Jeff Bussgang of Flybridge Capital joins Nick to discuss Calculating the Market Size for AI, Building an “Experimentation Machine,” the Bull Case for Non-Technical Founders, and the Key to PMF in the AI Age. In this episode we cover:
- Exits and Liquidity in Venture Capital
- The Experimentation Machine and AI Opportunity
- Evaluating AI Businesses and Founders
- Challenges and Opportunities in AI Investing
- Entrepreneurship Education and Startup Ecosystems
- Product-Market Fit and Customer Discovery
- Future of AI and Venture Capital
Guest Links:
The host of The Full Ratchet is Nick Moran of New Stack Ventures, a venture capital firm committed to investing in founders outside of the Bay Area.
Want to keep up to date with The Full Ratchet? Follow us on social.
You can learn more about New Stack Ventures by visiting our LinkedIn and Twitter.
0:18
Jeff Bussgang joins us from Boston. He’s a General Partner at Flybridge Capital, an early-stage VC firm with over $1 billion under management. He’s backed companies including MongoDB, Zest AI, and BrightHire. Before Flybridge, Jeff co-founded Upromise (acquired by Sallie Mae) and held executive roles at Open Market. He teaches at Harvard Business School and has authored three books: Mastering the VC Game, Entering Startupland, and the new bestseller The Experimentation Machine. Jeff, welcome to the show!
0:53
Thanks so much, Nick, great to see you.
0:55
The long awaited interview. I’ve been looking forward to this one for many years, and I’m so glad that we finally have you, and we get to talk about the new book a bit, but before we get to that, can you walk through your backstory and your path to venture?
1:07
Yeah, I mean, Nick, you said it. I But I’ll tell you the story that most people don’t know. I grew up with a dad who was an entrepreneur, a tech entrepreneur in the 60s and 70s. My dad had been a Holocaust survivor who came to the US after the war, somehow found his way into MIT, somehow got his PhD at Harvard and applied math, and somehow had the chutzpah to start a company in the 1960s as a bootstrapped tech entrepreneur working, eventually in the defense industry and communications and electronics industry. So I kind of had this kitchen table MBA. And always had this vision of myself as a tech founder like my dad, and so that led me to major in computer science at Harvard. I actually majored in AI. My honors thesis undergrad was on AI and natural language processing table nets, which Wow, which is crazy. And then I went and got some business experience at BCG, got my MBA at Harvard so I could have some business experience, and then jumped into the world of venture backed startups, which I didn’t know very much about. I knew about startups and entrepreneurship, but not a lot about VC. And I joined this company open market, as you said, which was in 1995 an internet darling. We went public in 96 we had a two and a half billion peak market cap in the in the mid, late 90s, big number for the time, very big, big, I mean, a reasonable number today, and a crazy number at the time, yeah, and, and that just kind of launched my career, and in venture backed tech startups. And after five years there, I co founded you promise, as you noted, and that was another venture backed company. Both of those companies, among others, had Greylock as the lead, one of the lead investors. And so I got to know the Greylock team very well. And there was a young peer of mine who was a friend from business school, who was at Greylock at the time, named chip hazard. And when Greylock shifted West, Chip spun out and left and co founded what is now flybridge with and then I joined him, and another Greylock Boston partner, David Aronoff soon joined as well. So the three of us and another gentleman started flybridge and about 20 years ago, and have been up and running and investing ever since,
3:18
unbelievable. So tell us a bit about the thesis, like, what’s different now versus, you know, when you started about stage and kind of focus,
3:27
you know, it’s so funny. Nick like, everything is different, and yet some things are so similar and timeless. I have this phrase in my book, which I know we’ll talk about, timeless methods, timely tools. And I think venture capital investing is like, you know, timeless principles, but timely application of those principles. And so what’s different? When we started in the business, there were a few 100 active firms, a few dozen on the East Coast. New York was invisible as a market, and we had a thesis that East Coast investing, old fashioned seed investing centered in MIT and Harvard, and then also yes, that New York would be in this incredible ecosystem. And that’s really played out. We led the series A and MongoDB and helped ride that journey. My partnership is still on the board there, 12 years or so later, we did number of other amazing companies in New York, like Code Academy and NS one and cloud, better cloud. And so we’ve had this, this other contrarian thesis, you can build enterprise software companies in New York City. So that was a contrarian thesis that we, we hadn’t pursued and followed. You can back MBAs. That was a contrarian thesis, not just technical founders, but MBA founders, and that’s played out very well for us. And you can stay small as a niche firm, which we’ve done for 20 years. We are three general partners our fund sizes, you know, even though we have over a billion under management, that’s across seven funds, our fund sizes are 100 plus million. So we’ve. We’ve kind of done some things differently, I guess, than many, and we’ve done, you know, pretty well with it. I love it. So,
