Where Top VCs Are Investing in AI — Felicis, Kleiner Perkins, NEA & DVC | AI Rabbit Hole 2025

All right, for the next panel, I'm privileged to be joined on stage by the most successful venture capitalists in the world who continuously score big in their early stage bets. I'm also very lucky to be on the cup table with some of the companies they backed and I know their founders giving them very highest marks. Please welcome on stage Peter Deng, GP of Felicis who also previously led Charity product and was CPO at Air Table and Uber. Please also welcome Ilya Fushman GP at Kleiner Perkins who invest in a very small number of deals but somehow they all score big. he was an early investor in Harvey AI, Codeium, Slack, people and many more. And please also welcome Ann Bordetsky. She is a partner at new enterprise associates who did deals like contrai and perplexity and before led corpdev at Twitter and Uber. All right, I will start my questions with Peter.

Sure. we already mentioned that Peter has amazing product experience. How this experience help you to shape your investment approach in AI companies and what do you say where you invest? Do you think that application lay is something that you're looking on and focusing on or JP anyway will eat the world? It's a good question. happy to be here. Thank you for having me. so yeah, I've I've worked with a few I've gotten lucky. I've worked at a few what I call rocket ships. it's been awesome because the analogy I use is that I've had the fortune of feeling a lot of the G-forces of the takeoff. So, I was at Facebook, I was at Instagram, Uber with an we, we worked together there, Air Table and ChachiBT. and it I think one of the things I learned is just like how awesome it is to work with different founders.

And I think that one of the things that I loved about all my jobs is I got to support these founders and they were all I think just kind of exceptional founders. Like being able to work with Markx closely on a bunch of projects was awesome and Kevin Instagram and and Howie and and and and Travis. I think that that thing that I learned most was just like how different amazing founders can be and getting a sense of the pattern recognition of like ooh that person's like so and so in so and in such and such way. So I think that's the one thing I I kind of learned. your question about chat eating in the world like I wouldn't be doing what I'm doing if I thought that that was the case. First of all I want to say attach is an amazing product.

I am not a daily active user. I am an hourly active user. it is it's incredible. The team is awesome. I'm biased. to help build the team there. but really there's so much value that is to be created and and the reason I think that you can just really believe that is that yes, there's, you know, with any company and every builder here knows this, you have a limited number of hours in the day and how do you focus and if you're on at OpenAI, you're going to focus on the most broadly applicable things because they have access to the best models as they're coming out. And that was the one of the magical things. But working there, you can really see how many opportunities are left on the table, right? Because these simplest ideas in in my mind, the simplest ideas and the founders that obsess over them, that's those are the people who are going to get the the businesses right.

And every I can talk for hours on this, but the companies I've been at, they're all simple ideas, but they really obsessed over the product details, and that's what really got it to really grow. So, I'll leave it at that. thanks for having me. Thank you. Thank you. I mean Ilia first of all can I congratulate you with exit with codium or you can't talk? I have no idea what you're talking about. Okay. Okay. Well, you've been investing in lots of company before AI wave such as Slack which is amazing company. How new generation of AI companies like Harvey and Codeium are different from them? Yeah, it's a great great question. I think that what's really incredible about the wave of AI companies is just the sheer momentum and scale that they're able to hit. and that's, you know, truly unprecedented relative to both their their team size and when they've been founded.

It took Slack about three years to go from the first million dollars of AR of Slack post their pivot to getting to 100 million. And, you know, we're seeing these companies eclipse that in a year, two, you know, two years, a year, maybe even in six months soon. So I think that's probably the most exciting part of investing today. and folks are able to do that with tiny teams which is you know thanks for Peter for building all these cool technologies that that power that. so I think you know and it it really changes fundamentally the the early stage venture landscape in many ways and what it means to identify really good early teams what it means to figure out which ones are going to win. So it's actually pretty fun to be in venture over a long window of time and see how the business has evolved.

Okay. Thank you. And at NA you did a lot of great deals. but I think you know that personally one of the my favorite one is perplexity. how did they stand out to you when they were raising series A so much that you close a deal like in one day in a week? Yeah, something like that. Let's say quickly quickly quickly quickly. no, it's a great question. I'm I'm definitely a DAO and probably how like perplexity user. and I think you guys heard from Arvin and Dennis earlier today. I mean, look, for us it was a combination of a couple things and I feel very fortunate that we get to work with this incredible technical team and follow their journey. One is, you know, obviously love consumer tech generational consumer companies. Those are extraordinarily rare. but oftentimes they begin with a moment of magic in the product, right?

