Zoom to Salesforce: How Emergence Capital Picks Billion-Dollar Winners

29:39Everyone has their debates on if it's going to be as good or not. And having someone on your team as an application layer software company that is staying ears to the ground on the latest developments on the model side has become an imperative for a lot of app player businesses. And so when I think about 2025, like the DNA of an app player company is very different than what a traditional SaaS business used to look like 10 years ago, five years ago, even two years ago. You're seeing more and more job openings for applied research roles, applied scientist roles in a lot of these companies because they're marrying the the infrastructure and capturing data to the end user experience.

30:19And by people using more and more of the software, like Gordon was alluding to with the Delta metric concept, you're now harvesting your own proprietary data set from your existing customer user base and powering a lot of the new improvements to the model experience, whether using an open source model or a closed source model. Would you say SaaS as we know it, traditional SaaS is dead? I would say SaaS is a bit... What does SaaS mean? SaaS means you have a recurring per seat software business model. I don't think those are going away entirely. There are going to be new business models for software.

30:52There's going to be the AI enabled services. We can talk about one in our portfolio that we back that looks more at the outcome based pricing. So SaaS as we know it, is it debt? Probably not. But every SaaS business is going to have AI at its core. Whether they price per seat or they price on outcome is really hyper dependent on the end market they're going after. But I don't think SaaS in its entirety is dead. I'm curious to get your take. No. So, no, here's why. SaaS is the latest incarnation of workflow, right? If you say that workflow is dead, that's really what the other way to say it. Is workflow software dead?

31:34No. And that is because the only way until we are either neural linked, which maybe we will be, but even then there's going to be some prompt that goes into our neural link that says, I should think about that. Okay, and then I'll respond and then that'll get into the overall next version of software that doesn't involve typing. But the bottom line is the only way to both understand information for humans is to have something presented to us. Maybe it's a chatbot. Maybe it's form fields. Maybe it's Neuralink. But something presented to us and then we react to it. And we use what is still a very powerful neural network, which is our own lived experience and our brains to figure out how do we then reply to that query.

32:28That is workflow. Where that goes away, when that goes away and everything is simply bots and we're on the beach or, you know, within the matrix, that's another story. But as long as humans are critical, and we are, to the advancement of these technologies and the advancement of the IP within our companies, that workflow, some form of workflow is going to continue. And that's what, call it SaaS, call it workflow software, it is going to continue because that's where you get those unique habits back, both to and from the users. Well, you've heard it here. SaaS is not dead. It's just evolved. That's great.

33:16Okay, so I want to talk about Together AI. Together AI has raised a tremendous amount of money. And when this first came out, it was a bit of a party round. So from the insider perspective, how has this come to be? Is it still, I don't know, I know the answer. But tell us, Yaz, how is Together AI doing and how do you feel about overall? AI multiples. Yeah. I mean, Together is one of the most foundational companies when it comes to building AI. When you think about the ways in which AI is being used today, there needs to be a core infrastructure layer that's end-to-end. And in order to build models, work with models, whatever it may be, wherever you are in your model development life cycle, some organizations want to own the end-to-end full experience.

34:03Others will like to take a Llama 3, a Llama 4, or whatever Lama is next or some of these other open source models and customize them, run them in their own premise or run them on Together's secure cloud API endpoints, they need someone that abstracts away all of the complexity with working with new models. And so what Together has been able to do is power that experience completely end to end from what used to be and continues to be a hair on fire problem, which is even sourcing the compute to begin with, to then pre-training slash fine-tuning the model, to then running inference on a model such that you have the best and most optimal speeds when you push something into production.

34:45What other approaches have been in this space is to take one thin sliver of this full problem and try to operate as a standalone business. Together's ambitions were a lot greater from the outset, which is why you saw a larger than typical Series A financing and why you saw other people come in, ourselves and some strategics and other financial investors come in and put money into the business because there was this high upfront cost associated with getting the compute necessary to power things like training and inference. And over time, what you're able to do is layer on all the expertise of a world-class research organization that is literally the world's best at this.

35:29If you look at the types of people, Tri Dao, the chief scientist, So it's saying the CTO, obviously a lot of the folks from Chris Ray's lab, given his involvement as a founder of the company, so many of these people from the technical organization on the research and product side are at the world's cutting edge. And they're taking everything that's so complex associated with working with a model and abstracting that complexity away through one API point. And that's the magic behind something like a Together. They'll find you wherever you are in the journey of using these models and powering your own consumer or SaaS application and making it as seamless and as easy and up to date as the next new thing comes out.

