The Exchanges

Q Exactly. Listen, I loved it because I also got so much context that I wouldn't normally get, even in like, you know, prep calls, which I think are generally bullshit, to be honest. Um, but I wanted to start with Zoom. This was your first deal at Emergence, and I just wanted to start there. Talk to me about Zoom. How did it come to be?

A We aspire to be a thesis driven firm. And before I even joined Emergence in 2014 and 20 13, the firm had developed a thesis around the fact that there was an opportunity to replace Webex, that Webex was a tired product that didn't, wasn't very good. In fact, when I was interviewing Emergence, the case they gave me to interview Emergence was for a company called Fuse. Fuse was video conferencing software, an early competitor to Zoom. And I was supposed to diligence that case and then make the recommendation should or shouldn't invest, and they were going to hire me based upon that. Did a bunch of work, ultimately concluded we shouldn't invest, made that case. Fortunately, I made the right call. That was also a decision they made, and they hired me. Fast forward a few months, I joined the firm, and the very first deal that we're pursuing, where I'm tapped to lead diligence, ism. So the good news was we had a prepared mind around the space. We also saw incredible early product-led growth. The company was around two or three million in revenue, was growing very quickly, but obviously very, very early. And we believed in Eric. Eric was the VP of Eng at WebEx before, so knew a lot about the space, and he'd rebuilt the core technology called the Codec, and it worked really, really well. My partner Santi, who ultimately led the deal, is from Argentina, and he used the product to call h…

AI assessment note: “We had a prepared mind around the space. We also saw incredible early product-led growth.”

Q I do want to go back to the element you mentioned about Zoom, which was freaking nuts, which was a hundred X revenue. 10 years ago. I mean, a hundred dollars revenue today is more normal. I still think it's crazy, and a lot still think it's crazy, but then it was completely unheard of. And so my question to you is, have the best always been expensive?

A They are not always expensive, but they are often expensive. So if I look back at our portfolio, Gusto was expensive. Zoom was expensive. Yammer was expensive. A lot of the good ones, Ironclad was expensive. But some of them weren't. Viva wasn't expensive because that was non-consensus at the time. Um, Sales Loft was also non-consensus at the time and was not expensive. More recently, my partner Loti led an investment in a company called Federato that's AI software to help insurers underwrite better. But she made that investment before the Zeitgeist, before people were like, oh, this is obvious and this is going to happen. And to her credit, there was a lot of, you know, questions and she pushed through and she got that deal done and she got it done at a pretty good price. And then the zeitgeist hit, and the company did a series B at a much higher price. So I do think that it is possible still in this world to be non-consensus and right and get a good price, but it is also true that there are increasingly higher, you know, more and more and more consensus deals, and you want to be in, you want to be in both.

AI assessment note: “They are not always expensive, but they are often expensive.”

Q What's been your worst deal and what did you learn?

A Thus far, I haven't had any zeros. I'm sure I will. Um, the exit I've had that was the worst exit was a company called Comfy, um, which was building energy efficiency software. So basically it provided employees within an office, the ability to change the lighting and temperature from their phone, wherever they were, and would actually follow them around and remember their preferences and change the building accordingly. It's actually quite cool. The business grew really quickly from a bookings perspective. Um, they would have these big seven figure contracts from Salesforce and others. Um, but the people in charge of deploying the product didn't care. And so there was a huge incentive issue between the buyer and the implementer in that business. And so we have huge bookings and we didn't have great deployed ARR and that gap bit us in the ass. We ended up selling the business to Siemens. We actually made a little bit of money on the deal and Andrew, the CEO and I stayed close. He actually bought me A gift certificate, um, to, uh, to the French Laundry, which I still haven't been able to use because the reservations are so hard to get, to thank him for helping navigate through the, the outcome he ended up making a good amount of money.

AI assessment note: “the exit I've had that was the worst exit was a company called Comfy”

Q How do you do knowledge management across partnership? And so what we do, for example, is we record calls so that I can listen to the call Jake had with the customer and I'm there in the room. How do you do that shared knowledge across diligence process?

