Episode 250: AI Special Featuring Sierra, Harvey, Windsurf & More

0:09Today on the show, we're doing something a little bit different. we've pulled together highlights from some of our most compelling interviews with the biggest thought leaders in AI, founders, builders, and operators from companies like Marvell, Sierra, Aurora, and more. These conversations reveal how AI is transforming industries like semiconductors, legal, customer experience, and autonomous driving, offering a front row seat to how the future is being built. Let's get into it. First up, we have Brett Taylor of Sierra. He's one of the most impressive people I've ever had the chance to interview.

0:41He reflects on AI's potential to reshape entire industries and how this wave could produce the first trillion-dollar SaaS company by moving beyond productivity gains into real labor markets. There are examples of non-obvious ideas that turned into really big companies. Like you could argue Facebook was one and Twitter. Airbnb, Coinbase. Yeah, and where you're really inventing something new, More often than not, though, the really, really large companies, if you look at the top five tech companies in the stock market that dominate the S &P 500, it was making an operating system in the core productivity suite for the early PCs.

1:28It was, in Apple's case, making the PC. And then, well, iPhone was a very novel product. The mobile phone market was established. shows. And then you look at web search, which was I put squarely in the category of obvious idea in the internet. I mean, it really was. It was a billion. There's like 20 companies. There's like 20 companies doing it. And Amazon, we should buy things online. Like, of course, the idea to start with books was a really savvy business move. But it was, so, you know, I think there's a loose, but I think correlation between total addressable market and obviousness, I think is the big thing.

2:12And so, well, there are some really important companies that started with just like, oh, wow, this is a great idea. Like, wow, I wouldn't have thought of that. most of the largest companies in the world are doing something that is very important and very mainstream and very widely used, like calling customer support. And as a consequence, everyone's staring at that market as everyone should be. This is what capitalism is. So it will drive down costs and make companies with better products, better execution, succeed. And that competition is, you know, as much as I would love to be the only company in this space, it drives.

2:47is why capitalism is amazing, right? I'm paranoid about all of our competitors. We respect some products. We respect some go-to-market teams. Everyone's got different strengths, and you're sitting there every day grappling with that reality. But I think it's because the market is huge, and I think that no company has the privilege of addressing a market that gigantic as an individual company, and the market just wouldn't enable that, right? Yeah. Yeah. And but I think the the reason why it's worth the stress of that kind of intense competition is, you know, I think we'll probably see the first trillion dollars software as a service company come out of this era.

3:32and a pure play, like not building infrastructure and things like that. And I think the reason for it is that these AI agents actually are doing jobs rather than just adding productivity enhancements on someone else doing the job. And I think that value is so quantifiable and so real that people will pay a lot more for it than you will a productivity enhancement. and I think similarly I always gave this example but the legal tech market I think was notoriously bad before Harvey it wasn't I can't think of one really really large legal tech company and but it's an important profession don't get me wrong it's just the addressable market of selling productivity software to lawyers it's just a relatively small number of firms relatively small number right you just do the math in a spreadsheet it's just like yeah you can make a decent company, but you're not going to make a Google out of that.

4:27But then you look at the amount the world spends on legal advice and like doing contracts for your supply chain or doing antitrust review or all these like, you know, long tail of things that are quite valuable. And all of a sudden that market that looked modest looks gigantic. And so the interesting thing about these agents is I think it's changed the total addressable market of software because you're essentially getting into labor markets and things that used to be backwaters of software, all of a sudden, like really interesting software markets. Uh, I think that's really exciting. So I think we're going to see some really important companies come out of this era in the same way we did in the birth of the internet.

5:08Yeah. On the, can I keep pulling on this Google analogy that you're using, which is like, um, obviously we have a history with Google. Um, John Doerr also did a bunch of the other companies that maybe weren't the preeminent ones and then got Google and Amazon right. But Google in this example was not the first search engine. Wasn't the... It wasn't even like the sixth. I was going to say, I don't think it was even the top 10. It was like maybe 15th or 20th search engine. So like, again, let's pull that forward. It's maybe history doesn't repeat, but it rhymes loosely here. I don't know. But like all of these companies that we're talking about, let's take Harvey in legal or Sierra in customer support, Cursor and Windsurf and coding, right?