36:10So that's what gets us really excited about together. Their growth has been astronomical. I mean, there was a step. I'm forgetting which podcast it is now. And he shared like the zero to 110 months is probably one of the fastest growing infrastructure companies in history. And they've continued to grow at just breakneck speed because they're solving really hard problems for some of the world's most important organization. So Yas, you mentioned something in your description of Together AI as you were laying out how it fits in the foundational model development cycle. So I want to go back to a piece that you wrote.

36:48I think it was like over a year ago now, but you wrote about the foundation model development cycle. Yeah. Looking back, what are your biggest takeaways since writing that and what has changed? Yeah, I think the biggest thing that's different from what we expected when we wrote that piece about a year ago is how fast we get to some of the things that we concluded. I think the major takeaway from that piece is that the best organizations are going to use multiple models in concert with one another to power their end experiences. It would no longer be a world in which you just plug in to open AI's latest model, and they would run really effectively.

37:30And instead, what we thought was that you would leverage a series of different models, whether it be from Anthropic, from Mistral, from Meta, from any of these different model providers, some of your own models that you've developed. In some cases, you'd actually choose to use small language models as opposed to large language models for some of the simpler use cases. So the way in which we thought about where the market is going is developing a core system that allows you to interchange between different models with the right tooling in place to move between them really seamlessly. And now when we talk to application layer providers, we learn that so many of them are tinkering with multiple models.

38:12Actually, no, we learn that Claude is better for this use case than GBT. We actually learned that we can fine tune our own thing here that can reduce our cost basis by 90%. So for these types of tasks, we'll use this. That to me was the thing we were pretty shocked about how quickly we get to a world where multiple models are actually starting to be proven to be best for each different use case. Now the challenge is what's the best way to route between these really, really effectively? And who's going to be the source of truth? Who's going to tell you when it's best to use this model versus another for this specific use case versus the next?

38:49That's where we're really excited to see 2020 to 25 heading, which is being able to build everything together in like a really cohesive way where the end user's experience doesn't change, depending on which model you're using. And it feels on the fly instantaneous where the best thing is used at the best price with the highest performance instantaneously. Because so much has changed in the last year and things are settling, what exactly do you see in terms of the kinds of profiles of founders that you would expect to win in this environment? Yeah, this is one I've been thinking a lot about actually.

39:29And really what gets me excited are teams that are just willing to work their tail end off and seeing what is going to change in this next iteration. It's let me stay as close to the market as possible. And the teams that are honestly doing that, and I know this has been a little bit more of a provocative one, it's like the younger teams are the ones that are growing the fastest right now. They're the ones that are willing to just break things, try new stuff, and sometimes they'll move away from some of the commandments that people learned when building great software companies. They're doing things that don't necessarily scale in the early days.

40:09They're putting things into contract, like opt-out periods, just to get their foot in the door because there's such a goldmine land grab opportunity to leveraging software. some of these younger teams are kind of doing away with a little less discipline on those fronts and saying, let's just go to market really quickly. And you push the risk away from pre-sales and being able to get your foot in the door, actually pushing the risk to post-sales. So the most important thing that we're thinking about for 2025, especially when we work with this type of profile of founders, how can we move into a world where you're making sure your post-sales experience is the best.

40:46Because so many times you'll land within these organizations and maybe you're not yet at the full expand opportunity within the account, but you have your foot in the door. And that customer might also be trying three of your competitors simultaneously. And so now the burden of proving value has really been pushed to the contract's been signed. We're no longer going through 12 months, 18 months enterprise sales cycles. Sales cycle is a month. You're in the door, but we're going to evaluate you for 12 months in a POC or POV. And I'm excited about the people that'll get there, that'll get into the POC, and the ones that'll iterate across those 12 months to make sure that their end customer is going to get the value that they want to out of the product.

41:28And it really is going to fall on having an amazing post-sales, solutions engineering, forward deployed engineering, whatever you want to call it, team that's going to make sure these experiences are great. As much fun as it is to always talk about exponential growth. Gordon, we need to talk about the plateau. We need to talk about preventing the AI plateau. This is a piece that you recently wrote with one of your principals, Wendy Liu. But could you just explain this out? Like, what were the core principles of this piece? And how do you see preventing a plateau shape out? So, first of all, and I've been around this industry for a while, unfortunately, seeing most of these firsthand, but there have been that most of us can remember these plateaus, these moments where there's a series of small factors going on in the world, either open source ideas in the tech space or just unique concepts that people are working on, even in the labs.