A Yeah. So we do record the calls. We also send out really detailed notes from every conversation. And then every night we send out an email with a summary of what's going on. So it's like, here's what we learned today. Jake did this call. Harry did this call. We talked to this customer. Here are the outstanding questions. Here's what everyone needs to dive in on. We need help with, et cetera. And so it's this constant stream of information that's bookended with these nightly emails. The other thing we do is we have a lot of calls at night. So one thing we realized is that if we're having these diligence calls, like particularly the internal calls where we're processing all the information, if we're doing that during the day, they get compressed because we have 30 minutes and we're just getting into the meat of it in minute 27, and then we have to go do something else. The reality is like, if you do the call at night after kids go to bed, after you've had dinner with someone, whatever, you have theoretically an unlimited amount of time on the back end, which means like the reality is we do a lot of late night calls discussing what we've learned and trying to synthesize, you know, the day.

AI assessment note: “So we do record the calls. We also send out really detailed notes”

Q in AI. Um, we see people like Klarna, You say, ah, we're replacing all of our SaaS tools, and we're building them ourselves, and AI allows us to build all of these tools ourselves, and people are genuinely asking the question, really, especially vertical SaaS, are all of these tools dead, and we'll be able to have very custom applications that we build ourselves. How do you think about that?

A So I have a strong take on this, and I realize that it is a self-serving take, in that I am an investor who invests in B to B software vendors, and so I obviously hope that B to B software vendors continue to exist in this world. I think it's probably also helped you as well. But I believe that B to B software, I believe software vendors have an important role in the future, even if the bolts of the world, the curses of the world make coding cheap, easy, in some cases free. And there's three reasons why I think that's the case. The first is when you're buying software from a vendor, you're not just buying the code, you're buying an opinion perspective on how to solve a problem. And that's a really important point. Like ultimately, if there's a software vendor who has dedicated their lives to figuring out the best way to solve a problem across a bunch of different use cases, they're going to have a lot more insight on how to solve it. And they're going to have that proprietary data, sort of like I mentioned in the mortgage use case, that an open, that a closed source model is not going to have, that you can't just get off the shelf. So that's like your, the first reason you're buying it, you're buying an opinion perspective. The second is the very factors that are making the software easier to build yourself also make it harder to maintain, right? So you could spin up something …

AI assessment note: “I believe software vendors have an important role in the future”

Q A couple of things there. One is, um, margin. In a lot of cases, these are essentially funnels for open AI around Tropic. How do you think about margin improvement over time slash margin maintenance over time, given they are funnels for LLMs today?

A Yeah. So I don't know. If this is a commonly held belief or not, but I, and I think we in general are not super concerned about the margin that OpenAI and the closed source models are commanding for two reasons. One is you've already, there's a lot of competition amongst the closed source models, and you've already seen pricing decline a lot. So most of our application layer companies that are providing applications on top of these products are seeing gross margin increase over time because of that competitive dynamic. The second reason I'm not that concerned about it is open source LLMs are really, really good and getting better. And so the reality is if you're an application provider and let's say for whatever reason OpenAI comes to you and says, you know what, it's 10 times the price and that eats into your gross margin, you now have a credible ability to go and spin up an open source model and have almost no gross, basically have a hundred percent gross margin. And that's, that's actually what Together.ai does. So part of the reason why they've grown so quickly is because companies are like, you know what, actually we'll spin this up on my own. I'll have complete security, data privacy, et cetera, and I control my own margins.

AI assessment note: “seeing gross margin increase over time because of that competitive dynamic”

Q What deal have you lost? And who did you lose to?

A The first deal I lost, I think I've lost two or three deals in my career so far. The first deal I lost was Ironclad. And I lost it to Jess Lee at Sequoia. Um, it was super painful. So the context is I had gotten to know Jason, the CEO there, uh, before he started the company. So he was at a coffee shop and met my wife. Um, cause she was wearing like, I guess a HubSpot shirt and like, he was curious to learn about sales. This is what they tell me. Um, and they talked, um, And Danny was like, my wife, Danny was like, this guy's amazing. Uh, you should meet him. So, um, I met him. I really liked him. And he started the company at the time. We weren't doing much seed investing, so I didn't look at the seed. And then at the a, um, we had invested in a company called simple legal, which was like billing software for lawyers and on their roadmap, they had contract management. They hadn't built it yet, but it was on their roadmap. And so we made the investment and then I called Jason and was like, listen, man, I really love, love spending time with you, but I, I think I should not be part of the series A because of, because of this. He agreed. And so we didn't participate in the series A at all. And then the series B came super quickly. And to Jess's credit, she ran fast and got in front of him first and put the term sheet in and won it. Obviously the happy story is I ended up investin…