And a really incredible like ambitious visionary team. And at the time we'd looked at other kind of earlier search attempts, search disruptors, great technical teams, but didn't have that product magic that I think is really required for a consumer first company. And the chat GPT moment had just happened, you know, a couple months before that. So it was obvious that LLMs were going to disrupt search and just knowledge work in general. So when we met the team and you know they had kind of like the exact constellation of backgrounds that you would want for that type of product AI research engineering design and it was obvious even in the early version of the product the MVP which they had only launched three months ago that there's like the the attention to detail and craftsmanship in the product that could be really special and that is just so rare and so I think when we saw you know the team the timing the ambitious vision and the quality of the initial product, it was kind of a no-brainer for us.

And obviously everything they've done since then is credit to the team and their hard work and their tenacity and innovation. and you know, I think the thing that really sets them apart is their product velocity. The amount of product that they're able to ship as a lean team every single quarter is just unbelievable and they've been able to maintain that even as they've grown the team. Nice. Thank you. Peter, I know that Felicis is focused on AI that augments the labor and I think we are now co-investors in some of the companies who are great in that. I'm not sure that we can mention that but but still what specific industries and kind of labor you think are ready to be automated and where we are not there yet? Yeah, I actually this is just a core thesis that basically AI is just going to be something that we co-eolve with as humans.

In fact, I actually gave a talk at South by Southwest with my friend Josh who was here earlier today on stage. with with some folks, but basically the idea is that you know when you have tools like AI, it's just it's not going to make humans lazier. That's not my thesis. It's going to make humans do higher level tasks. So my piffy way of saying this is like you know when the calculator was invented people didn't stop doing math they just did higher level math and you can send people to the moon and that's really amazing right so there's so many bits of knowledge work that this again I previously mentioned that and we talked about the product craft like when you have a founder who really obsess over making this workflow really sing that's where the product is going to really take off right and the workflow the ergonomics to use a product design analogy really do matter matter like how does the product afford it?

But also with AI, the other thing is that the data flywheel is super important, right? When you have companies that have a unique piece of data that maybe is not accessible that on the internet that all the you can really create a lot of really good value. that combined with the workflow and if you have that flywheel of the more that you use the product, the more of that reinforcing data happens, you start pulling away in terms of your advantage. so I think there's just like again kind of just to to go back to your question around what's ready for disruption or automation. You can take any you can take any workflow that's being done today and if you actually applied the right level of focus you would be able to really create agents I hate that phrase but call it what it will you can create agents that are able to really augment any work that's out there.

So basically any industries. Yeah, a lot a lot of knowledge work and I think a lot of it will be done by the chatbts and perplexities but there's plenty of different industries that you know I'm sure a lot of folks here are working in that have a specific knowledge and data sets and flywheels that just are not available anywhere else. Okay, got it. Ilia, I know that at Claner you adopted an approach that you called craft approach when you keep your investment team very small and you investing in small number of deals. Did it change with AI? maybe you use like tools that keep you like active and help you to reach more companies. Yeah. So we have a pretty small team for for a venture capital firm these days. we have two funds. Our venture fund is 825 million. Our growth fund is about a billion and a quarter.

we have six partners. three next generation younger folks on the investment team. and we actually do have a pretty amazing you know reasonably sized operating team that supports the growth of our companies whether it comes to sales, team building, engineering, hiring, you know marketing, things like that. But our our core belief is that you know venture is an art and a craft. The venture business undoubtedly has been evolving pretty dramatically. There's more capital in venture capital than there's ever been before. we're seeing, you know, even for us, you know, two plus billion dollar set of funds is pretty sizable by historical standards, but hey, it's not as big as 20 billion. So, but at the at the heart of it, you know, the thing that it takes to build a iconic generational category defining company when you look at the team in our view hasn't really changed.

And these things will happen much faster. They'll be more democratized because it's easier to start a company today than it's ever been before. It's easier to raise funds than it's ever been before. It's easier to grow fast in some of these verticals and and different domains than it's ever been before. But at the end of the day, it's about meeting with a founder, getting that that resonance, understanding what motivates them, what drives them. and really, it's a birectional connection because it takes a lot to build and help build these companies. And so from that perspective, we want to have a team that can identify and really deeply resonate with founders across the domains that we think matter. but I think the other part of this is that as a small, you know, what's really important in venture, it's how quickly can you make really good decisions.

That's pretty much it. making really good decisions. If you can fit your whole team around one table, you'll make really good decisions. If you have a giant team, you'll always kind of revert to consensus, right? Right? So, a small team helps you make really good decisions. And a small team helps you make really fast decisions where if you need to do a deal in a day or if you need to do a deal over the weekend, you can call up your partners. They can all, and by the way, we're all here in the Bay Area. You can all get together, you can go meet an entrepreneur, and you can do a deal in less than a day. and I think that's a unique advantage. The question is how do you scale that given that there are more companies than ever before? It's it's harder, you know.

