Pod Insights The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch 20Product: How Linkedin Does Product Reviews, A Post-Mortem on Stories, Linkedin Messenger and Spam & Why the Data Advantage in AI is Diminishing with Tomer Cohen, CPO @ Linkedin About this episode Tomer Cohen is the CPO @ Linkedin. Since joining in 2012, Tomer has served in key leadership roles, helping launch and scale new innovative member and customer experiences. He previously led the growth and development of LinkedIn’s Marketing Solutions portfolio and LinkedIn's consumer and mobile products. Prior to LinkedIn, Tomer worked as an entrepreneur with Greylock Partners and founded a company in the personal CRM space. In Today's Episode with Tomer Cohen We Discuss: 1.) From Israeli Military and Chip Design to CPO @ Linkedin: How did Tomer make his way from the Israeli military to being CPO @ Linkedin? What does Tomer know now that he wishes he had known when he became CPO? What have been some of his biggest lessons from working with Reid Hoffman? 2.) Product: Art or Science: How does Tomer determine whether product is art or science? If he were to put a number on it, what would it be? How does Tomer determine whether to go with his gut vs go with the data on product decisions? How is AI changing the role of product managers and product leaders? What do product leaders and PMs need to do to stay up to date with the latest changes in AI? 3.) Linkedin: Review of Current Products: Feed, Stories, Messenger How does Tomer analyse the success of "the feed" in Linkedin? What worked? What did not work? Why did "Stories" not work in Linkedin? What went wrong? What did they learn? What is Tomer doing to tackle the spam issue in Linkedin? What are the biggest challenges associated? Why does Linkedin still have such poor messaging service? Why is it a difficult problem to solve for? 4.) AI Changes Everything: Why does Tomer believe this wave of AI is the most significant technological shift in our lifetime? Who will win the race in AI; startups or incumbents? Which model will work most efficiently; open or closed? Will we see large enterprises prefer bundled AI options or unbundled with specialised providers?
Transcript Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
0:00 Speaker 00 If you ask me a couple of years ago, I would have told you that data is everything in AI. What's happening right now is technologies like GPT are already trained on all public data. The whole idea of the whole pre-trained is one of the biggest inflection points in this technology. It's already trained on every available public information out there. The data advantage that used to exist, I think, is getting diminished.
0:23 Harry Stebbings? This is 20 Product with me, Harry Stabbings. Now, 20 Product is the show where we sit down with the best CPOs in the world to reveal how they build products and product teams. Now, today, I'm thrilled to be joined by Toma Cohen, CPO at LinkedIn. Since joining in 2012, Toma has served in several leadership roles, helping launch and scale new innovative member and customer experiences. Prior to LinkedIn, Toma worked as an entrepreneur with Greylock Partners and founded a company in the personal CRM space. But before we dive into the show today, this episode is brought to you by Linear. Let's be honest, the issue tracker you are using today, it's not very helpful. Well, Linear is different. It's incredibly fast, beautifully designed, and it comes with powerful workflows that streamline your entire product development process.
1:06 Harry Stebbings? From issue tracking all the way to managing product roadmaps, Linear is designed for the way modern software teams work. Its powerful workflows let you customize and automate processes to fit your team's unique needs. And they have integrations with your favorite tools like GitHub, Slack, and it takes just minutes to set up. Linear is the default tool of choice for high-performance teams, from startups to a wide range of large established companies like Vessel, Retool and Cash App. So see for yourself what makes Linear magical. Visit linear.app/20vc to try for free with your team and get 25% off when you upgrade. That's linear.app/20vc. And speaking of tools we cannot live without,
1:47 Harry Stebbings? I need your input on the 20VC Miro board about new guests we should feature for 20VC in 2023. Just head on over to miro.com/20vc. Miro is a tool I consider to be truly game-changing. It's a visual collaboration tool packed with the right tools, tech and templates to help you think of and create that dream product. That means you can brainstorm the perfect Product with your team, vote on the best ones and explore with your customer journey roadmap all on a Miro board. Whatever you need, Miro's infinite whiteboarding capabilities help you get there. It acts as your team's single source of truth, and now I'm using it to hear from you. Go ahead and add your suggestions for 20 VC guests on our Miro board at miro.com/20vc.
2:32 Harry Stebbings? And finally, did you know that 90% of information processed by the brain is visual? And with Canva, it's never been easier. Canva empowers teams to design impactful documents, websites, videos and presentations that will wow clients and colleagues with no designing experience needed. Teams of all sizes can design together from anywhere on any device. And you can start designing today for free. What will you design today at canva.com? You don't need any designing experience. There are hundreds of thousands of free templates and a huge content library. And that's why the largest teams trust Canva. It's used by 85% of Fortune 500 companies. Whether you're a team of two or 2,000, Canva empowers teams everywhere to create their best work together. What will you design today?
3:16 Harry Stebbings? Start designing for free at canva.com.
3:24 Speaker 02 You have now arrived at your destination.
3:27 Harry Stebbings? Toma, I am so excited for this. I've spoken to just so many of your friends before, so I know an incredible amount about you. And I'm ready, so thank you so much for joining me today.
3:35 Speaker 00 I'm excited to be here. I'm a big fan of the show and I'm excited to be on it.
3:39 Harry Stebbings? That is very, very kind, but I'm fascinated. CPO LinkedIn, one of my favorite products. How did you come to be a CPO at LinkedIn, Toma?