AI assessment note: “The first deal I lost was Ironclad. And I lost it to Jess Lee”

Q I do want to go back to the element you mentioned about Zoom, which was freaking nuts, which was a hundred X revenue. 10 years ago. I mean, a hundred dollars revenue today is more normal. I still think it's crazy, and a lot still think it's crazy, but then it was completely unheard of. And so my question to you is, have the best always been expensive?

A They are not always expensive, but they are often expensive. So if I look back at our portfolio, Gusto was expensive. Zoom was expensive. Yammer was expensive. A lot of the good ones, Ironclad was expensive. But some of them weren't. Viva wasn't expensive because that was non-consensus at the time. Um, Sales Loft was also non-consensus at the time and was not expensive. More recently, my partner Loti led an investment in a company called Federato that's AI software to help insurers underwrite better. But she made that investment before the Zeitgeist, before people were like, oh, this is obvious and this is going to happen. And to her credit, there was a lot of, you know, questions and she pushed through and she got that deal done and she got it done at a pretty good price. And then the zeitgeist hit, and the company did a series B at a much higher price. So I do think that it is possible still in this world to be non-consensus and right and get a good price, but it is also true that there are increasingly higher, you know, more and more and more consensus deals, and you want to be in, you want to be in both.

AI assessment note: “They are not always expensive, but they are often expensive.”

Q Do you think we'll see the specialization of LLMs? I mean, if I'm Anthropik now, I'm like, for fuck's sake, just appreciate that you have Cursor, Codium, and you have an unbelievable coding machine. Be that, and that's a huge business. Do you think we'll see that specialization or not?

A I can't speak for anthropic. And it seems like Dario is focused more on the long game of like, how do I do this AGI think safely? And so my guess is his ambitions are focused there, but I do think that you're going to see a lot of specialized LLMs. And I think that a lot of them will come from open source back to the earlier point. Like you're going to see people who say, you know what, I'm trying to solve a problem in the mortgage world, and I'm going to build on top of an open source LLM, a tool that helps me analyze and make recommendations on how to write the best mortgages, you know, in a very specific way. And the cool thing about that, and this ties into, so we had a thesis back in 2017, my partner Gordon started it called coaching networks, which was, um, a poorly branded, but I think correct insight that the way AI will take place in business, business software is as a coach that'll show up and say, Hey, I see that you're about to write this mortgage. Here's all the data you should actually be using. And here's some suggestions on how to do it. It learns on what actually happens. You write the mortgage. You don't, does the person take it or not? Do they pay their loans or not? And then based upon those outcomes, it makes better recommendations to anyone else in that situation in the network. So we call that coaching networks. The reality is copilot is a term that took …

AI assessment note: “I do think that you're going to see a lot of specialized LLMs.”

Q How do you do knowledge management across partnership? And so what we do, for example, is we record calls so that I can listen to the call Jake had with the customer and I'm there in the room. How do you do that shared knowledge across diligence process?

A Yeah. So we do record the calls. We also send out really detailed notes from every conversation. And then every night we send out an email with a summary of what's going on. So it's like, here's what we learned today. Jake did this call. Harry did this call. We talked to this customer. Here are the outstanding questions. Here's what everyone needs to dive in on. We need help with, et cetera. And so it's this constant stream of information that's bookended with these nightly emails. The other thing we do is we have a lot of calls at night. So one thing we realized is that if we're having these diligence calls, like particularly the internal calls where we're processing all the information, if we're doing that during the day, they get compressed because we have 30 minutes and we're just getting into the meat of it in minute 27, and then we have to go do something else. The reality is like, if you do the call at night after kids go to bed, after you've had dinner with someone, whatever, you have theoretically an unlimited amount of time on the back end, which means like the reality is we do a lot of late night calls discussing what we've learned and trying to synthesize, you know, the day.