5:06
you know, before we jump in here, you’ve been at this for a while, and most people I have had on the show have not, you know. So we focus a lot on entry, we focus a lot on investment, but we we don’t talk a ton about exits, how to manufacture them, how to think about them, how to coach entrepreneurs, when to exit, right? Like I mentioned, your partner is still on the board of MongoDB, you know, after all this time and you’ve had companies go public. Give us maybe some quick, you know, thoughts and advice, you know how a VC at the seed stage in particular should think about exits and think about, you know, portfolio getting some some harvest and some liquidity on that portfolio. So
5:54
there’s a view in our world that you got to ride your winners power law business are only going to be one or two massive winners each fund, and you should ride them. And I think that view is right, but also wrong. And it’s wrong because you’ll never generate returns, and if you never generate dpi, you’ll never get to fund four, fund five, fund six and fund seven, like we have. I think it’s also wrong because I’ve seen unicorns evaporate. I’ve had two IPO companies go to zero, effectively. I’ve had another unicorn, multi billion dollar unicorn on paper, go to zero. And if I had sold along the way, I would have been much happier. And so generally speaking, we have a belief that you should look for liquidity windows. They’re very rare, but look for them. And when you find them, hit them and hit them with even 5% 10% transactions of your position to generate some liquidity along the way, so that even if you have a winner like MongoDB, which for us was a 60 times return winner, we still sold in a secondary pre IPO. Now, in retrospect, there was another 10x to gain from that position that we sold, I think we or maybe it was an 8x to gain. We maybe sold it sort of seven or 8x and then we had another seven or 8x to go. But, but if you sell 10% of your position, you maintain 90% of position, your position, you don’t regret it. And if that thing went to zero, we would obviously, you know, still have that strong liquidity that we had generated. So we have a general philosophy of liquidity windows are rare when they when they’re there, you should hit them. You shouldn’t just, you know, hold forever. Awesome. Super helpful.
7:39
So I think most in the audience know, mastering the VC game. It’s sort of off recommend, oft recommended in the VC industry, but the new one is the experimentation machine. Why did you write it? Jeff, though
7:53
I was, you know, I have these two hats that you described. I’ve been teaching at Harvard Business School for the last 15 years, and which has been incredibly fulfilling. I’ve had 2500 students during that period of time, which is kind of crazy. And I was seeing our students, like many students, be really on the cutting edge of leveraging AI and their founding journey, and then at the same time as an AI focused early stage VC fund, I’m working with AI native founders. And I was observing that these two populations were operating in a very different way than founders had a few years ago. They were effectively operating in what I term as 10x founders. So like the mythical 10x developer, who’s not 10% or 20% better than the average developer, they’re 10 times more productive. These 10x founders, both in my portfolio and in my classroom, were leveraging AI tools to be builders and to make progress on startup experiments so much faster, so much more effectively, so much more efficiently. And I felt like there were some lessons there that maybe I could synthesize and share with people, but do it in a way that it wasn’t just about the whizzy new AI tools, which we’re all getting exposed to in our it feels like it’s in the water. Your inbound email is just full of all the whizzy new AI tools that are out there, but also harmonize and connect those AI whizzy tools with timeless techniques for Product Market Fit and enduring value creation. So that was the, the idea behind the book. So it’s, you know, the experimentation machine, as you said, it’s, it’s trying to codify these Timeless Lessons of product market fit, customer discovery, finding pain, getting those value propositions right, doing your go to market experiments, but doing it in the context of these turbocharged AI tools.
9:41
Well, do me a favor, Jeff, you and I had a bet before the show, and you won the bet, and somehow, now I owe you lunch, but if you could bring a signed copy of the book, you know, and make me feel a little bit better,
9:52
it’ll be a good exchange. You got it all
9:55
right? Let’s do this. Let’s do that. So, so Jeff, talk to us about. About the size of the AI opportunity as you see it. You know? How do you how do you calculate it?