3:47 Speaker 00 Ever since I was a kid, I loved building. I love every aspect of building. I love the entire cycle, I love the problem solving side, I love the design side, technology side. My journey to LinkedIn was actually a very special one. So I became a LinkedIn fan long before I joined the company. I came to the Valley in 2008. I went to a lecture at a Stanford Engineering School. It was about social networks. This is 2008, social networks are a big deal. They're not as big as they are today. They're like the hot topic. All the big founders of social networks that you can think about were there. And it was all the rage. The hottest topic was Facebook time spent on the Internet. On stage, there was more of an older founder. His name was Reid Hoffman. This is where I got to meet Reid for the first time. Reid talked about the power of online professional communities
4:31 Speaker 00 and how it can create economic opportunities. It's the first time I heard it and it deeply resonated with me. The idea that a professional community becomes a powerful growth engine for the economy just inspired me on a whole new level. And over time, Reid himself became a personal mentor of mine. And it was only several years later, I had a conversation with who back then was the CPO of LinkedIn. And this was early mobile days and I came from a startup. And if you came from a startup in the valley, it was all about mobile. And larger companies were still trying to deal with this mobile thing that was happening. And he asked me, he said, how would you rebuild LinkedIn as a mobile product? I was excited to share my thoughts. And then he said, instead of talking about it, how about you come and build it? And the rest was history. I joined the company in 2012 and in 2020, I became the CPO myself.
5:16 Speaker 00 So it came full circle since then.
5:18 Harry Stebbings? I do have to ask you, and I ask quite unfair questions because they're generic. But I'm fascinated on this one always, which is like, when we think about products today, especially given your experience working kind of within semiconductors, within consumer, and then also like incumbent LinkedIn style, do you think product is more an art or a science? And if you were to put numbers on it, attach it to them, where would you put it, Toma?
5:38 Speaker 00 I heard you ask this question on a previous podcast I listened to. And I must say, credit some tension for me because I think it's impossible to delineate science from art. I think they're driven, they play off each other. There's a lot of science in art and there's a lot of art in science. Some of the best scientists in the world are very imaginative. After I thought about it more, I think when we talk about science in product, I think there's a tendency to think about it more as the best practices of product, the skill set. How do you learn the know-how around design and data and experimentation and business? And that for me is a foundation of what it means to be good at this job. And honing that craft, the scientific part of it in quotes, can take you a long way. And it has to be applied learning, right? You can't sit in a classroom, learning how to do product.
6:21 Speaker 00 You have to build. But I think what sets you apart in the craft of building product is your ability to bring vision and creativity and intuition and the judgment. The imagination. I think that's what sets you apart from the group. It's where you listen to customers or users or you just observe them. And some would just go and say, oh, this is what they want. And some will come, I think that's what they need. It's that uncovering profound needs that they haven't even talked about. It's being able to anticipate where technology is going, where the industry is growing, because you see a few steps ahead. And that takes a lot of intuition and imagination. And I think the further you grow in your career, there is an expectation that's what you'll bring to the company. The best product people that I enjoy working with have this combination of knowledge and creativity.
7:05 Speaker 00 They can do vision, they can go all the way to execution. And when I think about the intersection of product people who are dreamers, who are doers, who are learners, you really get the holy grail of product craft and ability.
7:17 Harry Stebbings? Toma, we chatted before about content creation on the platform and how much I loved it. I think the shows are also successful because I'm also quite honest, more so now than I have been before. Can I ask a blunt question? Which is when you look at the desktop site and you look at the mobile app, do you not feel that it feels outdated from a UI perspective?
7:36 Speaker 00 I think there's a lot that we can do there to improve. And there's actually we're working on right now to really innovate within the constraints of what we've done. In fact, the use case of LinkedIn dramatically evolved over the last few years. I do feel like in many ways, the current experience constraints us. And there is a lot we can do to improve it and to rethink it. That's what's happening right now. In a world of generative AI right now, there's a lot of complexity you can potentially unwind by building a simpler experience, what playing AI into it. If you play the last years for us at LinkedIn as a company, there's been tremendous evolution in the product that led to some pretty remarkable results. Revenue for LinkedIn more than tripled, our memo base more than doubled, our engagement is reading record levels.
8:20 Speaker 00 We're growing the fastest we ever grew right now.
8:22 Harry Stebbings? Can I ask an interesting one? I love the comment you said, what is something when you think about products, you've been wrong on but not confused on? And what did you learn from that?
8:31 Speaker 00 Yeah, so taking a step back on being wrong but not confused is my notion is if you're confused, then only luck will save you. It's like there is. Very little chances you'll be successful. But you have conviction, rally around that conviction. Yes, you might not be successful, you might be wrong. But ultimately, you have clarity and focus. And that for me are critical for execution. So it starts with clarity of thought about the problem you're trying to solve and goes all the way to the clarity on the solution with some strong principles. I'll give you an example of one that we were not confused about, but we were wrong. We launched Stories a while back, and it was a short test in many ways. But we launched Stories because we thought that we can basically unlock more creation on LinkedIn, allowing for ephemerality to play a role.
9:16 Speaker 00 We would sometimes hear from people that they think that it's sometimes risky to share on LinkedIn because their boss is on LinkedIn, their colleagues are on LinkedIn, their customers are on LinkedIn. We thought ephemerality might alleviate that concern. And we launched it and we saw some pretty, I would say, mediocre performance. It was far from the unleashing of sharing that the team thought they would get. When we did follow-up sessions with members and we were clear about what we were trying to achieve, we thought ephemerality specifically would unleash more sharing because it would alleviate the concern around sharing on LinkedIn. When we did follow-up sessions with members after that, it was clear that we completely misunderstood the job to be done for creators on LinkedIn. It wasn't about wanting things to disappear. It was almost like the opposite. People wanted things to last on LinkedIn.