5:56There's like a few categories that we're talking about here. Maybe glean and search, right? Like there's like the mother of all the markets today. They're first, you're first. Do you think about it differently and what it means to be first in these markets today versus somebody waiting in the wings, waiting for the models, maybe a different breakthrough. Yeah, how do you think about that? Let me go back to the Google analogy for a second and then answer the question because I think the nuance matters here. I think there was a narrative that is not completely untrue that Google won because PageRank was better.

6:35You know, a lot of search engines were reductively based on analyzing the content of the webpage, which made it really prone to spam and made it really hard to figure out which of these web pages talking about this topic was actually canonical. PageRank and looking at links and basically sort of anchor links between web pages had essentially a notion of reputation and importance and credibility in addition to analyzing content and the combination of those two just made it work. You could search for something and be relatively likely to hit in the top results, something that actually was what you were looking for.

7:15So it's a really important technical insight. I think some people say, okay, that was Google one because of Patreon. Yeah. Well, there's a couple other things like Google also signed deals with all of the major portals around the world. Google powered search on Yahoo for a really long time. And if you went around almost every country and every market, you know, essentially Google distributed its search engine and sort of like, you know, basically got distribution through a really effective partnership structure. Had they not, that would have opened the market to a lot of other competitors like InktoMe who ended up acquired by Yahoo to go into that market.

7:48Secondly, when I got to Google, I don't believe AdWords existed, but there was banner ads at the top. You could buy a keyword as my recollection. I might be embellishing history here. And then going to not only cost per click based ads, but the way AdWords ranked it, which was proportional to the click-through rate too, which incentivized high quality ads in addition to optimizing for sort of an auction so people could bid on those keywords, created one of the greatest business models of all time. It was essentially like mapping the demand of intent on the internet to the supply of people who could fulfill that intent in a very efficient auction.

8:34And so then you look at the Inktame's of the world that were licensing their technology to portals, and you realize at that point, you know, the Google's business model was just better. It was just generating cash, like it's going out of style. And you just couldn't reproduce that in a B2B context. So you have the partnership strategy, you have the business model, which was really important if Google's business model had remained selling that yellow enterprise search appliance, it would not be the company it is today. It wouldn't have been able to finance things like Google Maps that I worked on.

9:05You wouldn't have had a business model to support that. All of it's important. Business model, distribution strategy, Google's costs were much lower than a lot of competitors because we built our own data centers and the cost to serve was meaningfully lower. All those things mean you can just do different things with your technology or product. Okay, so now let's go to the AI market. How you price matters. Are you charging based on consumption? Are you charging for outcomes? Are you charging just a site license to use the product? That matters a lot. Because, you know, my intuition and the reason why Sear does outcomes-based pricings is outcomes are valuable.

9:48You know, people know the value of an outcome. If you end up charging proportional to technology, you're probably going to end up marginally, just you have a thin margin on top of the price of AI, which is going down rapidly, like even as we speak. Similarly, what is your cost to serve? I think that there was a lot of companies that created two and a half years ago where they went into places like Kleiner and said, we have all these great AI researchers. We're going to train our models because we're really smart. Well, it turns out that spending tens of millions of dollars to train your own model when the foundation model market is sort of commoditizing turns out to be a really unintelligent use of capital at least in the applied ai market most of those companies have been consolidated now at this point uh you just can't have the business model of a pharmaceutical company where you spend all of your capital and license it out um and so you know and then similarly like what is your go-to-market model is it product-led growth is it a developer motion like Cursor?

10:48Is it an enterprise software motion like Sierra? That also impacts your cost to book. It impacts what parts of the market you're able to address. And it also impacts things like attrition when new people come around. The reason why Google was able to overtake AltaVista was you just visited a new URL. It's really easy. That's always been Google's argument for when people say they have a monopoly. It's, well, just go to a different search engine. They're not completely wrong on that, right? And, you know, in contrast, you know, if you look at, say, a large ERP deployment, you're, you know, it's hard to rip out an ERP system.