10:04
So look, everybody knows the SAS business is in the US, about 260 billion globally, just under 700 800 billion. So that’s sort of the obvious thing, which is that all software is going to be aI native, AI enabled. And some of the incumbents who have done very well in the SaaS industry are going to be disrupted. Other incumbents are going to make the transition and really lead the lead the pack. That’s the baseline. What we are seeing is something totally different, though, because we’re now going after not just that three or four percentage points of operating expenses, that 260 billion we’re going after the entire budget, staffing, labor, process, services, the whole thing. And so we see the opportunity to be a multi trillion dollar opportunity. We did an analysis where we sized the market as 16 times the SaaS opportunity to just go after the payroll portion. That’s automatable by AI. And if you think about it, it’s pretty common sense, like, obviously software engineering is one of the first areas that’s being automated. And so AI for CO generation, very clear rules and and paradigms for CO generation, a very forward thinking, early adopter market and a very technically sophisticated market. Well that’s resulted in massive dislocation in engineering talent. You know, engineering hiring is is at a low for early, early state, for entry level engineers. So next you go to customer service again, clear rules, a clear space where AI is going, just like in coding, there’s dozens and dozens of startups in that space. And customer service, there are dozens and dozens of startups, and so those startups are all going after staffing budgets, not just software budgets. So
11:52
let me play devil’s advocate for a second, market is a function of price and volume, and one tenet of disruption is you get a much bigger market because you have access, like you said, you don’t have access just to software. Maybe you have access to services now as well, and and labor and a bunch of other things. But function of disruption is price dropping substantially, right? So can we not? You know, as we think about this market, the summation of all these Won’t it be, like a fraction of each of these put together? Because, you know, the price is going to exponentially drop.
12:29
Yeah. I mean, what you’re pointing to, it’s known as Jensen’s paradox, is that as pricing drops, demand expands, and we’ve seen this in the cloud computing world dramatically. I call it the cloud dividend, where this radical reduction in cost of cloud and cost of storage and cost of compute has resulted in a massive growth of the market. When we first invested in MongoDB, we sized the database software market at the time as an $8 billion market. Today, the database market has grown to be nearly $100 billion market because you know how in it, how efficient database software is, the rise of data, the the reduction of cost and compute and storage. And now suddenly database market is dramatically, dramatically expanded. Similarly, in the AI market, we’re in the middle of an AI dividend, where computational infrastructure, you know, the Nvidia’s of the world, foundational models, the open AIS, the anthropics, the Google Geminis, those prices, the price per token, the price per AI capability, is dramatically dropping. And so we’re seeing expanded usage and expanded demand. And so even though you’ve got this price compression, you’ve got a growth in the market. So yeah, I think what, I think what, I think what you’re pointing to is we’re going to see potentially, an explosion of capabilities and maybe ideas that we haven’t even thought about Price Waterhouse, when they did their analysis of the AI market, they put it at 16,000,000,000,009 of that trillion was based on some of the things I’m talking about, which are sort of substituting existing labor and processes, but 7 trillion was new capabilities of the things you’re describing, when, if you have this radical reduction in cost, suddenly usage and capabilities go up
14:10
amazing so, so in the book, you emphasize the importance of transforming startups into AI powered experimentation machines. And you talked about that a bit with some of the founders that you had observed that we’re building, Jeff, how is AI changing the way that these startups are being built, and what’s not changing about the process? So
14:33
first, I find that a lot of founders, they say to me, you know, I’m not an AI company. So is it really for me? Do I really need to embrace AI throughout my entire organization? And I point to the example of top line pro in the book, and I’ll describe it here. This is a company that makes websites and does CRM software for Service Pros. So think landscaper, plumbers, electricians. These are the most non technical professionals you can imagine. These guys. Don’t have desktop computers, and they’re not sitting at their desk building spreadsheets and building websites. They’re on a roof fixing a leak, and so those those pros are being serviced by an AI bot that top line Pro provides, which builds their website for them, manages their CRM for them, handles all of their billing and operations and back office for them, handles marketing for them, scheduling like this series of bots that top line Pro has created, and top line pro itself is building an AI native company so that even its outbound marketing is AI driven. It’s they’re using tools like hey Gen and 11 labs for customized, personalized outbound video generation. They’re using all sorts of tools for personalized demo. They created a chat bot for customer service so the Pro can text a photo of a job they’re in the middle of when they’re underneath the sink fixing a leak, and that job suddenly goes up on the website. So it’s a it’s a really sort of mindset that founders, even non AI software company founders, should have about imbuing AI throughout all of their processes, everything they’re doing. Hiring, you mentioned bright hires, one of my companies in the hiring space, hiring, customer service, sales, marketing, communications, of course, engineering, all the things that you do, even the finance function, you should have aI infused in every, every part of your workflows.
16:29
How much should a starter, a startup take on and try and build in a proprietary way versus, you know, find something off the shelf?
