Two VCs On AI’s Possibilities — With Joe Marchese and Michael Mignano

0:00LinkedIn presents.

0:14Welcome to Big Technology Podcast, a show for cool-headed, nuanced conversation of the tech world and beyond. This week on the podcast, we have two venture capitalists talking about the opportunity that they see in this latest wave of generative artificial intelligence. Joe Marchese is a general and build partner at Human Ventures, and Michael McNano is a partner at Lightspeed VC and the co-founder of Anchor. With all the hype we've been hearing about generative AI, I figured it was just time to have a grounding conversation. You know, what is the actual business opportunity here? and is it going to be incremental or a sea change?

0:50These two VCs you're about to hear from believe it's the latter, that this is going to be big and they help us flesh out a bit about where this is going. As you listen, I suggest paying attention to three main categories. One, content creation and entertainment. Two, search and discovery. And three, and this is probably the most interesting one, the AI acting as an assistive agent. So stay tuned for a nutrient-rich, fun conversation. And this one's a bit on the shorter side after a few longer episodes over the past few weeks. My conversation with Joe Marchese and Mike Magnano, coming up right after this.

1:24Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge, and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. The truth is AI security is identity security. An AI agent isn't just a piece of code. It's a first-class citizen in your digital ecosystem, and it needs to be treated like one.

2:05That's why Okta is taking the lead to secure these AI agents. The key to unlocking this new layer of protection and identity security fabric. Organizations need a unified, comprehensive approach that protects every identity, human or machine, with consistent policies and oversight. Don't wait for a security incident to realize your AI agents are a massive blind spot. Learn how Okta's identity security fabric can help you secure the next generation of identities, including your AI agents. Visit Okta.com. That's O-K-T-A dot com. Joe and Mike, welcome to the show. Good to be here. Sure. We're going to talk a lot about the opportunity that this new wave of AI has.

2:44And one of the things I think a lot of people are trying to figure out at home is, is this, it feels like it's something, but is this going to be another fad like crypto was, right? Is this going to be something that has a tremendous amount of hype and fades away? Now I'm going to let you guys spend most of this podcast making the case like that maybe there's something and maybe we should be afraid of it. But let's first start with a theory of what, If this is a flash in the pan, what is the path for it just fading away as a bunch of hype? Joe, I feel like you'd be good to start this one. I mean, I don't know if I'm good for what would make it fade away.

3:22I think there's, because I just don't believe that there's a chance that this isn't fundamentally industry changing at every level. I think the difference is, you know, whereas Crypto and Web3, they have aspirations to make wholesale changes to the monetary system and, you know, basically new forms of government and new forms of governance. Like this AI is fixing the boring stuff. And that's how you know it's a real business. It's like it's just making workflows more efficient at these at these micro levels of like legal processes and HR. And, and like, so just that alone, I don't, I don't see a path that it isn't already going to have an impact on, on multiple, multiple business lines.

4:01So let me, let me rephrase, let me rephrase. So instead of fade away, like have a minimal impact instead of an outsized impact, because you see every day, like all these predictions of how this is going to change the world. And so what is, is there a way where it just like makes a few things a little bit better or like, yeah, what is the path for that? I mean, I think maybe you would want to start. So first of all, let me just say I agree with Joe. So I think if you wanted a podcast of two people like choking each other, this episode may not work out that way. But I agree with Joe. But maybe a framework for analyzing your question and thinking about it is to sort of think about the areas of business or really life in general that stand to be most disrupted by this.

4:50And then maybe the areas that are least disrupted by this. And think about those areas, right? And those areas that this technology just can't or won't get to anytime soon. Now, the question is, which areas are those? Well, I think let's start with the areas on the opposite end of the spectrum. Maybe we can work our way down. I think inevitably, these tools we're already seeing are really good at... doing things that humans can do, but just a lot faster, right? And I think that's why you're seeing things like, you're seeing a lot of these like virtual assistants, you're seeing things that write on behalf of people, you're seeing technology that analyzes text far faster than humans ever could.

5:35So maybe one place we could focus are like, what are the things that are uniquely challenging to do regardless of time, right? I don't know, do either of you have an idea based on that prompt? No, I mean, this is the struggle. I mean, I've spent the better part of obviously the last couple of months as everyone has been just captivated by the rollout of AI into people's hands directly. And the more I've dug in with every business, I'm convinced that it's not a single business. It's that every business is going to have to figure out what role it plays in it. It's like saying that you're a cloud business.