Hey, I'll tell you like you know, Chad GPT is probably better at writing memos than I am. not even in deep research mode, right? you know, we we're we're long-term investors in Glean, which is an enterprise search and aenta company. We leverage that quite a bit to really understand our companies and our portfolio. we use a bunch of tools to do search and get smart on companies. But at the end of the day, you have to meet the founders and you have to connect with them and you have to really find that that resonance. So that that's where the business is still the business. Yeah. Something cannot be automated. Well, in in our fund, we also we have a small team, but we came to the conclusion that we don't need associates or analysts. Now we have only partners and agents who can actually do better investment memos.

and my question to you taking into account your experience at high growth companies like Uber and Twitter, do you have any advice for the founders of new AI wave on scaling how they can do their modes and keep their culture at companies? Yeah, it's a really good and complicated question. I think part of building an app layer AI company today whether it's consumer proumer or just kind of launching an AI company into this it's like fighting gravity like everything about it is like constantly fighting gravity right and the stakes are so much higher today the competition level is so much higher than I think it's ever been but I think there are things that you can do to sort of increase the probability of success and ultimately scale your company a lot of it is you know it's a momentum game, right? You're fighting for attention.

You're fighting for relevance. And yes, you have to build an incredible product and iterate on it very quickly, but you also have to have that distribution mindset of like what can I do to put myself on the map, to put my company on the map. Things that are working today is founder-ledd marketing. It's building in public. It's like gorilla marketing around events like the perplexity cyber truck, for example, right? that they've taken to different events. It's finding and creating like those viral moments that are going to continue momentum. but at the end of the day, you know, on the other hand, like the only thing that really matters in app layer consumer products is like do you have user love? Do you have a really retentive product? And do you understand like what takes your users down that happy path of if I you know do x number of searches or if I take x number of rides like I'm going to stay with that product for a really really long time.

And so it's also understanding your own user base and their behavior and their engagement patterns so you can build into that and ultimately scale your company not just through top offunnel growth right but great retention in the long run. and then in terms of culture, I think look, you have to move insanely fast and you have to build a culture that gives people a lot of autonomy at the edge to move fast, make their own decisions, like protect that culture of autonomy, experimentation even as you add more people to your team. And you know that has to really come from the founders and instilling that culture in the organization and just accepting the fact that you know we talk a lot about fundraising and venture. Fundraising is important. You don't want to run out of capital. You know you you want to amass resources and keep going.

But it's not really about how much you raise. It's not really about the valuation is like the long-term value of your equity. And your employees have to believe in that and also believe in the user and be obsessed with them where you're going to build something generational versus something that yeah has its moment on Twitter or X and then fades six months from now. Like that's no fun. To me that feels like throwaway work. I'd much rather work on something that is durable and generational and people will still talk about like five years from now or 10 years from now. Yeah, agree. questions to everyone. this conference is second year and I have a question. What sounded maybe crazy just a year ago but now seems like a working product. I guess I'll go first. I think I think it's just I think it's crazy how how steerable AI has become.

I think that is something that has been really surprising because you know back I would say a year ago you'd be like oh oh it got it wrong like hallucinations there's not a thing anymore right and and so but also the steerability part of it is interesting you you know something like runway you can actually produce like really high quality professional audio video and just and launch that right and that is is pretty big and I think the steerability is like crazy is is getting crazy good it's kind of having its point of inflection I take a look at Chhat GPT as well. So yeah, I mean I think one I'm blown away every day like by everything. This is pretty amazing to see the rapid progress and the quality of products. I mean my daughter has like a dinosaur AI toy that she talks to continuously, right?

Wasn't around. It's not very good, but it's good enough for her. but I think you know, the re just the reasoning models and I think the fact that you can now have a real full-fledged conversation with the chat agent to me was probably the biggest aha moment and unlock the fact that you can really almost interact with a quasi human level response mechanism. That that to me is pretty mind-blowing. you know, I was chatting with somebody about the New York City versus San Francisco ecosystem yesterday and asked Chad GBT a bunch of questions and refined it. You know, just what was the answer? Well, the query was how many how many nerds what's the number of nerds in San Francisco versus New York City? good question. It turns out the Bay if you take the Bay Area and and New York City, the all the the burrows, it's roughly the same.