16:36
It’s a great question, and I am generally of the view that startups need to be really good at selecting and implementing new tools and having the flexibility to be always evaluating what other tools are out there and bringing them in. So I’m a big believer in use the off the shelf tools that’s there. There’s so many companies and so many tool providers and existing companies, you know, HubSpot is infusing a lot of great AI into their tool. Notion. I have a founder who I did an HBs case on in my class. He says that, look, I run my, my entire company on notion because of the AI tools. It’s, it’s so incredibly efficient. He exposed his notion to his investors and said, Look, here’s all, it’s like notebook, LM, like, here’s all the documentation about the company. Just put it in this sort of, you know, firewalled environment, and then go investors as part of their due diligence process. Go ahead and look at everything you want. You can look at all of my, you know, segment data and all of my customer calls and all of my internal analytics. So I think there’s this idea of being good at identifying tools, bringing tools in, using those tools throughout the organization, even a five person, 10 person organization, but always be on the lookout for the next cool tool, and always be willing to experiment.
17:53
How do you think about infusing AI into your operations, which is kind of what we’re talking about here your operation stack, versus infusing it into your product. Like, all your customer facing, facing, you know, the product side of the business, like, you need to be aI first you mentioned the roofing example, you know, like, does there need to be an application that the customer is interfacing with, and whether they know AI is behind it or not? You know, it’s, it’s got that magic feel like, like software did 10 years ago. So any thoughts on that? Yeah,
18:26
look, you know the there’s this new trend, the tiny teams trend, and the notion of seed Strapping, where founders can do so much more with less that you can imagine companies. And there’s this website now, tiny teams and a couple of our portfolio companies are on it, where they’re celebrating because they’re at a million of AR per employee. I did this rocket ship list that I do every year with the hot non public startups that are great places to join. And there are 12 companies that are over a billion dollars in value with under 100 employees. So this idea of infusing it in all your operations, it creates a it requires a mindset of, Do I need a human to do this job? Or can I actually use an AI Toby, the CEO and founder of Shopify, wrote this manifesto that he published on X but he sent it originally in an email to all his staff at Shopify, and he said, you know, we’re going to be an AI native company. AI is now required. Everything we do is going to be aI focused. And what that means is, if you come to your manager asking for headcount, the first question you need to be able to answer is, why can’t you use AI to do that function? And so that’s, I think, the mindset for internal operations is, how can I scale without growing? That’s the new entrepreneurial mindset.
19:49
Have you seen examples of true agents? You know are the agents here yet? Because we keep talking about the agents, but I’m not quite seeing them yet in. Even, you know, my own team, we’re starting to stitch together a bunch of, you know, AI powered workflows. And so you could call that an agent, you know, something might be in chat GPT, but then we have APIs to our CRM, and then we have some RPA with Zapier, and then we have some, you know, Deal Memo creation. And so you put them all together, that’s really an agent, but it’s, you know, it’s one thing to have a chicken wired solution. It’s another thing to have, like a true agent off the shelf.
20:30
Yeah, and you’re totally right. Nick agentic AI is still in the super early innings, and the best and most and most advanced companies in our portfolio are starting to do it. Coding agents is one of the first areas our company, blitzy ai has 3000 coding agents that work alongside 12 employees to generate millions of lines of code to help refactor old code bases, so mainframe migrations and cobalt and Fortran code that needs to be modernized or upgraded. So yeah, they’ve got 3000 agents working for them, and those agents all are very focused, very narrowly tasked, and they’re doing real work. We have another portfolio company in the customer service space called melody arc. That’s another area, as I noted, where agentic AI is really happening, and decagon and others are doing some great work in that space. So, yeah, agentic AI is just beginning, and it’s just beginning in a couple of these early adopter markets. One of the challenges is that the tooling is so early to build agents to, you know, create agents on your own, to create the boundaries for them, the objectives, to evaluate them, to authenticate them and secure them. And so we’re actually investing at flybridge in a lot of agentic AI infrastructure. We recently invested in a company called arcade, which was founded by a former entrepreneur who we had backed, who sold his company to Okta, stayed at Okta, learned a lot about authentication, and has built a new company and a new platform for agentic authentication and securitization, for for agents. And so I think a lot of enterprises are nervous about bringing agents and independent actors who have power and agency literally into their enterprise for security and obvious other reasons, but it’s, it’s coming. So
22:22
we’ve, we’ve talked now about the 10x founder. You mentioned that at the top, we’ve talked a lot about these AI native startups, you know, whether they’re at the application layer or or otherwise. You know, talk to us about evaluation of prospective investments. You know, how do you identify the 10x founder? And how are you really evaluating these, these AI businesses?