6:14You're not a cloud business because you worked on AWS or Azure, like that was just the most efficient way to host your servers. And so I think that this is, I'm pretty convinced it's pervasive. I guess what you could make the argument is that if it becomes so pervasive that everybody has it and it's, then it doesn't mean anything to anyone. Right. And, but like, you'd still make, you'd still have to make the argument that those that are most efficient at using it will end up winning. Like, I think the line was AI isn't going to replace humans, but humans who use AI will replace humans who don't use AI.

6:44Yeah, we just had Palmer Luckey on the show last week talking about like, AI won't kill humans, humans with AI will kill humans. Sorry. Same philosophy. Yeah, go ahead, Mike. Yeah, I mean, I personally like I think the companies that are most interesting right now or that will be most interesting over these next few years are actually the ones that look to take advantage of the AI more aggressively and more quickly than other companies. Right. The companies that are more quickly going to be adopting things like Copilot to help write their code, you know, the marketing teams that are more quickly going to be adopting ChatGPT or, you know, our products that are based on large language models to help with writing.

7:20those are the companies that are going to have massive efficiency gains. And it's almost like if you're not using those things, you're going to be at a disadvantage. And one thing, one new lens that I've been looking at this through is like, I think, you know, someone said once, I don't know if it was the exact quote, so I won't butcher it or who said it, but because I don't want them to not want credit for this, but it's basically like strategy is doing the hard thing, right? Like everybody has ideas, but like execution is the hard thing. And it used to be the technology was the hard thing, right?

7:52Purely like, you know, we needed enough engineers and enough coders. Now, if everyone's coding on top of these large language models and like using co-pilot, so one engineer could do the work of 10, or maybe even you get to a place where non-engineers can do like a lot of the technical architecture. The question you have to ask yourself is what's the hard thing? Like, what is the new hard thing if everyone has all these tools at their disposal? and and you know i i would make a big case for you know cultural relevancy and brand uh and or you know things that we talk about like in the humanities like what are the human insights that you're that are driving your business and and how do you how do you maintain an edge but it for so long was the ability to go from i have an idea of how to put a button on a screen to needing to build a lot of things to get to do it now that's the easy part so now what's the hard part Maybe the hard part is places where technology can't go or has a hard time going, right?

8:45It almost feels like where technology can go will be completely, AI will be completely deflationary to those areas of commerce or business. But actually the things where technology can't go, maybe, I don't know, whether it be construction or agriculture, and obviously technology can reach these industries, but maybe more like the more offline aspects of these industries are the things that will be most untouched by AI. Yeah. And the reason why I started at the beginning talking about like the bear case here, and I know like, I think the three of us all believe that this is going to be big. But like, yeah, I think there is some concern that, okay, how much should people buy in?

9:26It goes anywhere from CEOs deciding whether they need to input this in their business, marketing teams trying to figure out whether they should use it. Reporters, for instance, thinking about whether they should write about it. And of course, VCs thinking about whether they should invest in it. The one thing that I'll say, and I want to get into some use cases, but the one thing that I'll say, you know, as we think about the bear case is I'm also curious if you guys could help put into context the leap that we just saw, because, you know, if you think about it technologically, ChatGPT was just taking technology that was already there, GPT 3.5, and putting into a chatbot, you know, so we were talking about an incremental step in, and you can push back on me on this one, but we're talking about an incremental step in terms of, you know, computers being able to read and write and that combined with image generation with things like Dolly and Mid Journey, which has gotten good.

10:20So, okay, that's interesting, but why would this represent a moment of like potentially unprecedented economic opportunity if the technology change is incremental? Well, I'll give you one thing on the bear case side that could freeze us is we have no idea how trademark law is going to work with AI right now. We have no idea what the models are being trained on. It's going to be so hard to figure. So there could be a little bit of a delay as we figure out what's investable. Because to your last point there, to what VCs invest in, we're looking at everything we can here. And I will tell you it's hard to figure out investable unless you're going all the way to the arms range of you're going to build your own LLMs.