10:00 Speaker 00 When people share on LinkedIn, they want it to be seen. In fact, they want to attach it to their professional identity. They want to feature it on their profile. Honestly, it was a really big, groundbreaking insight for us because we started doubling down on the role of creation as part of your identity. We also learned that when they share on LinkedIn, yes, everybody enjoys their likes and views. It's part of the chemical response you get in your brain. But what they appreciate most about LinkedIn versus other platforms, it's the reputational part. It's the opportunity, it's being able to showcase my values. It's very different than any other social platform.
10:33 Harry Stebbings? How important is being the first? I had Alec Schulz on from Meta, CEM at Meta. And I asked him this question, but how important is being the first? You mentioned Stories. Facebook obviously followed Snap and Snap followed Kakao. There's always who did first, but how important is being first in product release?
10:50 Speaker 00 It's first to launch. I think first to launch is not the right concept. I think there is a halo effect of who is launching first, right? That's sometimes and as you would see product managers getting really excited about the launches that they've made. But for me, that's not interesting, the launches. I think it's first to product market fit that is amazing. If you're first to product market fit, you build an amazing leg up in terms of both insights and momentum and speed.
11:13 Harry Stebbings? When we look at stories, we've got Snap product market fit first. And like LinkedIn were very far behind in like race to product market fit on stories. Do you think that was a core component as well?
11:25 Speaker 00 I don't think it's the same dimension because we were not trying to compete with Snap. I don't think it was the notion of being first to the market that made it special.
11:32 Harry Stebbings? You mentioned like the data not being amazingly exciting. And I just wanted to ask, how do you know how much data is enough to make a decision? Is it two weeks of data where you're like, okay, we see the writing on the wall? How do you know post-product launch when enough data tells you the answer versus when you need more?
11:51 Speaker 00 Yeah, I always like to start with before you launch, what is success? Once you launch it, what should be going up into the right that should be excited about? Don't tell me in retrospect because then you're just trying to fit the data into what you're trying to do, but tell me ahead of time. And it's okay to, again, you might be wrong, but not confused, just have a strong opinion around it. So we can see if we're building towards the intuition we have. I think for me, there isn't a notion of enough data. Is it the hypothesis you have isn't being validated? Like ultimately, you have success with a product when you have real adoption and retention. That's when you know you really have something. And adoption by itself, if you have a large enough base, could also be in a way a game in a while. Because discovery is really strong and you can always get people to try out something. But will they stick around? Will they come back to use it? That is your real test.
12:36 Speaker 00 We do a lot of processing around the notion of evidence versus conviction. You can build with strong evidence that you had before. Or you can build it with conviction with a hypothesis of showing evidence at a certain amount of time. And ideally, over time, you start showing evidence. But some of our biggest bets we've made, for example, investing in skills as a way to connect the talent marketplace, started with some big, bold bets around idea of conviction. And then over time, we started measuring, are we doing a good enough job there? Are we seeing enough validation for the early hypothesis to invest more and more? It's really a gradual incremental process that starts with conviction when you don't have enough data. But ideally, you come back with data to showcase the responses again.
13:18 Harry Stebbings? Speaking of that conviction and data, I do want your advice. Product reviews are so core to all product teams and to all companies. How do you do product reviews? How often do you do them? Who's invited?
13:29 Speaker 00 Yeah, so we do multiple product reviews every week. It starts each quarter. We actually review all of our big rocks, our big investment areas across the product and the business. And then we determine what do you want to see as a product team coming in? And for that, I decide what's the products I want to cover for review. And what I tell the team is you've put your best thinking forward. You know, you worked hard on the product, you worked hard on the thinking, you worked hard on the design. Now you're putting it to the team. And our goal as a team in this meeting is to make that thinking better, is to make that product better. So the only currency really is feedback. And what's nice about this team is you really have kind of this 360. You have the team presenting, they provide that local expertise, they're the one who are recommending. And then they're highlighting a problem.
14:12 Speaker 00 You have my team, which is a product executive team, they cover the whole business. It's multiple areas of the business coming together. It's a very broad. You have cross functional leaders who can provide other perspective. So by the time the meeting is done, they're getting some pretty diverse, deep feedback from multiple areas of the product. And that allows for that deep thinking to come to it.
14:34 Harry Stebbings? Do you see a difference in product review discussion quality and feedback quality? When comparing remote and online versus in person and whiteboarded?
14:45 Speaker 00 I do. I actually do my product jams in person. I feel there is an energy of creativity and discussion that you get in the room in person. It's you can do it remote as well. It's not that remote does not work, but I think there is an edge to doing it on a whiteboard, discussing, opening, pointing, having free form conversations. The pila for that creativity and discussion to be a lot more natural. So I move my. Product jam sessions to be in person. And with COVID, we moved them to be obviously remote because there was no other way. But once we came back from COVID, we moved it to be in person. And across the room, people will tell you the energy levels are just higher. The creativity, the ideation, the velocity of discussion is elevated to a whole new level.
15:29 Harry Stebbings? Who sets the agenda and who's invited?
15:32 Speaker 00 I set the agenda for the quarter in terms of what I would like to see as the product areas being covered. And we have a structure for the session itself. So you start with a problem definition, ideally as articulate and nuanced as possible. You share your job to be done. What is that insight that leads your work? And then you share your principles. If those are there, that's already a great session. And then ideally, we spend most of the time on the demo.
15:55 Harry Stebbings? What are your principles? So if the job to be done is lower amount of spam, say onsite, what are the principles? What would that be?
