have pretty long discussions, like one or two hours for a topic to really empty it out. My personal experience is, you know, one to two hours may save you Months of building the wrong thing. So I, I think it's worth it.

AI assessment note: “I actually have pretty long discussions, like one or two hours for a topic”

Q Second thing, whether it's sleeping in the office with your team, whether it's the late nights running to get the pizza, whether it's the going back to school to learn ML, your family respectfully are the ones at home without you. How do you communicate the value of what you're doing to them, where they see the importance of why dad's not there?

A Uh, I ask them to listen to the podcasts. No, it is a great question. So It's, it's a, and this is something that I think many people, I've heard many people talk about this. I've certainly felt it myself. You're so engulfed by your work and you get a lot of feed, positive feedback from work as long as, at least as long as the company's going well, but then at home, and you know, you try yourself to be, try your best to be a good leader and inspiring and so forth. And then at home, you're actually not inspiring at all. You kind of suck. And so, uh, my My wife, who, you know, she, she is the one who said this, like, I need to come and see you at work every now and then. So I understand what it is you're actually good at because at home, I don't see it. Like, and I think she has a really good point there in like, you need to be impressed by each other. Right. You know, and, and remember that because when you see yourself day to day, you mostly see the, the not so impressive sides of, of, of, of each other. So we, we try to do that. You know, I try to make sure I see her doing the stuff that she's really good at. And, and she's made sure that she sees and cares about some of the stuff I do. And then you find this like, You know, uh, love and respect for each other, and you re-find it. That's the best way I've found. I try to do the same with my kids every now and then, bring them …

AI assessment note: “she is the one who said this, like, I need to come and see you at work”

Q What does retooling the company mean, Gustav? Is it, again, this is kind of my question.

A It's what I spoke about. We literally had, have had, we have a bets board where we kind of stack crank what we're doing. We literally had a bet called AI is the product. For like two years to really emphasize the shift, getting people in that mindset, um, educating them to understand how this works to start to use these things and get empathy, not just for users, but for the models, because people first underestimate the technology and then they're excited and they're overestimate what it can do. So they, they create products that are not mature yet. And so there's a lot of work in getting people realistic about where the, where these things are right now. That, that was probably the hardest thing to get everyone sort of on the same level of knowledge and understanding.

AI assessment note: “educating them to understand how this works to start to use these things”

Q elegant. Um, uh, the final element that I do have to touch on before a quick fire. I spoke to Mignano before the show, and he said, There are many, like, near-death experiences that are kind of wonderfully entertaining slash, you know, uh, you know, uh, educating to hear. Uh, if you were to choose your kind of favorite near-death experience to tell, what, what comes to mind for you?

A Well, like, uh, Mike said, there are actually a few. Um, I've, I've seen Spotify have negative growth, actually losing users and contracting at least twice. And I, I, you know, I speak to people in the industry. It's, it's very rare. People have seen that, but the company never survived to talk about it. It's very rare to actually see it and survive and do it twice. But I think the favorite, um, near death experience story that I don't think is widely known is quite early on when, you know, we, we have a really You know, we, we talk about the, the labels and negotiation with the labels as sort of being frustrating, but, but it's not in, in a sense because it is their job to represent these artists and try to maximize the value of their IP, right? And I actually think they're doing a good job of it. It can be frustrating as a product person, but we've developed this really good relationship with the music industry and, and these labels, but early on, It was different. You know, I don't know that they actually believed that Spotify would work. Maybe they, uh, felt that, you know, this, you could get access to cheap VC money here, expecting us to sort of, you know, belly flop over. And, uh, so it was more tenuous relationship, and they certainly didn't understand the model, right? And so they were very, very scared of the free tier, the desktop free tier. And their notion was that…

AI assessment note: “the favorite, um, near death experience story that I don't think is widely known”

Q I'm in London. It doesn't work that way either. But no, I get you totally. I'm sorry to interrupt. Can I ask that? Is there a time when actually You sat around the table, and you went, oh shit, we should have copied them. It could be YouTube, it could be any of the other competitors. We're actually doing the different didn't work out.