22:47
So it’s a super hard question. On the people side. We used to have as a criteria at flybridge, we needed to see a deeply technical founder, typically a PhD in AI and machine learning was our sort of standard background for our machine learning and AI founders and portfolio companies from 1015, years ago on, but recently it’s it’s kind of a moment, and I’m a little contrarian in my view. I think on this, but I think it’s a moment where it’s the revenge of the MBA, where we can back business founders who are builders, who have become self taught, AI native founders and builders, and can achieve a lot, including an MVP or a version 0.9 and get going without a technical founder. They can provide, they can they can create the build capacity on their own using these modern tools. In fact, I’m teaching at HBS my students to be aI native builders and to be that 10x founder. And again, Revenge of the MBA like you don’t need to run to MIT to get a technical co founder. You can build that MVP yourself. Just grab lovable or replid or some other tool and build. And so the founders we’re looking for now it’s less about PhDs and AI and it’s more about, do they have an insight into the business problem, you know, the earned secret, which some of us in the industry, you know, talk a lot about, do they have a really deep understanding from months and months or years and years of customer discovery, some nuanced insight that nobody else has about an opportunity and a problem. So that’s sort of on the founder side. I can talk about the business model and application area, but I’ll just pause there
24:31
perfect and then give us a sense for you know, the evaluation criteria you’re using on on the AI side, on the business side, yeah.
24:39
So on the business side, we tend to look at three things. There’s there’s this common refrain, and I agree with it, that there’s lowered barriers to entry, and it’s easier and easier to build, and so the competitive moats are harder to find. But there are three things that seem to be relatively enduring, and. I might even add a fourth. The first is proprietary data, making sure that you’ve got access, as a company to proprietary data that is not available to the foundation models that are out there, the public foundation models or the open source models. So for example, we have a legal AI startup in our portfolio called noetica. And noetica ingests 1000s and 1000s of pages of legal contracts that their law firm customers put into their model and train. And those legal contracts are corporate debt financings, and so now they know all the terms of every corporate debt transaction ever done by that law firm. And so now they’re able to generate new contracts and do terms benchmarking and best practices super efficiently. Chatgpt is never going to have access to those contracts that sit behind the big law firm. So that’s a proprietary data set that they can train their model and have that model be purpose built in something that’s highly valuable. The second thing we look for is some sense of being a system of record, having that workflow that allows the business user to really live in that application. And my example for that would be one of our companies called allspice, which is the GitHub for hardware engineers. It’s a collaboration platform for the hardware designers, and using AI to help, of course, recommend hardware design and find flaws and help make connections. The customers of all spice live all day in that platform. They are literally addicted to the workflow of hardware design, just like they’re addicted to CADCAM workflows. So that feels like it has that power and potential to be very sticky. And then the third and final thing is insight into that human AI interface. AI is so powerful, but it’s so complicated for many end users, and so companies and founders that have a really strong product sensibility, design sensibility, and an ability to manifest and that AI capability and really nail that human AI interface. So the fourth thing that we’ve been spending more time on, as well is some distribution edge. Now, in an early stage company, that’s harder to discern because you’re projecting, will they develop a distribution edge because of the nature of the founder or the nature of the model? But eventually, founders have to obsess over distribution, and they have to really nail it in an environment that’s more and more cluttered. And so when we talk to these early stage founders, part of what we’re asking ourselves is, okay, I get what they’re trying to build, and I think it’s really cool, and I think it’s proprietary, and I think it’s gonna have great workflows and be a great system of record, but do they have an edge in distribution? And so that’s another criteria that we’re evaluating. What would an example
27:42
of that be? The distribution
27:43
edge, so channel partnerships or something else? Yeah, so
27:49
one of our portfolio companies is open FX, which is a company that’s applying trading techniques in the crypto and stablecoin world to help with FX and allow fintechs to do better and more FX trading and more efficient transactions. Well, the founder from that company was previously the founder of our portfolio company Falcon X. And Falcon x is this institutional crypto trading platform that sought liquidity and provided a trading of the prime brokerage platform. It’s a very successful company, $8 billion value in the most financing of years ago. That founder, therefore at open FX, he has a clear distribution edge, because he knows where all the pools of liquidity are, and he knows all the fintechs and firms that care about on ramps and off ramps in the stable coin ecosystem. It’s like earn trade secret. He’s got that earned trade secret, but he also has the relationships and credibility to instantly, you know, establish those those relationships. So not all young founders, or first time founders are going to have that type of a distribution edge, but maybe they’ll have a different edge that they can convince us of. What’s the biggest blind