11:03But the idea of what makes this a step change is two things. One is AI has been directing our attention for a long time. We famously all talk about the TikTok algorithm all the time. We all know what the Facebook news feed is, right? So basically, consumers have been talking about AI without even knowing they were talking about AI. But that was AI directing our attention to things made by other people, right? Now, AI is literally creating the content. When you ask ChatGPT, it doesn't go search the web and then give you a reference. It creates an answer for you. So now, what that does is it absolutely turns on its head search, which how much of the internet do you think is based on search most of it right it turns advertising on its head which is the entire internet economy is built on advertising so like so that means that like just trillions of dollars of market cap are are up for grabs in some ways and but the chilling part of that is we don't really have a replacement model for that yet so like i as much as anybody am very excited for the day the internet is not wholly ad-based because it's done some horrible things to the internet including listicles and you know alex we've talked about these wait a second All right.

12:17You're a former ad tech executive. I'm former BuzzFeed reporter. That hits home for both of us. I know. I think we have some sins to it. To attone for? The internet being fully ad-based was a problem for a lot of reasons. But for better or worse, it is an ad-based economy. And AI, by saying, okay, not only will I direct your attention, but I'll create the content for you. I mean, it's hard to overstate how big a deal that is. Yeah, I think this is a really important point. And it's something I've been thinking about as well. TikTok, per your point, Joe, has proven, TikTok and Instagram, and frankly, companies like Spotify have been doing this for a while.

13:00They've sort of proven to us how good the models can be and the algorithms can be at giving us exactly what we want at the exact time we want it for as long as our attention will stay focused on it. And I think that those models are only going to become more and more needed as AI creates this sort of explosion of content. However, per your point, Joe, they're sort of all driving at the platform business model of today, which is effectively ad revenue or some sort of proxy for ad revenue, right? Let's just call it attention. if the platform can now make the perfect content, if the platform can deliver the perfect content at the perfect time, and now it can make the content at the perfect time, I mean, platforms are going to be, they're already, they already are really, really, really good at keeping us addicted.

13:49They're going to be literally perfect at keeping us addictive. And so I do think new business models are needed to break away from that. I guess the question is, you know, will, will users pay for new models, right? Like, will I pay for a model to give me the content I think I want versus the content that the TikTok algorithm has sort of proven that I want? It's kind of like, you know, I know I should eat more broccoli and eat more vegetables, but I still eat the cheeseburger every once in a while. Can vegetables win? Can a new sort of vegetable model win? Because I think it's going to have to, or something's going to have to give, otherwise we may all end up uh like the the final scenes in wally where we're just floating around on these machines staring at our uh at our screens all the time it seemed bad in the movie but every now and again i i do wish to be that while they looked happy they seemed happy i mean yeah so so i'm just gonna well i don't know if this is reading between the lines or whether you guys are explicitly saying this but hearing you both talk it does seem that content creation ai based content creation and then getting that to people is like the number one big opportunity here like just is just companies creating content through ai how right or wrong is that i mean i don't know how much how much your listeners like have been really going deep on ai and what what the llms are but just let's pause for a second and think about how insane this is when you ask it a question when you ask chat to do a question it doesn't actually understand the answer that it's writing out and it's not copying and pasting from the internet what it's writing out it's taking each individual word and giving a probabilistic outcome to the next word that goes after it and then they end up forming whole ideas and thoughts and and you know you can say you could say i did this yesterday you are david ogilvy come up with 10 taglines for new banana oreos that do that and it comes up and it writes it like him and but he didn't go find it anywhere it created that so it's not it's not just creating content it's it's creating the answers to our questions you know based on the sum total of numbers.

15:53I mean, that should be awe-inspiring to people. And it works right now. This is an example, back to before, where with the crypto and the Web3 side of the world, we're saying, hey, we're imagining a world that's going to change or become this. Well, the world has already changed, and now we're just figuring out what to do with it. But it's unlike anything that we've seen. So it's hard to answer what wouldn't be affected by that. But it's not just content creation. It's answering our questions. That's crazy. I mean, search is a big business. Yeah. And through... Yeah. Maybe it's more than that.

16:31And it's not just answering our questions, but through... I'm sure you both saw the plugins got announced recently. There are companies that are building models specifically for action-taking, not just language. So it's very, very soon going to start. I mean, it already is possible, But more of us are going to see very, very soon at making the leap from words on a screen to taking specific actions that either we direct it to or it directs itself to based on the answers to our questions. Right. Think of one small vertical, right? Like travel, travel industry, OTAs, Priceline, TripAdvisor, Expedia, hundreds of billions of dollars in market cap.