AI assessment note: “we do record the calls. We also send out really detailed notes from every conversation.”

Q in AI. Um, we see people like Klarna, You say, ah, we're replacing all of our SaaS tools, and we're building them ourselves, and AI allows us to build all of these tools ourselves, and people are genuinely asking the question, really, especially vertical SaaS, are all of these tools dead, and we'll be able to have very custom applications that we build ourselves. How do you think about that?

A So I have a strong take on this, and I realize that it is a self-serving take, in that I am an investor who invests in B to B software vendors, and so I obviously hope that B to B software vendors continue to exist in this world. I think it's probably also helped you as well. But I believe that B to B software, I believe software vendors have an important role in the future, even if the bolts of the world, the curses of the world make coding cheap, easy, in some cases free. And there's three reasons why I think that's the case. The first is when you're buying software from a vendor, you're not just buying the code, you're buying an opinion perspective on how to solve a problem. And that's a really important point. Like ultimately, if there's a software vendor who has dedicated their lives to figuring out the best way to solve a problem across a bunch of different use cases, they're going to have a lot more insight on how to solve it. And they're going to have that proprietary data, sort of like I mentioned in the mortgage use case, that an open, that a closed source model is not going to have, that you can't just get off the shelf. So that's like your, the first reason you're buying it, you're buying an opinion perspective. The second is the very factors that are making the software easier to build yourself also make it harder to maintain, right? So you could spin up something …

AI assessment note: “I believe software vendors have an important role in the future”

Q Do you think that's great? Sorry, I'm interrupting that. Like compare that to Finn, which is Intercom, who actually do it on like outcome based, which is like solution granted.

A The direction that this world moves over time is solution granted. Um, I've spent a bunch of time, um, learning about the Finn approach and that it's hard for now. And the reason it's hard for now is, uh, there are a few reasons. One is, uh, back to the accountability part. It's hard to establish causality In the sense of like, if many support tickets, particularly higher level support tickets have multiple touches, right? Like, so someone, you know, a bot touches it and then maybe a human weighs in a little bit over here. And then how do you establish, you know, who was the winner? You don't want to create an antagonistic relationship with your buyer. If you're like, okay, I did all this. And they're like, no, no, you only did some of this. I'm only going to pay you this. And all of a sudden, instead of having like a monthly, like an easy bill, it's like you're negotiating every month with the customer. That sucks. I think that like over time, we'll start to figure out some of those hiccups and bumps. But I think we're still in kind of early land on outcomes.

AI assessment note: “learning about the Finn approach and that it's hard for now.”

Q What deal have you lost? And who did you lose to?

A The first deal I lost, I think I've lost two or three deals in my career so far. The first deal I lost was Ironclad. And I lost it to Jess Lee at Sequoia. Um, it was super painful. So the context is I had gotten to know Jason, the CEO there, uh, before he started the company. So he was at a coffee shop and met my wife. Um, cause she was wearing like, I guess a HubSpot shirt and like, he was curious to learn about sales. This is what they tell me. Um, and they talked, um, And Danny was like, my wife, Danny was like, this guy's amazing. Uh, you should meet him. So, um, I met him. I really liked him. And he started the company at the time. We weren't doing much seed investing, so I didn't look at the seed. And then at the a, um, we had invested in a company called simple legal, which was like billing software for lawyers and on their roadmap, they had contract management. They hadn't built it yet, but it was on their roadmap. And so we made the investment and then I called Jason and was like, listen, man, I really love, love spending time with you, but I, I think I should not be part of the series A because of, because of this. He agreed. And so we didn't participate in the series A at all. And then the series B came super quickly. And to Jess's credit, she ran fast and got in front of him first and put the term sheet in and won it. Obviously the happy story is I ended up investin…

AI assessment note: “The first deal I lost was Ironclad. And I lost it to Jess Lee”

Q Can you talk to me about market pull being the most important thing of all for people listening, founders or investors? How do you think about that?