29:04
spot with MBAs? You’ve brought them up a couple times now, you know, Revenge of the NBA, and it could be their time. You know, with a lot of these AI enabled AI enabled tools, but what? Where do they most often go wrong? What I
29:18
find the ones that don’t nail it, they stay at too high level, and they don’t do the gritty you know, when I my first job out of HPs was as a product manager. One of my first meetings, I was handed a stack of paper six or seven inches tall, and each page of that paper was a bug report. And I had to go through each of those bug reports and triage with my engineering lead, which ones are we going to fix before the next release and which ones are we going to punt? And just the gritty, the detailed, the grind of going through 1000s and 1000s of bugs, the gritty, detailed grind of interviewing 150 customers to get to the value process. Position that makes most sense the cold calling and outreach of 200 300 400 prospects to get to the 10 that you’re eventually going to close. The grind of reaching out to 100 investors to get to the two or three that are going to lead your seed round. It’s a grind. It’s hard work. There’s no shortcuts, and a lot of MBAs look for those shortcuts, and the ones that do don’t succeed well
30:25
and and to your point earlier, the ones that really get deep and really get specific and spend the time with the key stakeholders usually have the best earned it earned insights, right? Because they’ve seen it firsthand. And it’s not this surface level problem in solution, it’s the true workflow that’s going to find the wedge, get the traction, and has potential to break out.
30:50
Yeah, one of my portfolio companies is h2 okay, and it’s a Harvard dropout, actually, Annie Lou Who started the company. And Annie the company is providing AI and machine learning for fluids to and process engineering to allow sensors to determine quality of fluids in manufacturing environments. So think in a Unilever plant, the quality of the milk in a Ben and Jerry’s plant, or the quality of the beer in a Anheuser Busch plant, or a PepsiCo, you know, plant, the quality of the soda and making sure that the the water is clean. And every meeting that I’m in with Annie, she’s coming from the shop floor like, you know, hard hat on, flying to some plant somewhere in the Midwest or somewhere in Europe, somewhere in Asia, and she’s just walking the shop floor, talking to the her end user, and just making sure that the software is doing the thing it needs to do, and then use and the end users are happy with it. And you know, Annie is 23 years old. She dropped out of college. She did not have a 10 year experience of selling to manufacturing environments, but she’s nailed these huge companies and these massive contracts because she’s done the greedy work of going to their manufacturing plants, putting on a hard hat, putting on the boots, going in the shop floor, and making it all work amazing.
32:07
So this, this is pretty relevant for the next question, but talk to us about product market fit. You know, in this age, how is that different? Do you have any advice or thoughts on, you know, how founders should find it and validate PMF, yeah.
32:22
Look, I talk about this in the book a lot as the main focus of the book is the experiments to find product market fit and the metrics to determine where you are in the product market fit journey. I don’t think it’s radically different today than it was five years ago, 10 years ago, 15 years ago. Maybe what is different is that you can run these experiments and get this data faster through AI. You can spin up MVPs, you can spin up prototypes. You can put customer discovery interviews into notebook, LM and query the AI about insights, about what you heard from customers. You can have sales calls that you’re recording on Zoom or on Gong, and bring those back into an AI. You can have your AI create a synthetic persona and query, and have your engineers query about different features and different trade offs. So you can do a lot of things using the AI tools, but ultimately you’ve got to do the hard work of finding a hair on fire value proposition, building a product that customers love, getting them to use that product and use it repeatedly over and over and over again, and doing it in a way that you can reach customers in a profitable fashion. So
33:33
when in your class, do you run like sort of an accelerator that may have similarities to, I don’t know YC or something else where you’re, you know, encouraging your students to get out in the wild and try and sell and try and observe, or, you know, what are some of the hallmarks of kind of your approach there?
33:52
Yeah, I’ll
33:54
speak about HBs in general, rather than my class in particular. But we do have a series of programs that not to try to compete with YC, but rather to be an environment where entrepreneurial students can come in and really accelerate. We have a startup bootcamp that occurs in the first year, where they learn fundamentals of customer discovery and value proposition identification, research driven innovation. We then, in the second year, have a startup operations class where they can be very hands on. We have a founder’s journey class where they learn how to go through the mindset of the founder, the fundraising et cetera. And then my class, which gives them a deep dive into the zero to one journey and running the experiments throughout every step of the process, from early customer discovery and value prop through go to market through business model experiments. And so we have entrepreneurship is the largest unit at the school. We have the most seat miles in terms of class discussion. HBS, for those who may have thought of it as a school that produces hedge fund managers, that’s yesterday’s your that’s your five. There’s HBS. Today’s HBs is an entrepreneurship school at its core.