17:12Right. And they really rely on that. You go to Google and you type, you know, where should I go on vacation? And there's a search results page of which they are one, two, three, and four. Most likely you click through one. I'm on Priceline. Oh, I know Priceline. William Shatner used to be the Priceline negotiator for me, right? Like out there negotiating my hotel room rates. So I can trust that this brand is getting me the best deal. Okay, I trust this. And then you make a purchase and there's a rev. That entire flow goes around the other way. If you just say, you know, talk to me about where I should travel.

17:44Here's some things about me. Here's my income. And it's giving you answers. And so then the process of booking, I mean, that's just one vertical of commerce, right? And that is an entire vertical that is reliant on the flow of consumers today that usually start at a search bar. So, Mike, you brought up the plugins. So I feel like it would be great to just talk out what those are because they go into ChatGPT. And part of it is, for instance, like having basically talking with a kayak, right, and having it then go book you a flight. Like, can you expand a little bit upon what it is? Yeah, basically, OpenAI built in basically, you know, a developer ecosystem now to ChatGPT where developers, applications, you know, product services can integrate directly with with ChatGPT.

18:35So, you know, if Joe says, hey, you know, I want to I want to travel to Barcelona next week. What's the best time of day? You know, give me some options, whatever. It can now go and plug in directly with I think actually Kayak has a plug in. Yes. And go do the transaction right from within ChatGPT. And, you know, there are a number of these Wolfram Alpha, Wolfram Alpha. But maybe the most interesting one is there's a plug in that just connects it to the Internet, which is which is a huge deal. because chat gbt um you know or gbt4 is trained on uh you know a corpus of data that goes up until i think you know somewhere in 2021 september 21 yeah september 2021 it's not live it's not connected to the internet it's just a snapshot in time but now it is connected to the internet and so it can go seek real-time information and take actions based on that real-time information so So, yeah.

19:30Think about that example. Think about the kayak example for a second. Like, let's just all pause here. Okay. So, an incredibly smart AI has now plugged into a kayak, you know, so it can go book a year travel. But why does it have to? Like, why doesn't it just plug into Delta and to Marriott and to, you know, like restaurants on the ground? Like, that layer in between doesn't really make sense if the AI can figure out what it is you want to go do, right? And so how long until that becomes disintermediated or risk of disintermediation? That is tons of industries. I think maybe the more interesting question is, okay, so everything's going to plug into this.

20:13Like ChatGPT will be able to go anywhere. Where can ChatGPT go? Right now, it is in a box on my computer. I have to navigate to whatever, chat.openai.com or whatever. Where am I going to be able to access ChatGVT from a year from now, two years from now? That brings that capability so much closer to me, to the thought in my brain, right? Like, oh, wait, I want to go to Barcelona next week. How can I act on that immediately? I think it's probably pretty inevitable. It's going to get much closer to our smartphones, maybe our watches, maybe our smart speakers in our kitchen. Where is this going to be distributed to, I think, is the next question.

20:53And I can't stop thinking about this, and I brought it up on the show before, but this Benedict Evans tweet talking about people have been awe-inspired by the fact that you can have GPT-4 in Microsoft Excel and then ask Excel questions about your data. And what he says is the old way of thinking about this is it's just a feature on programs you use already. The real question is, why do we even need Excel? because if you just upload your data into some AI cloud and dialogue with it, you might not need to have those spreadsheets like you did before. How do you guys think about that? Yes. I mean, I think that's kind of my point is that there's a step in there that might just get skipped entirely.

21:40If AI is plugged into point-of-sale data at restaurants, they might not even have invoices at the end of the month to have to upload because they just won't need to do that. And that is such an amazing efficiency addition for small businesses, for supply chains. I mean, this could be very, very good news. And I know we'll get into the scarier sides of it before we finish here, but it could be very good news in terms of efficiency for the economy, like literal GDP up and to the right with less people. But we don't really have a replacement business model for that yet. Yeah, I agree with that. I guess the thing, you know, getting back to your question, Alex, the part that I struggle with is like, what are the things where I actually do need to touch and feel and see something for it to be really useful to me?

22:37Right. Like, you know, GUIs were created for a reason. Right. Like an actual user interface was created for a reason. And it was because it became easier, faster, more convenient for me to tap some buttons either through a mouse on my desktop or on my phone than it was for me to type into a command line. So I'm curious to see. And I wondered a little bit about this like five or six years ago when the sort of command K within Slack or whatever it is, that started to show up everywhere, right? Like the command bar. I'm curious if chat and voice and interacting directly with an AI will replace this notion of a user interface that has been so pervasive over the past 30 years.