16:03 Speaker 00 Principles are basically your opinion and approach for how you would solve the problem. They ideally have teeth, they have trade offs. So if your goal is to alleviate the amount of spam or bad activity on the platform, at the cost of what? So for example, there is a trade off between trust and growth. Because you want to allow for a lot more growth in the system, you want to reduce friction. When you reduce friction, you also allow usually for bad activity to engage as well. And the more friction you have to alleviate bad activity, you also take down good activities. So there is kind of, there's usually a trade off between the notion of engagement and growth with trust. And when an idea for Lila, you find a great way to build an efficient frontier where you have high growth and high trust going together.
16:47 Speaker 00 For me, the principles is that the product leader for that area comes with an opinion approach of how they will solve it. And they're stiff to it, right? They're basically making an opinion of how they solved it. And what's the trade off they're going to make to actually solve it in a great way? Once you have that, you can actually react to something really well.
17:03 Harry Stebbings? When you have bluntly this structure and you have a lot of ideas thrown around, you also have to make decisions and you have to prioritize. When you come out of a product review, how do you prioritize what to do, what not to do? And what's a luxury but next quarter?
17:18 Speaker 00 So if we just finished a Great product jam, then what happens is there's tremendous amount of feedback being shared. Usually the way I end the meeting is I summarize the areas that we covered and where I would like to see progress on. I try to make sure it's between one to three, so it's not the whole list. And then the working team, we have a new process we started a while back, which is called Briefback. The team itself, which presented, they send a summary of the session with the feedback and all the key action items they have and the ETAs for the key areas they're going to work on. So it's really up to the team presenting to take that feedback and act on it. And we literally have those sessions that really allows for clarity and execution, but it's really up to the team leading it to go and do it.
18:00 Harry Stebbings? Toma, what's the most controversial product decision you've made?
18:03 Speaker 00 If I go back, the feed at LinkedIn now in hindsight doesn't seem controversial, but at the time it was very controversial. LinkedIn was actually one of the first social platforms to ever have a feed, but it was more of an activity feed at LinkedIn. The feed was really a promotional feed for teams to showcase their products. The growth team would show people you should connect with and the jobs team would show jobs you might be interested in. And everybody was using the feed as a promotional way to show members recommendations about that they can do across LinkedIn. The first change I made there was that the feed was first and foremost about people that matter to you, talking about things you care about. The feed belongs to the member. In some organizational chart, we start from the member's job to be done. And I can tell you internally, that was a controversial decision
18:46 Speaker 00 because the feed was already used by multiple teams as a way of discovery. That was a pretty big pivot internally and externally.
18:53 Harry Stebbings? It's an innovative dilemma question. Is it where you've got to cannibalize your existing product with a new product?
19:00 Speaker 00 A hundred percent. The feed was a massive discovery engine for so many products at LinkedIn. And it was a revenue engine for so many products at LinkedIn. For many companies, it was their main discovery channel. So then comes this not confused, but might be wrong product leader and he says, I'm going to change it. I'm going to change the core of how this works. There was high conviction there and I had to show evidence along the way, but that was a highly controversial decision early on.
19:23 Harry Stebbings? I think we should do an internal review, me and you here, where I say a feature or product and you can give me Rating out of 10 on how you've done, okay? So if we start on the feed, what would you give your rating out of 10?
19:35 Speaker 00 I think we still have a long way to go. I think we're far from building the product that we know we can.
19:41 Harry Stebbings? What I also think through when I think of that is shit, AI fundamentally changes how we build products. Do you as a product leader need to fundamentally change how you lead product teams and product organizations? With AI moving faster than ever, as we touched on when we chatted before.
19:58 Speaker 00 Yeah, this is something which is very new to my heart because I've been working on AI with AI for many years. And same as we had the mobile revolution happening 15 years ago and how we changed dramatically, how we build and how we use products, AI is going to be much larger. And if you're building product today, I have this analogy, when you have a rear rafting boat, you have the guide on the boat sitting on the back and that guide usually has two big pedals. And those pedals pretty much navigate the boat. It pretty much dictates success or failure for your product. Those pedals for me are AI and that guide better be you as the product leader. So if you lead products in your company, you'd have to have the knowledge and skill set to know how to use AI, how to use those pedals to navigate your team and your company towards success.
20:43 Harry Stebbings? What do you need to unlearn? What are the bad habits that you need to forget as we embrace it?
20:48 Speaker 00 Oh, I love that question. There's a lot of things when it comes to AI where you have to need to unlearn and relearn again. For me, one of the biggest ones, and this is where I usually see the hardest mindset shift for product people, is that with AI, you don't control the experience. AI is not deterministic. If you ever had an interaction in chat GPT, most of the time you're trying hard to anticipate how it would respond. And I give the analogy of a chef in a restaurant where before, imagine as the product leader, you were the chef of the restaurant, you dictated every part of the experience from the music, the ambience, the ingredients, how the plating is done, the flavors, the amount of salt. And now with AI, it's as if you're giving the ingredients, you're giving the guidelines, the philosophy, the principles, but AI really learns on itself.
21:32 Speaker 00 You can always refine and help it understand, but it learns on itself. So you have to forego, have that control, trusting you can Build a much better experience with AI and how you do it.
21:42 Harry Stebbings? Toma, do you and the team use Copilot?
21:45 Speaker 00 We do. I've been using this technology all the time. It's incredible.
21:48 Harry Stebbings? Yeah, I had dinner last night with two of our founders and they said they're now fifty percent more productive. And my question to you is, if we think about now GitHub published forty one percent of new code creators, AI code creation. What is that in five years?
22:02 Speaker 00 Wow, five years is a long time.
22:04 Harry Stebbings? Have we ever seen pace like this?