Oh, really? And that's why that was in response to somebody tweeting out that there are more series A companies now that in in New York City than San Francisco. But yeah, I mean it's just that's that's crazy. Absolutely crazy that you can do that today. I think SF is the nerd capital and nothing will ever convince me otherwise. okay. One thing that's really surprised me, I have to say, because I think part of the acceleration, it is like the obviously the AI kind of tech development, but part of it is like habit change, consumer habit change. It's the things that we thought were like really weird before, but now we take to be totally normal. and there was a time when like ride sharing was kind of weird, you know. so we've accepted a lot of things over time. The thing that I've noticed is like a year or two ago like nobody really wanted to be recorded all the time and now like we have meeting recorders like the AI kind of like notetakers.

We have pendants that you can wear that will record everything you say. It's the idea that we're willing to be sort of recorded always on in an ambient way. And I think when I like first encountered this concept you know probably two years ago now I was like I would never do that. And now I'm like, "Oh, yeah, that sounds so useful." You know, and so it's like the psychological change that has come from experimenting with different AI tools and agents and like seeing the utility of it, I think, is completely changing what we think is status quo or like socially acceptable. And I do think we're going to be recording everything like that second brain memory going forward. Still creep creeps me out a little bit actually. You're not there yet. Well, I actually that's a really good notice because I've actually noticed I'm less creeped out.

You're totally right about that. And that's one of those things I feel like I'm that frog that's boiled. I'm like, "Oh my goodness, it is a little hot in here." But yeah. Yeah, that's great. All right, the last question, final one. we still need an alpha. where is it? Is it now in applications, infrastructure? Maybe it's still in foundational layers or biotech, robotics, or something absolutely different where I got two. I'll try to be quick. and it's really exciting because at Felicis we're an early stage fund so we get to like think about that future. the first thing is that we're going through a transition and it's not yet complete between from thinking to doing and that is that's happening because right now the thinking thing again I said the host nation stuff is solved and you know you can you have deep research you have other tools that you you're done with the thinking but now the doing part is going to be really interesting and there's going to be a lot of companies that come out that really have to figure that out.

Related to that is my second one, which is I don't think any of us really knows what happens when you have the AIs do their own thing and talk to the AIs. And that's a really kind of fun thought experiment that I just I I really want to find enough time to lock myself into like a a coffee shop, put on some headphones, and really think about because what does it mean when you can en you know when your agent does this and you know can talk to you know my my agent can talk to your agent. Again, I hate that phrase, but we'll find a better one, right? like what does that mean and what does that look like and there's so many things that will be reinvented as a part of that but this is sort of what I mean by like when you have founders now that are going to express obsess over each of those details of what that interaction looks like there's going to be a whole new ecosystem so I think something around there looking for that sorry one more thing to cheat I think consumers going to have some moments it's been a long time coming and it's been kind of boring in that space I'm a consumer person but I'm waiting and I'm ready Yeah, I mean I think we're on the part of the exponential curve where we're losing the ability to predict the future because it's evolving so fast.

and every time, frankly, we've tried to predict the future and make investments on a prediction basis, it's really backfired. so, so I think the alpha is ultimately in the people that are that are in the space and I think there are folks who are deeply native and understand how these technologies can be applied to all sorts of industries and all sorts of applications from consumer to deep enterprise. I do think you know things like robotics and and more you know hard domains have a natural timeline to them. but software I think is dramatically eating everything and accelerating very quickly. So we're just kind of looking for the smartest people to come and tell us what the alpha is in some ways. But I think the people themselves are are the alpha. Okay. I think that's very well said. It's something I think about a lot because I do think like the next frontier of value creation is around agentic applications and like very specific domain expertise and this can come in many different forms.

I mean there are incredible founders that were building for you know architectural engineering right people were thinking about how to build you know true kind of AI assistance for mental health right like there just so many domains with deep deep deep expertise where you can utilize Agentic AI build agents applications that are incredibly useful and and also kind of bring down the cost of those services potentially to consumers or to businesses. So, there's just a ton of TAM expansion and using AI to make something affordable that just wasn't affordable or accessible before. I'm also super excited about consumer AI. It's where I spent a lot of my time. but you have to find the founders that will be super adaptable and navigate whatever comes next because that similarly like I just I I always say like I don't know VCs are terrible at predicting the future.

Like we try but it's really hard. It really the founders are the visionaries. But my third I think there's a third and this is around like the agentic internet which is like imagining a world in which like the open web as we know it the open internet like it's you know it's not humans interacting with that it's agents right and that changes everything it changes commerce it changes how businesses and brands begin to sort of think about where they get discovered where they transact how do they advertise to and reach consumers and AI in and of itself becoming an entirely new marketing channel so there's a lot of alpha out there. Yeah, probably I can summarize that alpha is overachiever not only thinking but also doing probably something in consumer with strong domain expertise but we are uncertain still not sure. Thank you so much.

It was incredibly insightful. Thank you. That was fun. Yeah. Thank you all. Thanks for being here.