42:34and then something brings them all together. So TCPIP as a protocol, I was around, well, I wasn't around when it started, but certainly around when it really became ingrained within the internet broadly. And it was a series of underlying activities that came together in a protocol that opened up the world. And that TCPIP has been compared to crypto in terms of if you could just have made the money that you could make in crypto by TCP IP, it would be far bigger than anything that we've seen. But it was an open source platform that created a series of incredible markets after it. The next example is obviously the browser.

43:20The browser, I do recall when that came into the world. And FTP for file transfers and other technologies like TCP IP all came together to make a human readable version of all of these underlying technologies. And it opened up a whole new avenue of both business acumen and value for consumers and businesses. And then the most recent before ChatGPT was the App Store. The idea when mobile devices, even old flip phones, used to have simple little mobile apps on it. And all this technology, people were trying lots of things, but until Apple created the App Store and a phone that was ready for apps to roll out, all of those ideas then looks like they happened overnight, but they didn't.

44:17It was years in the making. Then ChatGPT comes along November 22, opens our eyes to what we, you know, scientists knew this technology has been around a while, BERT and others. So it's just just very importantly, they took it all together and brought it to the masses of people to understand what's possible. So what the article is about is, all right, are we plateauing after that first wave of AI? And the concept is, and the numbers show it, that the rapid growth where we're like, oh, my goodness, this is mind boggling that ChatGPT can answer these questions or can write a sonnet for us. That's all sort of calmed down because those things haven't dramatically changed since it started.

45:10The models have gotten more specialized or the ChatGPT, for example, has gotten more specialized in things like math and physics and legal. They've gone deeper into certain general areas. But fundamentally, the harvesting of publicly available data has, for the most part, run its course. And Ilya Sutskiver said as much. I love, you know, it's an interesting quote by saying, AI, data is fossil fuel for AI, and we've run out of it. And I generally agree. It doesn't mean there aren't going to be new evolutions, but the next evolution is going to come from sort of two areas. It's going to come from commercial data.

46:00But remember, commercial data sits behind firewalls. It has things called patents. It calls things called intellectual property. The reason all that applies to commercial entities, not to the three of us, is because consumers hand over their IP all day long. We do it on Facebook or on Instagram, on TikTok. And we're not looking for protecting it. We're looking for just giving it away. That isn't going to happen, at least near term and maybe never in the commercial realm. businesses have to differentiate. They have to have something. It doesn't mean everything, but they have to have something that is absolutely core to their future.

46:41And so the next wave, we think, is going to come from companies like Together and RC bringing sort of private model opportunities, whether it's one model or dozens of models, to the fore. And we think that is a whole new opportunity and then what i commented earlier on is also then understanding how you can take advantage of the employees within your business they're inside your ip rights your your protected area how do you learn from them and develop new ip and new value for your ultimate customers and not give that away to the uh open models because if you're handing everything out to open models, you're going to be leaking your IP out the door and that cannot happen.

47:38I was just going to say a great example that we spent a lot of time, we haven't made an investment out of yet is the financial research space. When you talk to a lot of really savvy machine learning and AI experts and you tell them, okay, if you were to go into a domain, what do you want to build? there was no surprise that there were two dozen, if not more, financial research companies. Oh, we're going to help you build a model or we're going to help you mine SEC filings. It's because you have this large corpus of data that's readily available, it's public, and people can train against it. Millions and millions and millions of artifacts worth of data that's publicly available on the internet.

48:15And that's been existed in unstructured and structured document formats. What we're seeing now with these startups is they're running into great value and And then they reach a hurdle, which is we've now mined all SEC filings data. We've run out of things to look at. Now we're going to go crawl social media to see like, are there Twitter feeds, things, podcast transcripts that we can glean some alpha off of? Then there's like that next hurdle. And so businesses that have been in this space for longer, like an AlphaSense, has what I'd call like semi-proprietary data. They moved past the firewall a little bit.