22:06 Speaker 00 Not in my lifetime. I can give you an example. I've been teaching how to build with AI first mindset for many years. But it surprised me how fast. In fact, when you think about what's coming, I think we're all going to be shocked by how more and more invested, more and more capable, because you are setting new baselines that allow for a lot more innovation. You know, you think about all the constraints of AI today, algorithms investment, computing power, allowing for more information to flow in and out in the form of tokens, those will all be alleviated. And when those are all alleviated, you can build faster and better and more accurate. You can allow for more creativity. That five year horizon is is almost like far too long, but I think we're going to see massive amount of productivity coming up.
22:47 Harry Stebbings? There's a brilliant quote. We've mentioned some great quotes. The big quote that I always think about is will the incumbent acquire innovation before the startup acquires distribution? Brilliant quote. The really interesting thing now is, I think incumbents are innovating so well with AI. You look at Microsoft, you look at Adobe and how Adobe are integrating into core products and product suites. And so my question is, where do you think value accrues in the next wave of AI? Do you think it is actually incumbents? Or do you think it's startups?
23:15 Speaker 00 So every tech revolution brings with it new winners and new losers. This one is no different. In fact, I think this one, the window for many startups to actually start playing with ingenuity and innovation. To your point, I'm sure many incumbents today have learned the mistakes of the past and they're racing to innovate with this technology. However, they do tend to be locked to their way of thinking, and I think that's an opportunity for startups. So let's start with the classic dimensions that usually incumbents have a leg up with. They have access to resources. This requires compute. This requires capital. This requires talent. So if you're an incumbent and you already have the infrastructure for AI, the talent For AI, you you already have a leg up in this race and you should be using that. And comments have thin market share, they have already an existing customer base.
23:58 Speaker 00 That customer is expecting them to start leading with AI. So they're potentially racing to that. Usually have a data advantage. They have some proprietary data that is unique to them, so they can build with that. But that's a big but, because I do think this is the biggest unique advantage of startups is startups can be extremely innovative right now. I think you take every problem that existed before, every job to be done, and you can go back to the drawing board and rethink it. Harry, this is not just in tech. This could be healthcare, retail, fashion, entertainment, supply chain, education. We can go industry by industry and think about the transformations that the technology can do. I think we're going to have a moment of being mind blown. You really go to the drawing board.
24:39 Harry Stebbings? Do you not agree though? So I was with a company this morning and they were like, we're going to sit on top of OpenAI and ChatGPT. And we're going to provide this additional slice and say it's in healthcare. So we're going to provide this additional. Doesn't work. The ones who actually work is when you've got five years of healthcare data that you can bluntly utilize the models on. Where it then is actually thirty percent OpenAI or forty percent, whatever it is, and sixty percent new. But this thin layer of I'm seeing so many sales tools, onboarding tools. I don't think value accrues onto that. And so I'm like, where does value accrue then in this wave that you mentioned?
25:12 Speaker 00 If you ask me a couple of years ago, I would have told you that data is everything in AI. Because computing power is accessible and the models are accessible to all, they're open source. So it's really about the data you have. And there was a people coined data as the new oil. And there was a reason for that. Data was the fuel that helped AI become better. And then the more data you had, the better products you can build, the better products you have, the more users you have, the more data you have. And you have this virtuous cycle of success. It was also vicious cycles for startups. So startups has to innovate around how do we get data? And there's ways to do that as well, by the way. Some very amazing scrappy startups were able to bring that in a very innovative way that helped them compete. Now, what's happening right now is technologies like GPT are already trained on all public data.
25:56 Speaker 00 They've read every book, every healthcare book, they've trained on. On every piece of literature, art, best practices. The whole idea of the whole pre-trained is one of the biggest inflection points of this technology. It's already trained on every available public information out there. So the data advantage that used to exist a couple of years ago, I think is getting diminished.
26:17 Harry Stebbings? How is it getting diminished? If you look at, say, TripActions or Nirvan, they have seven years of travel and expense management data, which is not a public resource, generally speaking, if it's high quality and rich in data sources. If it is, it's probably shit. So how is that not proprietary still? How is that pre-trained already? Like, they still have as just as strong as ever now.
26:40 Speaker 00 So like now, everybody has a similar baseline. If data before was the ultimate oil, it still has power, it's still important, but it no longer has the same power as it has before. Now, when you have proprietary information, you can start building your own unique application. So if it's healthcare and you have healthcare data as an example, you can build really unique specialized tools. So the example I give is GPT is a generalist. It's almost like you hire a generalist to your VC, they can do a lot of things. It's almost like you bring an athlete, but then you can start training that athlete to be a weightlifter. Athlete could be a long-distance runner. And then you specialize that athlete in a very, very, very specific with your own data, your own tools. You can build something unique. That process is called fine-tuning. That process is also expensive.
27:22 Speaker 00 So when you start fine-tuning, you have to build your own instance of this model, bring your own data, that requires computing power, and that by itself is a process. So there is an opportunity right now for startups to actually go and innovate.
27:34 Harry Stebbings? Do we need to restructure the fundamentals of our teams for these new processes? So when we think about moving to these instance creations, and when we think about fine-tuning our data on pre-trained models, but we're using our data, do we need scientists in our team? Do we need a completely different structure for how we build and develop products?
27:52 Speaker 00 I think you need new skill sets in your team, for sure. You need people who are working very closely with this technology. I think AI talent is going to be one of the most important talents to have in every company. So one of the most important interfaces is the prompts, right? What people use in ChatGPT as a fun way to write prompts. That's actually a really critical interface. That's the interface between the machine and the human. Being able to understand how you do, this is how you talk to AI. And proud by itself is a masterful skill today. The best practices playbook is still being written because in many ways, we're still almost like reverse engineering how this technology works. But the skill of understanding how to talk to AI, to your point, how to build my fine tuned data into the process,
28:33 Speaker 00 how to leverage it in the direction of building amazing processes is an extraordinary skill.