11:26Every large company in the world has had these like multi-year ERP system replacements. And that's an example of like a market that's like slow to address, but very low attrition when you address it. So I think all these things matter. So I don't think like a first mover advantage on its own is a sustainable advantage. And in fact, if you look at the dot-com era, it wasn't always maybe even less likely if you're the first mover. But I don't think it's causal necessarily. I think that what really matters is like, what is your sustainable sort of like technology point of view and advantage? What is your go-to-market model?

12:05How do you manage your customer relationships? What is your business model? I think like a pricing model, and I think all of these things matter. The fun part about new technology is, we didn't have cost per click ads in the early days of the internet, and now it's pervasive. Now we can even do conversion-based ads and things like that. I think the exploration going on in the industry on our approach to outcomes-based pricing is really novel. It's new. I think it's a really good idea. We'll talk in 10 years, and I'll tell you whether it was actually a good idea. I don't think it's going to take that long enough.

12:39I think we'll find out soon. How come based pricing meaning? We charge when our agents autonomously resolve the problem, and it's free otherwise. And we do that so that at Sierra, our business model is aligned to your outcomes. It's very new, and it's very much like the difference between impression-based ads and cost-per-click ads. These are things that excite me right now is I think we're going to find some key conceptual ideas that are hard for incumbents to copy, maybe give you a different angle than your competitors. And I think just like you can, the narrative I had around Google, which is probably at least half right, you know, on why they succeeded.

13:21With the benefit of hindsight, we'll know which of those key insights were actually like the meaningful ingredients to winning in these really big markets. Next is Winston Weinberg of Harvey on how AI is transforming legal work, automating tasks while elevating lawyers to more strategic advisory roles. An amazing example of this technology eating what we thought were uneatable traditional industries. You should take AI and basically apply it to X industry. Like I think that if your company isn't doing that, I don't think you're ambitious enough. If you are doing that, you need to partner with the industry.

13:55These industries are incredibly complex. Legal is one of the oldest professions known to man. There are firms that are over 100 years old. There are firms that are hundreds of years old. And having a brand that says, we are partnering with the industry to transform it versus we are just going to steamroll the industry is really important for us. And I think that it's led to one kind of a unique brand where it's like very tech forward, plus, you know, kind of like caring about tradition, right? And I think we've nailed that and we have really good brand folks at our company. I think that the other piece of it is just the importance of making that clear to clients and interacting with clients and helping, like letting them kind of help define the brand too.

14:45Can you help me reconcile something? Like Ilya calls it a co-pilot, right? You have this idea of an agent, which is in many ways doing a lot of the work of what somebody could be doing. You know, you see Salesforce that goes out and has their AI cloud or whatever, and then they go, you know, they're telling customers that they can replace all of their salespeople, and then they go hire a giant sales team to sell their AI cloud. Like, how do you see that? Maybe they're training off the salespeople. How do you see that world unfolding for, like, is it just, are we just saying that now so we don't scare away customers that we're going to, like, replace people?

15:19I think the reality is there is going to be a lot of task automation. Task automation doesn't mean job automation, right? And so I think that that's something that people miss a lot. Like you are a transactional attorney. Part of your job might be like looking in a data room and finding change of control provisions. Is that going to change in the next couple of years? For sure. These systems are going to be able to do that at a level where you are checking and you're probably not doing necessarily like hands on all of that work yourself. Right. Is that are you an M &A attorney now? Like that's it?

15:51No. There are so many different pieces, especially to high-level knowledge work, that I don't think these models are just going to be able to automate immediately. And so I think what you're going to see is increasing levels of task automation, which, again, does not mean job automation. Those are different. Do I think that jobs have to change? If 70 % of the tasks in the next three years of your job are automated, does that job have to change? Yes. But that doesn't mean that it happens overnight. It's not just your entire job is gone. It's the task get automated over time and then your job's going to change.