48:48And they're now talking to these aftermarket research, the AMR research firms, like the Bulge Bracket Banks. And they now have like exclusive relationships with the likes of like the big bank, the JP Morgans of the world, what have you. So they have a new repository of data that they can now work with. And so for a startup to compete there, it's become more difficult because there's this rich data asset that they can no longer glean insight from that's powering the AI. Then they go and say like, okay, well, actually you have your own data behind your own firewall. We can actually run our AI in a secure server within your premise that we can now pull data from.

49:24You have startups that are focused exclusively there like Hebbia and some of these others. And so when you look at that, you're seeing like this move towards, oh my God, we're starving for data. How can we find it either through partnerships or through actually getting access to someone's own commercial data that's valuable? The next phase and what gets me really excited is there's richness in the data of the users that is being created as they use your product. The more someone puts in an input that's unique as to how they should look at a certain asset class or a certain type of stock or whatever it is, now you start to gather even more rich data through all the analysts that are using the product.

50:09And so over time, you're building this proprietary data set. You start with some of the stuff that's readily available and move your way up to more and more proprietary data sets over time. It's a rat race, really, to see who can get the most fuel to power the best end experience. And I think financial research is a great example of what we're seeing. It's really interesting that you say readily available because the first, I'd say, swath of data was just scraped data. And so, We started to see some pushback on that and licensing, but how do you see regulatory and access and permission and licensing play into this next evolution of data aggregation?

50:53So this is a big area that we're focused on because when you think about the closed models, the big models out there, their goal is to slowly work their way from what is kind of publicly available. And even then, they've obviously crawled things where folks are saying that you didn't necessarily follow our agreements. But bottom line, publicly available is one thing, but their goal is to really try to get more and more of the actual inner workings of a company. and there are, you know, certain strategies they're using and it's, again, totally fair game that they can work on this. But if I'm a corporation, the last thing I want to have is any of my employees, let's say, talking to an expert network on a phone call and giving information away about the company, right?

51:52So some of them say, we don't want you talking to anybody. Others say you can talk, but you can't talk about any of these areas that could be at all used for nefarious purposes, that is going to be more and more of an issue going forward because these models, one person who has a chance to share information about how something is done that is a piece of either business process or IP, one person, and that is now not just out to somebody who happens to come across it, It's embedded in a model and that can then be accessed by pretty much anyone in the world. So what we're counseling larger customers to focus on is remember how critical it is not to let any of your business processes go out the door.

52:44And right now, the legal ramifications, the legal issues of regulatory issues are not written for how that's going to be done. Net-a-data within a model is okay, but not your data data. What does that mean? Is it HIPAA? You know, HIPAA says that if as long as your data about, you know, something, some drug I'm taking can't be, you know, kind of delivered back to me as a human, as long as it could be someone in this room, that's okay. That's not okay when you're a corporation trying to keep your IP secure and your business process is secure. So it's a lot of work in this area ahead. Super dense.

53:28Is that more of a cybersecurity concern or is that a contractual concern within software? I think it's a contractual. I think that's a really excellent point. Cybersecurity also has risk, right? Because if someone finds their way into this model that you've protected using, you know, together or RC and found their way in, even if they can't get outside your business, you've just, you basically have the keys to everything that company does. That wasn't the case before. Remember when it was just data? You can get access to our data. Well, data is nothing. It's the way your employees are using that data to make value out of it.

54:12Guess what models do? And good models are going to do it even better. And we want to create them. We just want to make sure our companies create them. We want to make sure they're secure inside the four walls of their customers. That just might have made me more bullish on AI lawyers. But maybe it's the big law firms. Going back to the AI plateau, I want to bring up the four key opportunities that you mentioned within the article because I think they're really good points that we should address. So you mentioned four different areas. One, engage experts, which we've talked about a little bit. Two, leverage latent data.

54:54Three, capture in context. And four, secure the secret sauce. Got it. And I think maybe you can touch on secure. Maybe you can talk about both capture in context and latent data. But on secure, I think I already touched on that. These models are your company. They are your business. You're kind of nothing. If you're really bringing AI into your company, everything is held by those models, especially as time goes on and we engage experts within your business. They are the absolute secret sauce. And without it, you really will lose a lot of your IP from there. So how does AI affect pricing models and product features?

55:48Yeah, there's a number of different emerging pricing models that companies are tinkering with. And I'd say we're still in the early innings of what pricing model is going to become the ubiquitous model that people use in the age of AI. What we are seeing is that it depends on the way in which the AI is delivered to the end customer. So, for example, we just invested in a company called Bolt or previously the parent organization called StackBlitz. They launched a product called Bolt.new. They're a unique business that's really exciting in that they've democratized the ability to create web applications.