28:38 Harry Stebbings? I'm throwing you in the deep end here. I'm going to get you in trouble with comms, Tomer, but LinkedIn is a publishing engine as well. I publish to LinkedIn my posts. Many people publish amazing content to LinkedIn. OpenAI and ChatGPT and many of these models will be scraping, extracting your content and your value, paying you nothing. How do you think this looks for content publishers moving forwards?
29:02 Speaker 00 Yeah, I think there's the question around the role this technology is playing with publications in general. Like if now I can get an answer for this tool, do I need to go all the way to the site? And I think that evolution was still being developed and built. I think we have the same thing with search to an extent where if I can get the answer for search, do I have to go all the way to the site and actually get the response in the site?
29:22 Harry Stebbings? Well, no, but if you think of search as a reader, if I search Tomer Cohen LinkedIn or CPO LinkedIn, it takes me to your page on LinkedIn and then you retain me, you have the chance for me to follow someone else or follow someone else. Here I can go, who's the CPO at LinkedIn? What was their background before? LinkedIn never gets my visit. It's different.
29:41 Speaker 00 Yeah, and I think that will potentially start changing the interaction model. But this is where you start playing into at what level is every information of every company appearing in this model. There's some specific like lines around what is accessible and what is not. But you do expect this technology to start playing a much bigger role in interactions. So for example, when Bing launched their experience with Copilot, it gives a lot of annotations to where the information came from and reference back. So you can actually get the information back and it starts to highlight a lot of the value of those platforms. I think we're still at the early innings on it, so we can see how this develops. I do think it poses the question of where is the internet evolving towards in terms of information? Discovery and then information in a way, business models that are going to appear as part of this one.
30:25 Harry Stebbings? You said very early there in terms of where we are. I just have to ask you, Elon said, we cannot let it out of the bottle. Because unlike most technologies before, if you can put most of them back in the bottle, you won't be able to revert this progression. Do you agree with him? And how do you think regulation looks for AI moving forwards?
30:43 Speaker 00 I think when you think about this technology, your spectrum for me lies from high excitement to concern as well. Because it is with any powerful technology, it could be used to do amazing things and as an amazing productivity tool. It could also be used as a weapon. The notion of being extremely responsible with how this technology is being used, especially for the ones building it. I could not agree with more. We know we've been investing for a while now, even long before this technology. And what we call responsible AI principles, which is how you build with transparency, how you build with exclusivity, how you build with privacy. You make sure it's built into the models, it's built into the interaction. And you don't release something to the public unless those principles are being met.
31:26 Harry Stebbings? Should models be trained to be politically correct?
31:29 Speaker 00 Models should do their best to provide the most accurate answer to your flow. If you're trying to have a buddy experience, then that's a very different product you're building. But if you're trying to access information, the models should try to be as factual and objective as possible. I think that kind of puts aside the notion of being politically correct. Everybody will have access to this technology. Everybody has access to it right now, so they can build whatever product they're trying to do. But models are, in general, they're just construct of technology. Question should be, should the applications be done in a certain way? And that's really up to what are you trying to achieve? That's really up to them, but the models ideally are as accurate and focused on information as possible.
32:08 Harry Stebbings? What do you think are the big questions that people are not asking? You've been deep in AI for many years. You see the questions that I'm asking. You see the questions that other people are asking. What are you going, Harry, I don't know why no one's getting this?
32:21 Speaker 00 Those models right now are very focused on existing knowledge. So they learned all available public knowledge in the internet and they have. Incredible comprehension and generative skill. And they were able to produce a result for you that is trying to predict what you're trying to answer. But then there is a question of what about new knowledge? What happens when those models start to hypothesize? They can come up with new ideas, new scientific discoveries. Imagine AI coming up with answers to some of the biggest scientific mysteries in the world, like what is dark matter? What's dark energy? What causes Alzheimer's disease? What is quantum mechanics? What is oneself? And that for me is you're moving from a place of those models are amazing in rebuilding and restructuring existing knowledge
33:06 Speaker 00 to coming up with new knowledge. When you start to come up with new knowledge, you're really talking about a whole new frontier, not just for business and the economy, but for society as a whole.
33:15 Harry Stebbings? Isn't that new knowledge, not AGI? Is that not the ability to deal with ambiguity? The ability to make subjective decisions?
33:22 Speaker 00 It is, a way you're trying to build as close to human intelligence as possible. And human intelligence, going back to our conversation about art and science, like the ability to be imaginative, to hypothesize, to have a vision. That's what makes us naturally human. There's why we start asking, when this technology reaches that level, what is next? And then it's not just going back to the drawing board. I don't know if there's a drawing board, let's go back to rethinking, but what is the role of this technology? And how do we really start interacting? The human and machine interaction is elevated to a whole new level that I don't think we have experienced before. Now, here's the part, I think society has yet to adapt. If you think about self-driving cars, we're still playing with the idea of self-driving cars. What about the self-driving doctor? A self-driving psychologist?
34:06 Speaker 00 Imagine this technology being able to be so intimate with you. It could access your phone, your computer, your photo, obviously with your permissions. It could read your voicemail, every piece of digital footprint you have. You know, you and I can start building some great science fiction movies as a result of that. But is it science fiction? There's a great saying that I really like, that the difference between science fiction and non-fiction is just a matter of time. And I think we're entering that phase of science fiction being a reality.