A You want people desperate for your product, and I think, like, that's something that is so overlooked when someone's starting a company. I think particularly when someone's starting a company because they want to start a company, not because they're trying to serve a specific need. You want people who have tried desperately to solve this problem themselves, right? It's not, it's not a desperate problem if someone hasn't, if your buyer hasn't tried to hack together something on their own to solve it, or if they haven't bought an inferior product to solve it, or if they're not spending countless hours themselves dealing with it. Otherwise, it's nice to have. So you need something that is just like, Oh my God, this is a massive problem I need to solve.

AI assessment note: “You want people desperate for your product, and I think, like, that's something that is”

Q Do you think we'll see the specialization of LLMs? I mean, if I'm Anthropik now, I'm like, for fuck's sake, just appreciate that you have Cursor, Codium, and you have an unbelievable coding machine. Be that, and that's a huge business. Do you think we'll see that specialization or not?

A I can't speak for anthropic. And it seems like Dario is focused more on the long game of like, how do I do this AGI think safely? And so my guess is his ambitions are focused there, but I do think that you're going to see a lot of specialized LLMs. And I think that a lot of them will come from open source back to the earlier point. Like you're going to see people who say, you know what, I'm trying to solve a problem in the mortgage world, and I'm going to build on top of an open source LLM, a tool that helps me analyze and make recommendations on how to write the best mortgages, you know, in a very specific way. And the cool thing about that, and this ties into, so we had a thesis back in 2017, my partner Gordon started it called coaching networks, which was, um, a poorly branded, but I think correct insight that the way AI will take place in business, business software is as a coach that'll show up and say, Hey, I see that you're about to write this mortgage. Here's all the data you should actually be using. And here's some suggestions on how to do it. It learns on what actually happens. You write the mortgage. You don't, does the person take it or not? Do they pay their loans or not? And then based upon those outcomes, it makes better recommendations to anyone else in that situation in the network. So we call that coaching networks. The reality is copilot is a term that took …

AI assessment note: “I do think that you're going to see a lot of specialized LLMs.”

Q Do you pre-investment think about dilution potential downline and what that does?

A Yeah, we do. So the framework we use internally to figure out, um, if we should do the investment is, it's called what you have to believe. So I was a consultant. I really like frameworks. And so the framework basically means, um, you try to identify what are the three to five things that are specific to this deal that you have to believe for this investment to return the fund. And there's a bunch of things that go into that. Like, if you think, if you unpack that, there's things like dilution. How much additional capital will they have to raise? Will the founder be able to raise that capital as well? There's obviously questions around defensibility. There's questions around market. There's questions around competition. There's questions around team. Like, and all those questions depend on the company. So when we do the analysis, They're always unique to the investment opportunity and to the fund we're investing out of. So when we're doing diligence, what we're trying to do is identify what those three to five, what you have to believe are specific for the company.

AI assessment note: “Yeah, we do. So the framework we use internally to figure out”

Q Do you think that's great? Sorry, I'm interrupting that. Like compare that to Finn, which is Intercom, who actually do it on like outcome based, which is like solution granted.

A The direction that this world moves over time is solution granted. Um, I've spent a bunch of time, um, learning about the Finn approach and that it's hard for now. And the reason it's hard for now is, uh, there are a few reasons. One is, uh, back to the accountability part. It's hard to establish causality In the sense of like, if many support tickets, particularly higher level support tickets have multiple touches, right? Like, so someone, you know, a bot touches it and then maybe a human weighs in a little bit over here. And then how do you establish, you know, who was the winner? You don't want to create an antagonistic relationship with your buyer. If you're like, okay, I did all this. And they're like, no, no, you only did some of this. I'm only going to pay you this. And all of a sudden, instead of having like a monthly, like an easy bill, it's like you're negotiating every month with the customer. That sucks. I think that like over time, we'll start to figure out some of those hiccups and bumps. But I think we're still in kind of early land on outcomes.

AI assessment note: “learning about the Finn approach and that it's hard for now”

Q Given, as you said, that the dominance of still, uh, companies, as you said, with like refrigerators then running their software, do you think we overestimate adoption of AI in the short term?