56:23now through a very slick interface, you in very natural language can build your own website, a web application, or anything that you can really think of. And so it's really democratized creativity on behalf of all people to build, not just coders. This is probably one of the fastest growing software businesses we've ever seen. They've gone zero to 20 million in run rate in two months. And we were lucky to lead their series, be there. What's unique about them as an AI-enabled software business is that it's traditional per seat pricing. And now there's some variability and usage gates on top of that.

57:03But the base layer doesn't look too dissimilar from what your typical PLG self-serve software product look like, like an ocean or an air table. You pay per seat. Once you get to certain usage gates, just from a margins perspective for the software provider, then if you start burning out new credits, there's credit limits that once you hit them, there's now a usage component on top. But that's a good example of how it's still workflow software. It's powered by AI, and this would have never been possible before without some of the new things around AI. But it looks and feels like traditional pricing models.

57:39But on the other end of the spectrum, there are businesses that are going purely off of volume and outcomes. And there is no platform fee or SaaS-based per seat pricing as we've come to know them. Think about companies that are in the customer support space and it's on per ticket resolution or companies that are truly AI-enabled services. There's one in our portfolio, Mechanical Orchard, that's working on legacy to modernization efforts, moving things from on-premise mainframes to the cloud. And that is very much on a project basis. It's once you take one application and move it from an on-prem mainframe, push it to the cloud with the technology platform that they've built for code writing from COBOL to Java to equivalence testing and a lot of other things they've built around the periphery.

58:25They're pricing based off of the outcome of being able to successfully move that application from one place to the next. And so people are now building different models around what the core end use case is. Some of it will look like traditional SaaS. Others will look like outcomes-based or volumes-based pricing. And a lot looks like a hybrid in between. We've really worked our way up to present day, and this has just evolved tremendously even in the last year. Gordon, I want to bring it back to the very beginning of Emergence's founding story. Could you tell us more about how you founded Emergence?

59:02Happy to. I'll try to fast forward. But, you know, the people you meet change your life. And Marc Benioff, my second company that I built, I sold it to IBM and Marc Benioff was looking for a launch partnership with IBM to get his, you know, then fledgling company, Salesforce.com, sort of its first real launch, you know, logo and customer. And we ended up backing them both from a marketing perspective early on uh and from a you know financial perspective really helping helping to to fund some of those early uh marketing objectives and mark benioff i 18 months later he recruited me out of ibm uh to start a new company with him a parallel company to salesforce called software service and it the concept was to be any force.com that we would create any of the new uh software that's that's needed to be created where salesforce had been a single application uh he wanted to and we wanted to create um in effect a a development platform for the cloud we worked on that for uh a while and then mark had to go back in and run salesforce day in, day out.

1:00:22And he had actually become chair and left it to another CEO for a period. Went back in during the really tough time after the bubble burst and took over the reins. And obviously the rest is history for Mark. The software service became the foundation of the Salesforce.com platform. And I loved what we were able to do there. But the question then was do i want to start uh another another sas company since i had heard it from the creator of multi-tenancy about how powerful it is or a venture firm so two amazing partners that i had known in the venture industry uh and i said let's let's consider starting a venture firm that can almost parallel what salesforce was going to be doing as a single company we can take a bet that cloud is going to matter in the long run.

1:01:18So Emergence was always a thesis driven firm. We've always said we're going to have an area that if we're right, we look brilliant. If we're wrong, we took a risk and it was unobvious. And I remember when we started Emergence, the idea of multi-tenancy and cloud software was considered a toy by not only Larry Ellison, because he said that about 100 times about Salesforce, but also by the venture community. It wasn't seen as, why have these long revenue streams that take six, seven, eight years to build any volume up for a company as opposed to licensed software where you get paid right up front.

1:02:00But again, the best opportunities come from not doing what everybody is doing. And we took a bet and it paid off. And so that's the beginnings. and we can touch maybe a little more on how we've worked on things since. Did you raise outside institutional capital back then? How did you actually get the formation together? Good question. We did. We raised$125 million first fund right in the dregs of the, after the big, the major dot-com bubble burst. People met with us, took us 15 months to get to our first close, mostly going to the east coast so we were on planes away from our families for for most weeks of the of the month and uh and you know they would meet with us lps would meet with us just because they wanted to see it's almost like we were a zoo animal we want to see who is trying to raise a tech venture fund in the absolute worst period in in uh in in the history of of uh venture investing to go out and try that.