16:27I think that will happen. Like I do think jobs are going to change. I mean, I think it's kind of like the early days of Excel, right? If you work in finance today and you can't use Excel, you can't work in finance, but it doesn't mean the job of finance went away. It's just you have to use this tool as a core part of your toolkit and be really good at it. And the models are going to... I think that's a good analogy because it's like there are tons of Harveys for financial services, right? And they're going to build systems that automate Excel, right? And does that mean that everyone that works in private equity is gone?

17:01No, right? It's just like - I'll bet your capital stays. Yeah. Our work is as subjective as it gets. Exactly, yeah. No, but my point here is like, yeah, that task is going to change, but they're probably just going to do different things. And I also think like, and this is maybe less about every domain and more about the domain that we work in. But a lot of what you want to do as a lawyer are these things that I think you will be able to do much faster in your career, right? So like 50 years ago, more people were like the CLO. Like most lawyers were like a CLO. Like they were business advisors, right?

17:40Like they were not basically like sitting in data rooms. Chief legal officer. Yeah. I mean, or they were an advisor, like they were a strategic partner to businesses, right? And now you have to be basically the tippy top of the profession in order to be doing that. Like most of what you're doing is basically giving all of those, giving those advisors all of the insights from all the work you're doing so that they can advise. That's like what you're doing, right? And my point here is the more that you can automate a lot of those tasks, the more earlier on in your career you can start doing the advisory work.

18:17And I think that's going to happen. And from a client, like I'm a client of a lot of law firms now. Right. And so my perspective on this has has even I've been more convinced of this as I've been more and more of a client of law firms of for that, like very expert advice, I would pay more. Like the reality is like for, should you buy this company? Or should you structure the deal like this? Or should you incorporate an XYZ state? Things like that are incredibly valuable, right? And I would pay more for that than I'm currently paying for if I could pay less for the stuff that AI could automate, if that makes sense.

18:53Makes total sense. It's also kind of an interesting question to ponder whether that's ever going to be automatable or managed by a model, right? Because it's subjective. I think that's right. Right. And there's also just like, there's so many pieces, like a lot of people have the argument of, well, these are the things that no matter what these models are not going to be able to do because of X, Y, Z. And I don't know where I fall on that. I do think that the models can do a lot if they have all the context, but the reality is like, are we going to carry a model around with us in our pocket that's listening to all of our conversations and all, and we're giving feedback on like what our instinctual responses to people are and all of those things.

19:33I don't know, maybe. And I think even with that, there's tons of things that humans do that these models couldn't do. But I also will push on the point that if, like, assuming there's more context for the models, they will increasingly be able to do more things. Now we hear from Matt Murphy of Marvell. He explains why semiconductors are at the heart of the AI boom and how just a few companies, himself, NVIDIA, AMD, are enabling this massive shift in hardware. Why is the semiconductor industry, why is Marvell, why is NVIDIA, why is it so important to this transition right now? Like why has basically all of the value so far in the last four or five years, sans like maybe OpenAI, Anthropic and a few others, basically all accrued at this layer?

20:25Yeah. Well, a couple of things. One is it's certainly, well, take a step all the way back, right? Fundamentally, at least for the last 30, 40 years, semiconductor technology has been one of the key growth drivers of global economic growth. and it's the foundation and basis for almost everything we do. And having been through all these different product cycles, there's different benefits it's brought to consumers and human beings. But in every one of those cycles, you know, there was a huge sort of value attribution, if you will, to a certain set of companies. Like notebooks, like notebook PCs when I started in the chip industry were a luxury item.

21:19Nobody, you know, they were very expensive. They were heavy. That trend happened. Hundreds of millions of notebooks started shipping. You know, Intel benefited big time, but so did the rest of the ecosystem, the memory vendors. I was at an analog company. We were doing all the power management back then for notebooks. Helped us tremendously. And in each of these cycles, the networking wave or optical communications, the cloud computing trends, smartphones, there's always this sort of big boon that comes from that into the chip industry. The interesting thing in those prior cycles is it was very broad-based.