34:32 Harry Stebbings? Toma, you have children. Do you feel a bit ridiculous sending them off to school every day? And I don't mean that rudely, but I just mean in like the decay rate of. Education. By the time they'll be like functioning adults in a workplace, what they're learning today is pretty useless, huh?
34:46 Speaker 00 I think the most important skill outside of being a good human being and a kind human being from a what do you learn to be successful? The only skill that matters, I think, is growth mindset. It's the ability for you to learn, to continuously evolve. You're already at a pace of accelerated technology. If you look at the skill set you needed for a job today versus five years ago, that has changed by twenty five percent. So a quarter of the skills that were required five years ago, they're different right now. You look at it five years ahead, you're going to be fifty percent different and we will know they're going to accelerate and accelerate over time. So then you go back to, is it learning a specific skill? Should you learn coding? What should you learn? And it's really, for me, it's you should learn how to learn. You should learn how to pick up new dimensions and new material
35:31 Speaker 00 as fast as possible and being able to immerse yourself with it. At least for me, in my household, growth mindset is like our second religion at home. It's the most important thing we invest in.
35:40 Harry Stebbings? Final one, I promise, then I'll do a quick fire. Is it a world of one model or is it a world of many models? Much more complexity, but tailorization.
35:48 Speaker 00 We are now in a phase of like foundational models. So the foundational sets the baseline for all other models to build on top of it. So you'll see a lot of applications and new models. There is now like AutoGPT that takes the GPT four and builds overarching objective function above it that kind of breaks it down to multiple subtasks. So I think we're now at the phase of foundational models and you'll see a lot more applications build as a result of that on top of it. And then gradually we'll see more and more foundational models in the future. You know, back to the notion around bundling and unbundling, I think we're now in the version of unbundling to multiple versions until we see another foundational model in the future.
36:24 Harry Stebbings? So I want to do a quick fire round and then I'm going to let you go. So I say a short statement, you give me your immediate thoughts. What is your favorite interview question when hiring for product?
36:34 Speaker 00 Two questions. One is, what's the most complex problem you worked on and how did you do it? And what you are trying to understand there is what was your job to be done inside? How clear you are around explaining a complex problem? How nuanced can you get? How much of a profound understanding did you bring to it? And I enjoy going I really enjoy going deep with with candidates. The other one is a growth mindset question. I'm looking for areas where you did not succeed, where you failed. Understanding how you dealt with that. How did you deal with failure? And I'm looking for learnings. The best responses I get is I've learned so much and here's how I would do it differently. It's not about retroactively trying to fix anything. It's more about seeing that there is an evolving mindset there of trying to do things better and evolving over time. Those are my two main questions I ask in every interview.
37:18 Harry Stebbings? What product release are you most proud of? And what are you most embarrassed of?
37:23 Speaker 00 I heard an artist recently talk about, a musician, talk about their career in albums. And I feel pretty cool to think about your product career in releases kind of thing. Releases that kind of mattered and made a significant impact. We talked about the feed before and how transformational it was. Transforming LinkedIn into a mobile company was an incredible moment. To rethink originally, LinkedIn was a desktop-first company. It had so many like desktop experience. I think we had hundreds and hundreds of pages on desktop. And here we are launching LinkedIn from scratch again with 20 pages. That was a big bold move and I really enjoyed that. With the feed, we talked about how we shifted towards more knowledge and economic opportunity in creation.
38:02 Speaker 00 I was really proud of how we dealt with COVID and how we rallied internally to meet the moment and help members where they need us the most. So many moments I'm proud of. And you know, looking back, even the ones that did not succeed, we learned so much. I've grown so much from those.
38:17 Harry Stebbings? Which one are you most embarrassed?
38:19 Speaker 00 I'm not embarrassed of anybody because I've learned a lot from, but there's many that I've done that did not succeed. You know, I've done one early on. I tried to launch what's called instant articles. This was early mobile days. Every article you clicked on LinkedIn took you five to seven seconds to load. It was a horrible experience. And I was like, I can solve this. I can cache all the information when you come in. I'll be the great example, a great experience to showcase it to users from publishers. We got publishers to work with us. It was a really tough one because ultimately we did, we were able to load it pretty fast. And it was a much simpler experience, but there were so many nuances to how publishers wanted to experience to show up. And we ended up thinking, oh, this is an enterprise tool and we're building. Consumer product have to simplify this dramatically.
39:02 Speaker 00 I've learned a lot about the different nuances of how you build different customer experiences for publishers. But we were way ahead of the market in that one. The timing was not there as well. The hypothesis was solid, but the execution was not.
39:15 Harry Stebbings? What's the biggest piece of advice you give to a new product leader joining an organization today? What do you wish you'd known the night before you became CPO of LinkedIn for the first time?
39:25 Speaker 00 I think when somebody joins a PM at LinkedIn and usually asks me like, what's my, what's the best way for you to be successful? The biggest piece of advice I tell them is if you just focus on your area, there is a chance that you'll be linearly successful and you can move your area forward and can hit your targets. Really, your ability to be exponentially successful at LinkedIn is really learning how LinkedIn works. LinkedIn is a beautifully complex ecosystem. We have a consumer platform that caters to 900 million members, 60 million companies on LinkedIn. Now, with that ability, how can you build something that is truly unique, is truly innovative? But don't just think about your swim lanes and what you're trying to do. Think more holistically about the experience that you can bring into it.
40:07 Speaker 00 Some of the best experiences on LinkedIn, they cross across the entire ecosystem, cross multiple products to build something which is really unique. And that's where I would love to see innovation come through, not how I innovate within my specific area or my feature.