A Well, I would say this. There are certainly tons of things. It should have been faster. I think lyrics is an example, uh, of that, which, which we now have is very engaging. That was obvious. So we've certainly been super slow for various reasons on these things. In that case, it wasn't really because we had a contrarian view. It was actually, which is often the case because of how the music industry works. So if you work normally at Twitter or Google or Facebook or something, You're, you're as a product person, What you do is you try to figure out what the best possible thing is that you could build, and then you try to build that. What you do when you work in the music industry is you try to figure out what the best possible thing is. Then you go to the labels and you try to license the lowest common denominator of what is licensable. And that often means that you can't build the best possible product. You have to build with much, much narrow constraints. And often actually the cost is a constraint because the job of a music label is to take a piece of IP And try to sell it to you as many times as possible. You know, foreground rights, background rights, video rights, sync rights, all of these rights. So you can see that as, as very annoying as a product person, or you can do what I do and see it as constraints are what makes the world interesting. You know, can you do someth…

AI assessment note: “I think lyrics is an example, uh, of that, which, which we now have”

Q Gustav, final one for you. What are you most excited about when you look at AI in the coming years and what it can do for everyone, but also for Spotify? Just what excites you most?

A For me, it's kind of what, um, Tamer said. We, um, experienced the shift to mobile. It was very scary. Our business model had to change, could have disrupted the company. It was also by far the most exciting time I had. And to be honest, it was a little bit too stable for a few years there. And I feel like we're back to that kind of change. So, you know, I'm, I'm secretly hoping for that business model challenge and, and those things happening, that everything changes again. And I think I don't know if you agree, uh, Scott and Tomé, but like, if you're, if you're a designer, product person, those are the exciting times. So we're right there now. I don't think we're at the end of it. I think we're at the beginning of it. So I'm kind of back to like more excited than maybe in the last Seven, eight years in terms of just, it's, it's unclear now what's going to happen. We were on a little bit of an iteration track for a while there. I don't, it's not an iteration anymore at all. So, so that's what I'm excited about for the future person, very egoistically. It's just an exciting time.

AI assessment note: “it's unclear now what's going to happen. We were on a little bit of an iteration track”

Q the next lesson, which is how do you know whether it's a sustainable fundamental shift that you're fighting against or a unsustainable one? Uh, point being that a lot of, like, consumer behaviors were changed in COVID that have now come back and actually we've gone back to original behavior patterns. How much data do you need to know whether something is ongoing and continuous versus actually time-bounded and temporary?

A It's a great question. And I think COVID is the perfect example because you, you knew, or you couldn't be certain, but you're quite certain that this would actually end. The question was, was this a, a fundamental shift in behavior or a temporary shift? And we had exactly this discussion when COVID started. It was much easier with something like the shift to mobile. There was no real model where you saw people going back to desktop computers and other stuff. Macro shift is, you know, from curation to recommendation. Or as Mike Mignano puts it from social media to recommendation media. That's not going to reverse. So those are quite easy, but COVID was hard. So we literally had the discussion. Should we model this as a temporary anomaly? Keep investing, for example, in the car, even though we saw car usage dip completely during COVID, or should we just say no one is ever going to drive again? Let's reposition for, for the home use case. We chose the former. We said that we're going to model this as an anomaly, and we kept investing in The world going back to normal, and I can tell you that that was incredibly painful, because we thought it would be over in six months, then in 12 months, and like 18 months later, it didn't look like it was going to go back. It's actually only now, in the last few months, that we clearly see it going back to the normal behaviors. You know, home us…

AI assessment note: “We said that we're going to model this as an anomaly”

Q people or even middle-tier people will not feel the security to express their true opinions against Gustav. You say something and you say it with energy and charisma and conviction. It takes quite a lot if you're new or junior to say, Gustav, I totally disagree with you. How do you think about creating environments of safety where more junior people can have debate and disagreement in a productive way?