A My guess is there'll be a little bit of a trough of disillusionment, just like there always is in technology adoption, right? We're sort of, A lot of, you know, everyone's trying everything right now. The good news is there's a lot of movement from experimental budget into real budget in these enterprises. But the bad news is a lot of these companies that aren't actually delivering and doing value are going to get cut. And there'll be some buyers who say this thing didn't work as well as I want. So I'm a little disillusioned. The other thing that could happen is there could be sort of an FTX moment in B to B AI as these agents come out. What I mean by that is these agents are incredibly powerful and they do things for you. They send emails, they buy things, they, you know, they can take action, which is very powerful, but with great power comes great responsibility. And it's very possible. In fact, likely that some big enterprise is going to deploy an agent and the agent's going to do something really bad, right? They're going to send a bunch of emails to customers or prospects that they shouldn't. It's going to buy a bunch of things. It's not hard to imagine what could happen. And there could be a bit of a backlash to say like, oh, wait, This isn't good. We shouldn't do it. And the reality is, like, we do need to figure out the guardrails for these products so that they're dep…

AI assessment note: “there'll be a little bit of a trough of disillusionment, just like there always is”

Q Do you have margin, margin degradation on AI enabled given the fact that you own the full vertical and you have to kind of ingest that all yourself?

A It depends on pricing. So this is a really interesting question. So if you're pricing on, um, labor basis, which is generally how most services are priced today, um, you in some ways are taking the risk upfront. Right. Because you're saying like, okay, it'll chart, if, if it's going to take me, you know, this long, then, you know, I'll pay, I'll, I'll charge you this. But if you, if your AI doesn't work, then you could be in a world where your margins are really degraded upfront. If the AI does work, then you actually capture way more margins. And so you have to be really thoughtful about how you price and you're basically taking a bet on yourself. Like how good is my AI?

AI assessment note: “It depends on pricing. So this is a really interesting question.”

Q What's an example of that and what did you get wrong?

A One of the big examples, like when we look back at, at failures in our portfolio over time, it's been when we thought there was product market fit, but there wasn't. And so there's the term I'm using for now is I'm calling it mirage product market fit, where it's like companies grown really quick. And so you can fool yourself into thinking this company has incredible product market fit, but there's a couple different, um, sort of downside cases where you think you have it and you don't. One of the cases in traditional SaaS is you're selling a product to a very diverse audience. Who's all using it for different things. And so you think like, oh, I've got product market fit, but the reality is like someone's using your thing for this over here and another person, a completely different type of customer is using it for this. And so how do you figure out your go to market motion? How do you figure out your product development motion when all these people want completely different things? That's like what can help companies blow up. There's a second sort of dynamic that's now happening with these AI enabled services companies, where let's say you go out and say, hey, I'm going to be an AI enabled accounting firm. And, um, I'm going to charge you, you know, 15% less than the incumbents and I'm using AI. And so it's going to be even better, higher quality, et cetera. So faster, better…

AI assessment note: “when we look back at failures in our portfolio over time, it's when we thought there was product market fit”

Q I, I often think about Keith Reboys. The best founders don't need you. They're made better, but they don't need you. How do you feel about that?

A I think the reason why a founder chooses a VC is because they believe that you'll help them bend the odds of success on the journey. And that can be in all sorts of ways, right? It could be in helping them with something like go to market. It could be them, you know, helping bring them people to hire. It could be helping give with advice. It could be because you're a great therapist to the founder. There's lots of ways where you can help bend those odds, but I think ultimately that's why someone chooses you because you need a reason, right? You need a reason to be chosen. I agree that in general, if the founder doesn't need to rely on you a lot, then, you know, that's great. But the reality is the odds of success of these things are so low that if you can even bend the odds of success, Incrementally, it matters a lot.

AI assessment note: “I agree that in general, if the founder doesn't need to rely on you”

Q Do you buy that? You know, Sarah Tavel's written before about, you know, paying for the work, not just for the software. Do you buy that? And I'm not disagreeing or agreeing with her, but I'm saying, do you buy that transition? I think a lot of buyers will find it difficult in their minds to justify paying for labor when it is software.