40:20 Harry Stebbings? Penultimate one for you, my friend. I hear you're an intermittent fastener. What's the biggest advice to someone considering intermittent fasting?
40:28 Speaker 00 Give it a few weeks. It takes time. I think the beginning is hard. My most favorite meal is breakfast and giving it away was not easy. And then
40:35 Harry Stebbings? How long do you fast for?
40:37 Speaker 00 I do everything between 18:06 to 24, 20 hours fasting, 4 hours eating or 18 hours fasting, 6 hours eating.
40:45 Harry Stebbings? Do you not find it impacts your brain activity and your mood? For me, my brain activity goes down and my mood goes down.
40:51 Speaker 00 No. How, how long did you try it?
40:54 Harry Stebbings? Probably three days a week for a month. So not every day, but like Monday, Wednesday, Friday for a month.
41:00 Speaker 00 So when I started, a hundred percent and I run and I used to run long runs and I used to think about, I'll have to eat before the run, if after the run. I can fast for a whole day with A long run and not feel anything right now because your body learns to produce the sugar you need, the energy you need, the glucose you need in a different way. But I've found it to be extremely powerful for me. And in fact, I'm sharper when doing my fast than when I eat. So it's actually it elevated my physical abilities and my mental capabilities. But there is specific science around the intermittent fasting, the ability for it to both improve your health span and your longevity.
41:38 Harry Stebbings? Listen, I could talk to you all day. I didn't know you were a runner as well as fuck. We're in for a long session. My final one for you, though, is what's the most recent company product strategy that you've been most impressed by?
41:48 Speaker 00 So this is a very easy one, but I'm obviously very subjective. So you have to excuse me on this one. I've been very impressed with what Satya and Microsoft has done with the OpenAI integration. It's just hard to look at it and not be impressed by the strategy, the execution, the impact. I've been able to see it firsthand, be part of it firsthand, and I know what's about to come, and I'm extremely bullish about it. But being able to really think through a groundbreaking technology, think through the execution of taking it to market in a responsible way, thinking through the elements of all the way from cloud impact to productivity to end user. For me, that's one of the most impressive product strategies I've seen in a long time. Plus execution, been remarkable to see and be part of it.
42:30 Harry Stebbings? Elon said at this point, Satya and Microsoft basically control OpenAI. Is that fair?
42:35 Speaker 00 Microsoft is a partial owner in OpenAI, but OpenAI is an independent company.
42:40 Harry Stebbings? I've loved this, man. This has been so much fun.
42:44 Harry Stebbings? Thank you so much for joining me, Toma. You've been amazingly resilient to my questions. So you've been a star and I really appreciate it.
42:51 Speaker 00 Of course, this was awesome. Thank you so much.
42:55 Harry Stebbings? I absolutely love that discussion. Now, if you want to see the full video, you can on YouTube by searching for 20VC. But before we leave you today, this episode is brought to you by Linear. Let's be honest, the issue tracker you are using today, it's not very helpful. Well, Linear is different. It's incredibly fast, beautifully designed, and it comes with powerful workflows that streamline your entire product development process from issue tracking all the way to managing product roadmaps. Linear is Designed for the way modern software teams work. It's powerful workflows that you customize and automate processes to fit your team's unique needs. And they have integrations with your favorite tools like GitHub, Slack, and it takes just minutes to set up. Linear is the default tool of choice for high-performance teams.
43:37 Harry Stebbings? From startups to a wide range of large established companies like Vercel, Retool and Cash App. So see for yourself what makes Linear magical. Visit linear.app/20vc to try for free with your team and get 25% off when you upgrade. That's linear.app/20vc. And speaking of tools we cannot live without, I need your input on the 20VC Miro board about new guests we should feature for 20VC in 2023. Just head on over to miro.com/20vc. Miro is a tool I consider to be truly game-changing. It's a visual collaboration tool packed with the right tools, tech and templates to help you think of and create that dream product. That means you can brainstorm the perfect product with your team,
44:21 Harry Stebbings? vote on the best ones and explore with your customer journey roadmap all on a Miro board. Whatever you need, Miro's infinite whiteboarding capabilities help you get there. It acts as your team's single source of truth. And now I'm using it to hear from you. Go ahead and add your suggestions for 20VC guests on our Miro board at miro.com/20vc. And finally, did you know that 90% of information processed by the brain is visual? And with Canva, it's never been easier. Canva empowers teams to design impactful documents, websites, videos and presentations that will wow clients and colleagues with no designing experience needed. Teams of all sizes can design together from anywhere on any device. And you can start designing today for free.
45:04 Harry Stebbings? What will you design today at canva.com? You don't need any designing experience. There are hundreds of thousands of free templates and a huge content library. And that's why the largest teams trust Canva. It's used by 85% of Fortune 500 companies. Whether you're a team of two or 2,000, Canva empowers teams everywhere to create their best work together. What will you design today? Start designing for free at canva.com. As always, I so appreciate all your support and stay tuned for an incredible set of episodes next week.
Previous 20VC: Why Being First To Market Does Not Matter, Why You Do Not Have Defensibility on Day 1, How to Analyse Market Size and Present it to Investors, Vitamins vs Painkillers; Do Vitamins Survive Recessions and Good vs Great Messaging with Guy Podjarny @ Sn Next 20VC: Why Financial Models at Seed, $5M Seed Rounds & The Fear of Signalling Risk is all BS | Why Multi-Stage Firms Have Destroyed Seed & Who Wins and Who Loses in the Next 10 Years of Venture with Adam Besvinick, Founding Partner @ Looking Glass Capital Page updated October 26, 2026 · transcript and summary are machine-generated