23:24I think that's a good point because I do think that audio architecture has not moved. I mean, a lot of people ask, why didn't Amazon Home get there? Why didn't Google Home get there? Why isn't this all prevalent? And it's because without menu listings, audio is very difficult for, because we all talk a little bit differently, right? We all think in slightly different mental models of how we form a sentence. And so without a command bar, it doesn't work. But this new system, like I would respond to ChatGPT and I'd say, number nine from above, give me this versions of that. There wasn't some architecture that said I had to press, you know, pound sign, number nine, you know, then this, or I had to spell out nine.

24:04It just figured it out. And I think that was what was missing from the last time audio. So audio didn't work before because there's no interpretation. And this has the ability to interpret. And so you won't need kind of a command hierarchy to be known by the user. Like the only commands that were useful on Amazon Home, you had to already know what the commands are. Right. And that's just not helpful because we don't all think the same way. All right. Let's take a quick break and come back here. We're here with Joe Marchese, general and build partners at partner at Human Ventures and Michael McNano.

24:32He is a partner at Lightspeed and the co-founder of Anchor. We're going to be back right after this to talk a little bit more about the opportunity for startups and then some of those worries that Joe mentioned. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge, and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed.

25:10That's how they step. That's technology at Capital One. These days, it feels like every dollar should be working a little harder, but figuring out where to put your cash can be confusing. That's where Wealthfront comes in. Wealthfront is a tech-driven financial platform built to help you grow your savings into long-term wealth. Their high-yield cash account through Program Banks offers a 3.5 % APY on your uninvested cash as of November 7, 2025. And there are no monthly fees, no minimum or maximum balance to earn that rate, and you can even make free instant withdrawals to eligible accounts in just minutes, 24-7.

25:45So your money can always be within reach. Right now, Wealthfront is offering new clients an extra 0.65 % APY over the base rate for three months on up to a$150 ,000 balance. That's a total of 4.15 % variable APY when you open your first cash account. Go to Wealthfront.com slash Big Tech to sign up today. This is a paid testimonial for Wealthfront. It may not reflect the experience of others, and there's no guarantee of future performance or success. Wealthfront brokerage is not a bank. Rate is subject to change. Promo terms and conditions apply. For more information, see the episode description.

26:19And we're back here on Big Technology Podcast. We have Mike McNano here with us, co-founder of Anchor. and a partner at Lightspeed VC, and Joe Marchese, a general and build partner at Human Ventures. Great to have you both on. So we actually talked about some interesting things in the first half that I think they require us to expand upon them a little bit. Now, we talked about the fact that OpenAI is releasing plugins for ChatGPT, and you're both venture capitalists thinking about how can companies actually go up and build on top of this software. So when the plugins came out, Some people were like, well, there goes a generation of, you know, chat GPT startups because OpenAI is just releasing it on their own.

27:00And then I think about Bing and Microsoft's offering. And I wonder in some ways whether these big platforms, like the model has always been to build platforms and allow other people to build on top of the APIs. And at first it seemed like chat GPT and Bing were demos. And I even wrote that that was the case. But now I also wonder, do these companies think they have something so powerful that their use case is actually the demos themselves, right? The ChatGPT and the Bing and not making this available to everyone. I'm curious how you think about that. Like, doesn't it make more sense for like an open AI to just make ChatGPT the platform as opposed to, you know, try to make money on its API because of the power there?

27:44And how does that influence your thinking as a VC? Mike, do you want to take this one? Yeah. The honest answer is I'm not sure yet. And I think they may not be sure yet, which I think is why they're probably doing both. And it's only been a few weeks or a few months or whatever, but it does seem like they are investing more and more into ChatGBT. it does seem like it's becoming more of a destination than it was. I mean, you know, there, I think there are some figures being thrown around about how many subscribers they have and how much revenue they're generating. And they're all pretty mind boggling.

28:21At the same time, you know, I have to imagine, I have to imagine that their API business is really, really healthy and robust as well, because it seems like pretty much every company I talk to is, is integrating with the API and looking to get access to, to more and more credits for their users and more GPUs. So I'm not sure. It's hard to say from where I sit. And if I was at that company, something tells me I'd be unsure inside that company as well. So it's possible they're still just pursuing both. Right, like Joe. I mean, and just to set Joe up, I mean, think about just the Bing example, right?