21:57Like almost every semiconductor company had their own angle in one of these types of applications. So think of the whole industry gets lifted. But the AI opportunity is so unique that it's a lot more value being created because the chip technology has a bigger lever on kind of the end result, which is ultimately productivity. I'm talking about global productivity. And there's fewer companies that possess that unique technology. So this is not a broad-based thing. So in the AI wave, it's a big disconnect between the top three, four, five companies that are really in there and then everybody else who has some exposure to it, but it's not enough to move their needle.

22:46Because they're another end markets too. And so that's coming back to my theme. I've just always believed like market always wins. Like that's what, that's what you pick. And then you put the best team together, the best products, et cetera. And so, and then, and then if you think about the value that Gen AI is creating and where that's coming from, it is coming from the fundamental at the base, at the most basic layer, the coupling of, you know, breakthroughs in high performance computing, right? which NVIDIA has really driven, but now other people are also coming in with their own homegrown solutions, the large cloud providers, but NVIDIA pioneered all this in terms of the compute layer.

23:25But then also coupling it and going back to Mellanox, which they did very well, coupling it with the networking, the connectivity, the overall system design. and we participate in that too, within some within their ecosystem, but also within all the other cloud providers, that it's just this very complex, very kind of tightly knit closed system that only a few companies have the basic technology to provide it all. And I think the breadth as well. So it's different than the other cycles. I'm just thinking out loud. Yeah. Tell me if you think I'm crazy here. I had John Chambers in this room like a month ago, who is the CEO of Cisco.

24:12Of course. And at one time in the world, they were the most valuable company. And hearing you describe the role that NVIDIA and you all play in this ecosystem really reminds me of the role that Cisco played in the days of the internet. It was all the piping and plumbing. Right. And again, for a long time, not forever, but for a long time, the value accrued to them. That's right. Now I'm sitting here now, like it's, you know, middle of 2025. It's definitely different. You know what I mean? Like, because at some point Cisco cratered like 90%. Right. I'm not, I don't, I don't know. But like so far, history has rhymed and maybe this tailwind is stronger, bigger, more permanent.

25:05I don't know. Yeah. I don't know. Like, how do you forecast that? Yeah, it's hard to forecast. You know, numbers are all over the map. But I think what's what's there. Then there are parallels in terms of, you know, as there been a build out cycle and CapEx and money being spent and then is there digestion and then there's economic, you know, there's a lot of factors that happen in the sort of networking boom. right which i referred to and you kind of brought back with the cisco story and now we're in a ai capex boom right now and at some point there'll be a digestion it'll be some moderation like like just you would normally expect i think a couple things though for that's really exciting about this one one is you know and kind of akin to the innovation that cisco drove it it really enabled the internet to exist, right?

25:58And they drove it. And now you have companies like NVIDIA and others and the cloud companies creating their own solutions to let AI exist. The interesting one is that on the AI side, I mean, there is, and you pick your number, you guys have a number in your firm and others, but there's, I don't know, four or five trillion dollars of productivity out there, right, that AI could really go after in the world. It really could create a massive, for those that win in it, it could create massive, massive economic opportunity. Next is Yamini Rangan of HubSpot. She talks about cutting through AI hype and focusing on what really matters, creating meaningful value for customers and the urgency that she feels to disrupt before being disrupted.

26:46One of the things I feel is there's a lot of hot takes and hype and all of these. And every cycle, whether it turns out to be real or not, has like hot takes and every day like some crypto this and Web3 that and now AI this, right? It really comes down to customer value. And you got to go back to, you know, it doesn't matter what all this technology talk does. You got to go back to who's your customer? How, what is their job to be done? And can you improve that and add value? That's how we look at it. We have like a North Star, which is all for the customer, SFTC. We talk about this all the time.

27:27And then it comes back to great, forget all the hype and forget all the cool technology. Can we take a neat feature of AI and make it a necessary feature in AI? If we can prove to ourselves with customer usage and repeat value that we deliver, then there's something in the cycle. If it's not, if it's the cool thing that you do today and then next week it's like churned and everybody is left, then there is no value. So I think you've got to like ground yourself in customer value and truth. And if you do that, then you can sift through it. But, you know, at the beginning of the cycle, you just have to like, you know, use some kind of pattern recognition to say, is this going to work or not?