35:04
So you know, there’s, there’s famous examples like Larry Ellison and Fred Smith and Phil Knight that you know, went through startup competitions in their university and got terrible grades on them, right? And so classically, we’re told, if you’re gonna do a startup, don’t do it within a university context. It sounds like, you know, things have modernized quite a bit, but should we be doing startups? You know, within a higher education context?
35:30
Look, I think it’s an and not an or some people, for them, an MBA is incredibly valuable. It’s a lifelong learning opportunity. It’s an incredible community. And in the top business schools like Harvard and Stanford and MIT and others, it’s it’s just a an amazing, transformational experience for everyone and and you can have an incredible entrepreneurship education. And so the two founders of CloudFlare, which is a $30 billion public you know company, Michelle zatlan, and Matthew Prince, they basically won every HBs pitch competition that existed at the time when they started the company out of HBs 1215, years ago. And so yeah, there are plenty of great success stories of launching in the university context. The HubSpot founders were out of MIT that worked out pretty well for them. That’s a $30 billion market cap company. And at the same time, there are plenty of people who can be very successful starting companies outside of these elite institutions. So to me, it’s not one or the other. It’s just depends on what path is the right path for you.
36:33
So Jeff, before we leave the the AI discussion, talk to us a bit about how one avoids the model moat trap, you know, how do you ensure that your portfolio companies maintain defensible positions amidst, you know, these, these giant players? Yeah, it’s
36:51
a great question. There’s this running joke, of course, you don’t want to back GPT rappers, and you don’t want this moment where every new release open AI kills your startup. And I think these areas I mentioned about proprietary data sets, workflows that really create this system of record dynamic within a company, and the human AI interface are really the things that we look at. But we also tell our portfolio companies like, the bar keeps rising, these platforms keep getting better. So you want to be in a position where, as the platforms get better, your solution gets better. It’s not eating away at your value, but enhancing your value. And our best founders are really figuring that out and figuring out how to architect their platforms to benefit from where the puck is going with these incredibly powerful foundation models, as opposed to seeing those models eat into their value prop.
37:38
I know that open source models allow you to sequester your data and build your own models, right? Whereas open or, yeah, closed source, you know, the open AIS of the world arguably get access to everything you’ve got. Do you encourage some of your, you know, vertical AI businesses that have proprietary data to use open source as as a shield, or is that not part of kind of your your guidance,
38:09
it’s a big part of our thesis. MongoDB was an open source database software company. We love open source. We’re investors in RC, which is a very important company in the world of open source AI models, we think open source is a great approach, but most of our companies, honestly, Nick, are doing multi model so they’re building their applications in a way that are flexible, and are choosing the model that’s best for the use case. And then as the models improve which they are every week and every month and every quarter, they’re allowing those improvements to adjust which models they use for which problem set. So Claude may be a little bit better for CO generation. Open AI comes out with a new image generation tool. Well, let’s make sure we leverage that. You know, you know, Gemini is doing some amazing things in writing. So you just sort of see our companies use this multi model approach, and have that flexibility embedded in their application architecture. So Jeff, that
39:05
the data suggests that a lot of deal making has shifted back to the Bay Area, at least in 2024 you know, we’re, we’re kind of in a flight to quality moment, so to speak. You know, with the advances in AI that we’ve we’ve spoken about so far, and that you write about so much in the book, will we see more net value created inside or outside of the Bay Area?