A Yeah, it's, it's a really good question, and I saw that shift pretty clearly when we started hiring more in the U.S. U.S. people were just used to hierarchy. Yeah. And, uh, I could find people, there were a lot of misunderstandings between Swedish people and, and, and, uh, Americans in, in many ways. Uh, you know, Swedish people are, Better at some things, but they're much worse at other things. You know, they do this, Swedes do this form of silent disagreement that Americans, uh, interpret as, as, um, actual agreement, but it's not, you're just silent and you disagree. So lots of cultural challenges there, but I think the way to do it is to model the behavior. I think back to, you know, what I liked about Daniel is he models that behavior. So in his team, you can debate quite fiercely. You can tell the CEO in a pretty heated voice that you disagree and you don't like that and you're not going to get fired. I think it starts there. And then I try to do the same thing. Uh, having, you know, sometimes heated debates with my direct reports, often in front of their direct reports. And sometimes it's a little bit shocking maybe for them. They're figuring like, oh, now my boss is going to get fired. And then If he or she isn't, and it's fine. And then they learn that, okay, it's apparently okay to disagree. That doesn't mean to your point of efficiency that you don't need to make dec…

AI assessment note: “the way to do it is to model the behavior”

Q Do you think there is a business model innovation? I cannot believe I'm saying this with Web three and crypto.

A It's a tricky question. So I, I love, um, I love crypto from that technology angle. You know, I'm a sucker like many people for this. This is like a true, You know, true computer science innovation, and I'm very excited about the, um, possibilities of, of, you know, trustless cooperation and so forth, but like everyone else in the very near term, maybe a bit disillusioned with the fact that while the promise is there for For sort of trustless cooperation. Actually, one of the biggest problems seems to be how you trust other people on this network. Uh, and you know, you sort of go back to the web to world to actually verify who someone is. So I think it's early. I am. I am. I am fundamentally excited about it from a technology point of view, but I don't yet have a, have an opinion on, on, um, If there is a big boss business model there, I have a feeling that just like AI and other things, it will take quite a long while, but it's going to be, there is a, there is a chance that it becomes a fundamental part of the infrastructure just further down that it actually powers a lot of things and is very valuable, but it might not actually be on the consumer end. That might be one of the mistakes that consumers don't care so much about cryptography, but if it can solve fundamental problems, like, you know, on the TCP IP level, it could be incredibly valuable.

AI assessment note: “I don't yet have a, have an opinion on... If there is a big boss business model”

Q Gustav, final one for you. Actually, penultimate one. Penultimate one. What's your favorite memory of working with Shaq, the many years with Shaq? I have many, but I'm intrigued.

A So what I think is special about Shaq and why he is such a core part of Spotify's history is that, as I said, Spotify is from one view, a tech company kind of grew up in this era of leveraging new technology. And as I said, that's, that's, you know, I think that is our core, trying to always understand new technology, but it is also right in between this world of, you know, nerds and hardcore tech like myself. And rock and roll. And, and that's always been really interesting at, at Spotify when you're sitting and you're writing code and then, you know, Kanye West comes and knocks on your shoulder and, you know, there are these surreal moments and, and Shaq happens to be responsible for, for most of them. For some reason, he has this incredible ability of finding people, getting people to like, he's just, I haven't met a person that doesn't Like Shaq. I'm sure they exist, but I haven't met them. And I think, um, he is the embodiment for me of marrying the technology world with, you know, Hollywood and, and the music world. And, um, it's very hard to To find that kind of person in, in a company like this and to get that person to be able to work in a company like this. As you can imagine, Shaq is not a big friend of, of like product processes and reviews and stuff, but Daniel has managed to, to, to, To make this work and have Shaq run these special operations for Spotify througho…

AI assessment note: “what I think is special about Shaq and why he is such a core part”

Q Final, final one on this, I promise. Does Zoom and COVID change the way that you do these sessions and these, like, you know, debates? And, and do you feel that there is a change in the quality of the debates in a virtual versus a in-person world?