A So what I've seen thus far, and it's still early days, is that most buyers of this stuff aren't firing people. What they're doing is not hiring new people. And so they're trying to be more efficient with whatever they currently have. Like I just invested in a voice AI company in healthcare. And, uh, we talked to a bunch of their customers and the customers, uh, were like, we love this thing. It's amazing. And we're like, okay, great. How many head count did you reduce? And they're like, none. I'm like, wait, why do you love this thing? And he's like, well, I love it because I've grown my business three times with the same head count. And so I think right now, and I think part of it's emotional, people don't want to fire their people, understandably, but I think businesses are able to grow more efficiently than they were in the past because of this stuff. And therefore these, these software vendors should be able to capture some of that leisure labor.

AI assessment note: “these software vendors should be able to capture some of that”

Q What's an example of that and what did you get wrong?

A One of the big examples, like when we look back at, at failures in our portfolio over time, it's been when we thought there was product market fit, but there wasn't. And so there's the term I'm using for now is I'm calling it mirage product market fit, where it's like companies grown really quick. And so you can fool yourself into thinking this company has incredible product market fit, but there's a couple different, um, sort of downside cases where you think you have it and you don't. One of the cases in traditional SaaS is you're selling a product to a very diverse audience. Who's all using it for different things. And so you think like, oh, I've got product market fit, but the reality is like someone's using your thing for this over here and another person, a completely different type of customer is using it for this. And so how do you figure out your go to market motion? How do you figure out your product development motion when all these people want completely different things? That's like what can help companies blow up. There's a second sort of dynamic that's now happening with these AI enabled services companies, where let's say you go out and say, hey, I'm going to be an AI enabled accounting firm. And, um, I'm going to charge you, you know, 15% less than the incumbents and I'm using AI. And so it's going to be even better, higher quality, et cetera. So faster, better…

AI assessment note: “failures in our portfolio over time, it's been when we thought there was product market fit”

Q Penultimate one. Very often, uh, older partners hog carry pools. How does the carry distribution look in the partnership?

A I am so, so grateful for this. Um, so as I mentioned, we grow partners from within. We think it's one of the things that makes us different. Um, part of the reason we're able to retain incredible people like my partners Loti and Yaz and others who have come up behind me and me and Joe and Santi and Kevin, all the people that have been grown within the firm is because our founders made the very generous choice to when they step away from the business and retire to forfeit their carry. And this was, this is not something that's talked about in venture. And I had no idea about this when I was considering which firm to join 11 years ago. But the vast majority of founders of firms, when they retire, they retain a meaningful portion of the ownership of that firm. That creates really bad incentives for the really high performers. Because if you're a really high performer, why would you stay at that place? You'd go start your own thing, right? And that's part of the reason why you've seen such a proliferation of new, new, uh, funds pop up. The amazing thing about our place is I have no reason to leave because the generosity of the founders who stepped down and said, you know what? We want to empower the next generation. We've made enough money. Here's our carry. It means like it's, it's ours to run. And it means I can look you in the eye if I'm recruiting you to be a principal and groo…

AI assessment note: “our founders made the very generous choice to when they step away... forfeit their carry.”

Q Can I be honest? You make much money from a course. Like you do the seed. It gets bought for half a billion, which is great. And I'm not, but is that in stock? Is that in cash? Like so many of these deals that actually behind the scenes, you kind of make three X and it's like not as good as it looks.

A We made, I forgot the specific multiple. It was definitely more than three X in that one. Um, I don't remember the specific number, but I think it was north of five X. The challenge with that is when you have a fund that's of a certain size, even if it's a 10 X, it's still not going to necessarily move the needle. That fund, I believe was fund three, which is the same fund that, uh, has Zoom. Uh, it's the same fund that has some other really large outcomes. And so that fund is already at a, I think it's a 16 times DPI. And so even with a 10 X on the chorus investment, it's not necessarily going to move the needle.

AI assessment note: “It was definitely more than three X in that one.”

Q That makes total sense. How quickly do you know your winners going back to this when you have them?