39:27
First of all, I don’t know what data you’re looking at, but I think that’s wrong. I think, oh yeah. I think that New York has risen quite dramatically in terms of companies, in venture backed companies, and certainly AI companies. And in particular, New York is on the rise for applied AI, the last mile, the customer facing, whether it’s legal, accounting, customer service, financial services, real estate, proptech, construction, New York is booming right now in AI, and I would say Boston is also. Are doing just a huge amount of activity in AI. I’m a little disappointed, frankly, in Boston, in that, if you look at the top 100 top 200 AI companies on any listing, their, their, yes, their preponderance are in the Bay Area, but there’s a good chunk in New York, and there’s not as many in Boston as I would like to see. But we’re working on that, but, but, no, I think, I think, to me, it’s less about like, where’s the value going to accrue and wherever, and do we have a healthy ecosystem? So obviously, the Bay Area is the gold standard for the greatest and most amazing ecosystem in in the history of startups, and that’s been true for decades and decades. I would say New York is clearly a strong number two and a very healthy ecosystem. And I would say Boston is a sort of, sort of healthy, although has a little bit of a fever, you know, is coughing a little bit, has a few bumps ecosystem. And the way to make it healthier is to retain the talent. Because all the talent is there. They’re just there for two years or four years at Harvard and MIT and northeastern and elsewhere. And then they go to New York in the Bay Area, too high frequency. Are you
41:04
seeing most of the talent focused on tech? You mentioned that before at Harvard, you know there, there used to be the banking tracks and sort of the path to hedge funds. Are the majority of the students, you know, enrolling in more of the entrepreneurial and in the chat the tech career path
41:19
one and on the faculty told me this week that in 2025 this last year, we had 43% of the seat miles, meaning 43% of all second year student course work was in the entrepreneurship unit. So yeah, we have a massive market share. It’s the largest unit, over 50 faculty members. When I was there 30 years ago, there were two faculty members in entrepreneurship. So it’s a pretty dramatic change. You
41:44
know, you fly bridge as a firm has over a billion under management. Why? Why stay at the seed stage, right? Why, why not expand to series A you’ve got a lot of experience now and a lot of access to capital?
41:56
Yeah, we just love zero to one. We love staying small. We love staying focused. And there’s just so many great firms that Series A, Series B, multi stage firms that come into the series a world. We just rather be great at that first, what we think of as the most inefficient part of the market, and then help our companies with the more efficient parts of the market. From there on. Jeff, if
42:19
we, if we can feature anyone here on the show. Who do you think we should interview, and what topic would you like to hear them speak about? It’s a great question. Can it be a founder? Can be a founder? It can be an investor, someone in tech or not. I would love
42:31
to have you interview Alex Karp of Palantir, and I’d love for you to talk to him not about his book, not about his sort of contrarian views on defense tech, but rather enterprise sales. 101, how did he build such a fast growing enterprise selling machine? Because so many of our founders, as I said, are great at product and have this market vision, but don’t nail distribution. Don’t nail the the growth curve and the enterprise sales and commercial capacity building. And he seems to have totally nailed it.
43:09
Awesome. Jeff, what book, article or video would you recommend to listeners? Quite a few. Well, you got quite a few behind you, and you’ve got one on the table in front of
43:19
you. Yeah, yeah, right. You know, behind me is a biography of Muhammad Ali, written by Jonathan Ike. I bet very few of your listeners have read this. Muhammad Ali was an elite athlete. Was an exceptional outlier, was was the kind of person you would want to back as an entrepreneur. He was extremely driven. He was a little crazy, but he was also a civil rights leader, and he there was a part and an amazing friend and amazing human. And there’s a part of that humanity that I didn’t fully appreciate until I read the biography.
43:52
Awesome, Jeff. Do you have any habits, tactics or behaviors that are a force multiplier,
43:57
but a believer in the quantified self, and so I do a number of things that are very quirky with regard to maintaining logs and tracking my behaviors to achieve my objectives. And so every year, I’ll put a set of objectives that I have for the year, and then I’ll sort of develop the business plan and the metrics and analytics to back that objective and create a system that allows me to monitor and maintain and be accountable for those objectives. So I sort of treat life management area, my area of life management, similar to what I might treat business management and company management. It’s a little quirky, a little a little wonky, I’ll admit, but it seems to work reasonably well.
44:41
That’s great. And then finally here, Jeff, what’s the best way for listeners to connect with you and follow along with flybridge?
44:47
Yeah, LinkedIn is probably the best way. And subscribe to my newsletter. Bus gangs, bullets and experimentation machine.
44:55
He is Jeff busgang. The firm is flybridge, and the book is the experimentation. Your machine. Jeff, thank you so much. This is long overdue, and look forward to doing it again.
45:04
Thanks, Nick, look forward to having lunch with you. Thank you, sir.
45:13
All right, that’ll wrap up today’s interview. If you enjoyed the episode or a previous one, let the guests know about it, share your thoughts on social or shoot them an email. Let them know what particularly resonated with you. I can’t tell you how much I appreciate that some of the smartest folks in venture are willing to take the time and share their insights with us. If you feel the same, a compliment goes a long way. Okay, that’s a wrap for today, until next time, remember to over, prepare, choose carefully and invest confidently. Thanks so much for listening.