1:03:07We tried to get, you can hire firms that'll help you fundraise, which now there are many of them and very easy to get access to if you need them. We tried and they turned us down. So we're like, oh, okay, that's not a good sign. But we just kind of got through it and it was a startup. It felt like a startup and And we really are very happy and so appreciative of those early investors that believed in us and got us started. I guess since you started in a dark market, that might inform some of your tenacity to not think about them too strongly and to just keep on going. But yeah, I guess I just I want to know, like, how do you think about navigating these huge market shifts or do they not phase you because of how you started?

1:03:59Look, they owe, you know, I said to our LPs a couple years ago, yes, remember he said, when everyone was asking about AI, it was brand new and happening. And I stood up there and said, you know, by the way, this is pit in the stomach stuff. Like, this is, is this going to be the next great, and yes, it's going to be something absolutely tremendous, but our, you know, we've got to be careful not to overreach, not to kind of take our eye off what we think matters, not to change our model dramatically because just change by definition means you're maybe not, you're not ready to do it properly. You can't like massively force fit an organization overnight.

1:04:40So we've applied the same strategies we did all the way along with, you know, from horizontal cloud was Salesforce, industry cloud was Viva and the growth there, Our enterprise collaboration was Yammer and Zoom. Desk's workforce was Doximity and a number of other companies in the mobile space. So we've been thesis every two or three years. We have a new primary thesis that we all focus on, along with the existing ones. And some roll off the back of the truck. New ones are added to the front of the truck. And we just keep motoring through that. AI was a pretty major disruption in a good way. And we've come up with new theses that we think are going to matter.

1:05:28And we put on a little bit of blinders, not to a fault, but if you're only trying to go with what is moving in the market. Some people say, venture capital, you should never try to predict the future. It's just don't predict the future, just go with what's right in front of you. That doesn't work, I think, to really create outside returns because everyone is seeing what's right in front of them. And it's not that hard to see that. What you need to do is see what's right in front of you and then say no to things. Say no, that isn't going to matter generally, and this is going to matter generally.

1:06:08And to me, that's where you create outsized returns. And you have to have a predictable way of doing that, which we do within the firm. Everyone not only has to be hunting for deals and great opportunities, but we're hunting for theses where we take what we're seeing in the market and raise them up into a thesis. And we beat on that thesis almost as much as we beat on a new investment as to whether it's the right thing to do or not, because they're almost as important as a great new investment because it gives us that ability to focus on a few critical areas. I love that you went into that deeper because I think it's very well regarded that emergence's theses are iconic and they're very thoughtful and they do set kind of a tone for how people are viewing the enterprise markets.

1:07:01And I was curious if they were more reflective or if they were predictive. And it seems like you kind of combine both of that and you just focus on continuously engaging and testing them out, but getting them pen to paper and actually expressing them and living them and then seeing it applied throughout the investments. Yeah. We have a whole process. We actually got together with all of our great seed partners at our seed summit and in this room actually walked through our kind of how we do our thesis development, how we start with hunches. Everybody, including associates, are coming up with hunches, presenting them on Mondays.

1:07:47The hunches then kind of get everybody talking about it. If it coalesces, then we take it to the next level and they present it as a minor theme or we take a little more time and then they go out and talk to a bunch of companies around that theme so it's a it's a an organized process that is again as important in many ways as just being out and seeing what's happening in the world because what's happening in the world informs theses and then theses inform focus which finds the next opportunities that others may not see that's the other fundamental part I always love to look at it if if every venture firm could bring in a company to our boardroom or go and meet with them and we'd all say about the same thing we'd be like hmm what's your CAC what's your burn ratio let me understand oh okay I see where you cut if you're all gonna be the same thing where is the arbitrage there other than we're just hunting to get access and access is important and we all work on that but it can't be the only thing that you're doing.

1:08:50There has to be some perspective on the world that informs the companies you really want to dive in with and spend eight to 10 years building together. And I think one of the key things too, really, is sometimes the thesis isn't going to be fully fleshed out. And the best way to learn is to make an investment. Like one of our more recent ones was around AI-enabled services. And I remember the first time we started talking about it two years was also in this room, also at a seed summit that we hosted here. And we just, I remember, it was you and I, we were sharing it with a bunch of investors. They looked at us as if we were crazy.