A It's interesting because I think I've been trying to figure this out. I actually think some of them have gotten better because the notion of seeing everyone on screen and having to sort of mute, it's given a bit more structure because you have the obvious problem that some people would dominate a discussion when it's live. And, uh, I have actually seen less of that, uh, problem over zoom. The other thing I've done, which is interesting during COVID is, um, I've simply put my AirPods in and then had like, you know, remote walking talks with individuals or with groups of people where you walk for a few hours and sort of group talk. That's actually been very, very fruitful. So I thought when COVID happened that what would really suffer was strategy discussions and That kind of ideation. But for some reason it hasn't. So I, I would say one of the things I, I also tell my teams that I think is quite different from what many other companies do is I find that a lot of, um, executives, they look at their org and they try to split it as much as possible. You know, they try to divide and conquer and, uh, desynchronize and find swim lanes and divide problems. They, they try to To sort of cut up the org in as many independent units as possible for the purposes of sort of parallelization. And that makes sense from a point of view, but we've actually chosen back to sort of trying to be contr…

AI assessment note: “I actually think some of them have gotten better because the notion of seeing everyone”

Q How radical do you think you get with UIs? Everyone here wants to be at the forefront of design, of, kind of, interfaces with consumers. How radical can one be in how far you push it quickly?

A I think I want to build on something Scott said there, because I think it's the definition of what is UI changes. So you took the example of the, of the AI DJ we did, and you're exactly right. It was the design team that user tested. What is this? Is this a utility like Google? Is there an AI person like Alexa or, or do we actually, uh, is the UI that we digitize the real person with the real personality that exists? And we chose the latter. So that is, that is design and, and it is brand. So I think designers need to need to think holistically about the experience in that sense. And that is still design. It is the user experience. And I also think something that, that designers, um, need to get good at in this world is to understand the, the capabilities they have. They need to understand basically GPT-IV as well as they understand a user, what, what, what, uh, what it needs. And how it works. So an example that I think is a good example is Mid Journey. So Mid Journey, the designers there clearly understood the performance of the model when they built the Mid Journey experience, right? It took like two minutes to generate an image. It was wrong one out of four times. They could have built a horrible experience. We waited for two minutes and got disappointed 75% of the time. Instead, because the designers there and product people understood the capabilities, they built a fault-…

AI assessment note: “I think it's the definition of what is UI changes.”

Q in a box, number one, and two, you have to earn the right to do the next thing, but what's got you to where you are won't necessarily be what gets you to the next thing, and actually you have to change and adapt, and it's one of the biggest transformations that I've had, again, thanks to Shaq. Which product leader outside of Spotify do you most respect and admire?

A So there, I thought about this a bit. Cause I get that ask that. I get that question asked quite a lot. I admire a ton of people for, for different things, but I would say, you know, it's always because people answer mostly the same things. And so I also admire the same people that everyone else does, but to say something maybe unexpected, there is a product that I I've specifically fallen in love with over the last Uh, I actually don't know the name of the product leader, but I've been trying to learn. Yeah. I try to learn new stuff every now and then. Um, and, uh, after machine learning, I took a break and then now I'm trying to learn a few new things. And one of them is, uh, playing the piano. It feels like when you're, when you're working at a music company, you should probably be able to play the piano. So I've been using this application called, uh, simply piano. Which you hook up to your, to your piano over MIDI. And I really like it because it is, uh, so for two reasons. First is really good application. You know, what you want to do when you start to play piano is you don't want to, you don't want to start with these child songs. You want to start right away with the rock songs. And so what they've done is they've painstakingly, and I have a lot of empathy for this because I work in the music is they've painstakingly, painstakingly licensed like all the popular music. …

AI assessment note: “I actually don't know the name of the product leader, but”