A So it's not always obvious. Um, and there's a lot of humility, I think, uh, that is important in this industry for lots of reasons, but that's one of them. So bill.com incredible software company been around for a while. We were one of the earliest investors in that company as well. And that was not a straight into the right company. So it grew nicely before the, um, financial crisis, financial crisis happens and business starts to stall a bit. Um, but then we help them figure out the channel partnership strategy for them. Partnering with banks, I think specifically Bank of America was the first one that unlocked, really accelerated that business to figure out like, oh, we can sell through these banks, whereas low ACB products, so doing traditional go to market can be expensive. If we find a channel partner, the whole thing can work. And that business absolutely took off and has been an amazing winner since then. So that's a great example of one that like, wasn't necessarily like this. It's kind of like this, and then this, and then this. That is very possible. And that company made fund one for us. And that it's part of the reason why that fund is so good. Another example that, um, that comes to mind, um, around this humility point is, um, a conversation I had in 2015 with a peer investor at another firm. I remember he came to my office and he said, I just made my career defin…

AI assessment note: “So it's not always obvious. Um, and there's a lot of humility”

Q Can you talk to me about market pull being the most important thing of all for people listening, founders or investors? How do you think about that?

A You want people desperate for your product, and I think, like, that's something that is so overlooked when someone's starting a company. I think particularly when someone's starting a company because they want to start a company, not because they're trying to serve a specific need. You want people who have tried desperately to solve this problem themselves, right? It's not, it's not a desperate problem if someone hasn't, if your buyer hasn't tried to hack together something on their own to solve it, or if they haven't bought an inferior product to solve it, or if they're not spending countless hours themselves dealing with it. Otherwise, it's nice to have. So you need something that is just like, Oh my God, this is a massive problem I need to solve.

AI assessment note: “You want people desperate for your product, and I think, like, that's something”

Q Do you pre-investment think about dilution potential downline and what that does?

A Yeah, we do. So the framework we use internally to figure out, um, if we should do the investment is, it's called what you have to believe. So I was a consultant. I really like frameworks. And so the framework basically means, um, you try to identify what are the three to five things that are specific to this deal that you have to believe for this investment to return the fund. And there's a bunch of things that go into that. Like, if you think, if you unpack that, there's things like dilution. How much additional capital will they have to raise? Will the founder be able to raise that capital as well? There's obviously questions around defensibility. There's questions around market. There's questions around competition. There's questions around team. Like, and all those questions depend on the company. So when we do the analysis, They're always unique to the investment opportunity and to the fund we're investing out of. So when we're doing diligence, what we're trying to do is identify what those three to five, what you have to believe are specific for the company.

AI assessment note: “Yeah, we do. So the framework we use internally to figure out”

Q Do you have margin, margin degradation on AI enabled given the fact that you own the full vertical and you have to kind of ingest that all yourself?

A It depends on pricing. So this is a really interesting question. So if you're pricing on, um, labor basis, which is generally how most services are priced today, um, you in some ways are taking the risk upfront. Right. Because you're saying like, okay, it'll chart, if, if it's going to take me, you know, this long, then, you know, I'll pay, I'll, I'll charge you this. But if you, if your AI doesn't work, then you could be in a world where your margins are really degraded upfront. If the AI does work, then you actually capture way more margins. And so you have to be really thoughtful about how you price and you're basically taking a bet on yourself. Like how good is my AI?

AI assessment note: “It depends on pricing. So if you're pricing on, um, labor basis”

Q Which is like your notions of the world?

A Notions, the ironclads, like the companies that are above a hundred million ARR, growing nicely, like, and still dynamic and young enough to make changes, but like they're not startups anymore. Um, I sort of segment the world into those three kind of buckets, just way oversimplifying. The biggest thing I've changed my mind around in the past 12 months relates to that, to this question, which is I was fearful when the power of LLMs came out that all of, most of the value would accrue to the incumbents because of their data and distribution advantages. What I, um, underappreciated, which is just the recurring lesson of startups, is the value of focus. The reality is like, it doesn't matter how much distribution Salesforce has, how much data they have, have. If you are a startup who's just focused narrowly on solving a very, very specific problem, if you're unified, you know, helping with the go to market stack in much more narrow way than Salesforce is, you're going to run just way, way faster. And customers are going to want your product more. And that's, we're seeing that play out. And so the thing I've changed my mind on is I'm less fearful that incumbents will be able to accrue most of the value. The reality is, like, it's still early days in this game, and things could change, but thus far, the startups are outpacing, the focus startups are outpacing the incumbents. The grow…

AI assessment note: “Notions, the ironclads, like the companies that are above a hundred million ARR”