1:09:28And then we made an investment in Mechanical Orchard. And now, I think two years later, a lot of other great investors have published content around AI-enabled services. And so, the best way in my mind to really learn is to continuously improve and sharpen your thinking. and the only way to do that is to continue to meet the best founders building in that space, but also being the most important partner to the founders that are building in that space. And so taking the leap, even if your thesis isn't 100 % airtight, is the only way in which you're really truly going to learn where the new opportunities lie.

1:10:03And so that to me is what makes this so fun. You can be as proactive as you want. A lot of times you're going to have to walk it back and relearn what matters, what doesn't, and then you're going to be better for the next one. Shifting into the organizational structure of the firm, from my understanding, it's an equal partnership. So how do you resolve the tension of being an equal partnership with, you know, a power law driven business and only very few deals drive the returns? How does this work internally? How do you resolve the tension? Yeah. And it's a great question. It's, is this is inherently an unbelievably human business.

1:10:46Like partnerships and the money and the difficulty and the risk, the amount of complexity around that I've been quoted as saying it's like a five-way marriage, you know, for five partners. It's five-way marriage where only one of the five has to have a challenge or something goes wrong and it can blow up a firm. I mean, it really is that sensitive. So we take it super seriously. I won't go into all the detail here, but we are in terms of having an outside coach and facilitator every quarter. we get together as a partnership and work through both business issues, but also issues personally as well, because it's that critical.

1:11:33And we've been doing it. It's all sort of getting ahead of things, but it's something that we believe has been really helpful to us really since the earliest days of the firm. So that's a bit of an aside. But the main part question you're asking is partially there's two big things. We work as a team and it's again equal partnership which is critical uh and we develop from within so when jason brian and i started the firm it was you know we all went through those 15 months and and longer to get to a final close and all the pain of that so we were equal from day one then as we brought on other now senior partners, they have developed from day one at this firm.

1:12:19Yaz is developing into an amazing future leader of emergence, but started very early and certainly in his venture career. So when you've come through the firm for many years, you're not as apt to say, oh, you know, I just landed this great deal and I'm not getting, you know, extreme amounts of credit for this. I think I'll just take off and go somewhere else because you've been part of the firm all the way through your career and you're now a full partner within that firm. That is sticky and stickier. You can never, you know, nobody is ultimately sticky in any firm, but we have had great success by basically people come up the stack and senior partners, my two partners have retired and rolled out with sort of no strings attached as they move on to their next endeavors.

1:13:15And I'll be doing the same thing at the appropriate time. Being one of the newer partners that's been at the firm for a few years now, so much of the magic is being able to take a risk. At the end of the day, we're early stage investors. And so much of the stuff, like to your point, we want to invest in things that aren't obvious. And the fact that we approach each investment with a we all mentality, we're all excited about the investment. It's not Yaz's deal or Gordon's deal. It's we all collectively as emergence are really excited about what this company is building. It makes you as a younger partner coming up the stack, feel more confident in taking that risk.

1:13:54If I was over-optimizing for, let me go for the safe ones, then you won't ever find that moonshot. You'll go after the obvious stuff that everyone else at every other firm wants to go and do. And so, when Gordon shares so much about what it takes to find those outliers, you're empowered to do that when the firm is structured in the way ours is. So, that's what I think is truly unique about being like a team first venture firm, which is pretty typical. Well, Yaz, again, congratulations on being one of the newest partners for the firm. We have to ask, what is the most recent risk that you took? Yeah, so we're really excited.

1:14:34It's one I'm actually lucky to be working with Gordon on. Made an investment in a new area we've been spending a lot of energy around, which is what are the next modalities of the future? And when we look at history and look at all the times automation has really powered net new, really amazing experiences for companies and their end consumers, most people have tried to do things that are text-based. It's easier, it's simpler to control. We decided to take a really big leap in voice and are lucky to be partnering with the best end-to-end voice agent platform, Bland AI. So we raised a$40 million Series B, and we're really excited about where they're heading.

1:15:16They're working with some of the world's best enterprises, like the Hertz rental cars, the Cleveland Cavaliers, a lot of other great businesses and healthcare support and what have you. And they're powering the most complex interactions between company and their end customer. And what stands out so much from the founders, Isaiah and Saban, that they continue to harp on is we want to power opinionated experiences that make our customers' customers happy. and so this is the big risk we're taking, going after a horizontal voice platform that wants to be the trusted modality in voice for these experiences.

1:15:55Over time, they'll layer on to all the other modalities, but chasing hard things is our approach and the hard thing is voice. After that, being able to add text and chat and all the other things, that's the easy stuff. Voice is what's really going to matter. It's the most innately human way in which people communicate. And we're lucky to be working with a company that is powering the next era of voice for 2025 and beyond. So exciting. And I think Upfront Ventures, my alma mater, was an early investor in that. Yes, they were. Shout out to Kaser.

1:16:33Okay. So as we close up, I have two final questions and then we'll let you go. So I have to ask with the partnership what was the one deal that you regret doing most one deal that you regret doing most uh regret not doing we can do that we can do that if you want oh yeah i'm gonna i'm gonna uh that was a trick question that was a trick question it is a trick question and it tripped us. No, let me start with the ones that we wish we had done. I think that's a good one. And maybe if we think of one, we'll talk about it. But I know you don't have to do that. That's unfair. And there certainly are more.

1:17:20And I'll add a third, maybe. But Figma was one that we all, back to we work as a team, we all looked at Figma back when their revenue was low, but their engagement was incredibly high and spreading within organizations. And we just couldn't get comfortable. We were a little biased on if we're going to pay that much money, we've got to have some reasonable revenue traction and growth. And we just didn't see through the fact that the classic line now we say, when we call it the Figma effect is engagement is revenue. If people are using it heavily and love it, you might as well just be getting paid for it, even if you're not, because you'll find a way.

1:18:02And in that case, we missed it. Second one was Twilio. Twilio, developer first, you know, kind of model. And we weren't as clear as to the power, the kingmaker or queenmaker of developers was not as obvious to us. And other firms basically had a thesis that was clearer about that buyer. So, or that leverage that developers had. and this was many years ago. So those are two examples. They're more, but I'll just stop there. Fantastic. And to close out, what are you both looking forward to most this year? For me, it's when you think about all of the new use cases that are popping up, all these new applications, I think you alluded to it a little bit earlier during the conversation, which is right now security is gonna matter, trust and privacy is gonna matter, understanding who has what IP, who doesn't is going to matter.

1:19:02In a world where voice feels as authentic as speaking to a human, where talking to a chatbot and they can help solve really complex user queries, there comes a lot of risk with that. So right now we're spending time thinking about what's that next problem that organizations are going to face when they're rolling out some of these technologies. And there is a whole suite of new early stage opportunities that are going to detect malicious behavior on behalf of these orgs when rolling out this new, scary technology. And that's something I'm really excited about. It's finding the guardrails to make sure this doesn't become the complete wild, wild west, and making sure that people can do things in a controlled environment and that doesn't take advantage of end users and consumers broadly.

1:19:51And it wouldn't surprise you that we'll dovetail on each other's theses because for me, it's sort of the human side of that one, which is how are we going to, what is the next generation of software going to look like from a UI perspective or UX perspective where we get, we're so used to form fields and typing everything into a form field, but how much can we both using Bland perhaps to use voice to interact with these systems in a way that's much more natural? Maybe it's Vision Pro from Apple. Maybe it's some other... Apple just talked to me. Oh, no. And, you know, there's going to be new modalities, new ways of interacting.

1:20:37And the key will be we'll be able to help users more if we know more about the kind of actions they're taking and what they're trying to get done. Back to that workflow comment from earlier in the day. So I think there are going to be whole new ways of users interacting with software, that we are looking at a number of really interesting companies that are doing that. And I think one of the things we believe is there's going to be a whole new, if the mobile device was something that opened up a whole new category of software for mobile devices, we've got devices, companies like Bland that may open it up for voice in unique new ways.

1:21:19They're going to probably be whole new categories that the Apples and the IBMs and the Microsofts are going to come up with that will be a whole new platform on which to deploy technology out to end users. And looking at those trends and trying to understand what's coming next is really exciting to us. Fantastic. Well, Yaz, Gordon, I think we could have talked for about four more hours, but I will stop this now. It was a pleasure and thank you so much for taking the time. Thanks, Molly. Thanks, Molly. Really appreciate it. See ya.