Entrepreneurship in the Age of AI with Jeff Bussgang, Co-Founder and General Partner, Flybridge Capital, Senior Lecturer at Harvard Business School
Willy Walker: Good afternoon and welcome to another Walker Webcast. It's my real pleasure to have my long-standing friend, Jeff Bussgang, to talk about AI venture capital and more specifically his incredible book, The Experimentation Machine, Finding Product Market Fit in the Age of AI. Jeff, it's great to see you.
Jeff Bussgang: Great to see you, Willy. Great to see everybody. Thanks for having me. It's great.
Willy Walker: It's great. Jeff, let me do a quick intro and then you and I can dive in kind of in a bunch of different ways. I think one of the most interesting things for me about your book is that it is written for entrepreneurs. It's all about entrepreneurial startups and technologically driven entrepreneurial startups. At the same time, as I read it as the CEO of a reasonably scaled publicly traded company, I sat there throughout it saying, “These are lessons that anyone who runs a business, anyone who is trying to either grow their business, stay in business or not be completely wiped out by this new age of entrepreneurship needs to read this book and listen to all the great things that Jeff is talking about.”
Jeff Bussgang: I appreciate you saying that. My audience definitely is the founder, entrepreneurial audience, Willy, but like you said, as I wrote this book and I thought about this kind of combination of timely tools and timeless lessons, I'm very much hoping that it resonates with folks like you and maybe some of your listeners who are thinking about how can AI be leveraged as a tool to make their business better, more efficient, more effective, more profitable.
Willy Walker: I will say, Jeff, just from an emotional standpoint, from reading your book, it's scary. I mean, it is fascinating, and we'll get into a number of the case studies that you talk about in the book. But as someone who has gone from a very small company to a reasonably scaled company and gone through all the different, if you will, processes of identifying your customer and then finding a value proposition that actually fits the customer and then going through all the pains of scaling and what have you, what is most frightening about it is that you mention in your book, Warren Buffett and creating businesses with moats around them. I think that the advent of AI and the implementation and implications of AI are such that a lot of people who've built some pretty deep moats around their business, as I think we have at Walker and Dunlop, need to be questioning how deep those moats are and how impenetrable those moats are around their businesses.
Jeff Bussgang: A thousand venture capitalists like me are investing in entrepreneurs that are going after those moats. The companies we invested in 8, 10, 12, 15 years ago, they're now larger companies, public companies like you all. Those companies are also looking over their shoulder and saying, “Holy cow, am I going to be the disrupted after having been a disruptor for so many years?” And the way I like to think about it, I think your point about it being scary is right. It's disorienting as well, because people who grew up in the internet age, like you and I did… we both graduated from Harvard Business School in 1995, shortly after the browser came out, shortly after the internet 1.0 era. We were trained in a certain way of doing business, and that way was very effective because we were disruptors. Now there's a new generation of leaders that are trained in native AI mechanisms to disrupt these new opportunities. That's the challenge for all of us—to try to reinvent ourselves and reinvent our businesses before someone else takes our lunch.
Willy Walker: Just on that, from a personal standpoint, Jeff. I am going to get to your intro before we spend the entire hour talking about fascinating things. As you personally use AI, I'm certain there'll be over 100,000 people who will listen to this. Some of them are as insightful on AI as you are. Others are complete neophytes. As they think about it, if somebody is listening to this and saying, “Wow, I need to get on the bus as it relates to understanding the implications of AI,” what would you suggest for him or her to do as it related to starting to play around with it, starting to engage in it beyond just downloading ChatGPT or becoming a subscriber to Grok?
Jeff Bussgang: I know from my teaching at Harvard, different people learn in different ways. So, some people are going to love just reading and listening to videos, and I will share with you afterwards some resources that I recommend to my students to help them get up the learning curve. For some people that are self-learners, that might be very effective. Other people want a buddy. One thing that I recommended is to get a reverse mentor. Go to a younger, perhaps, native AI person and ask them to be a reverse-mentor to you and sit by your side virtually or physically and show you some of the tools and walk you through how their own personal workflows have changed dramatically. I was talking to a VC friend of mine who's in their 60s and he told me that for the first time he was using ChatGPT as a mechanism to create a presentation to help one of his portfolio companies to talk about these new changes in their market. He found that in two hours of conversation, ChatGPT allowed him to create the presentation all the way down to the PowerPoint slides, as opposed to what might have taken a week or two of research and thinking and Googling around and writing. So you're never too old or too young to reinvent yourself and your workflows and your personal productivity, but you need help doing so, like you said. So either I can point to some resources to do it yourself, or strongly encourage you to get a reverse mentor and get them as like an AI Chief of Staff on your side.
Willy Walker: You talk about age, and in doing research on this webcast, I realized that your birthday is a week after next. So happy 56th birthday on the 20th of June, Jeff. Let me quickly read through the bio and then we'll dive in. Jeff Bussgang grew up in Boston as the son of two Harvard professors. He attended Harvard College and Harvard Business School and joined Open Market, a unicorn startup right after HBS. After participating in Open Market’s IPO, Jeff co-founded Upromise, a loyalty marketing and financial services company. He founded Flybridge with our mutual friend, Chip Hazard, where his bio reads, “former entrepreneur turned VC, HBS senior lecturer, author of two, dad of three, husband of one, civic leader, and a fan of all Boston sports.” He is the author of several books, case studies, and journal articles, including The Experimentation Machine: Finding Product Market Fit in the Age of AI, which we will discuss at length today. Before diving into that, I gotta ask you, Brad Marchand— I mean, Boston Bruin par excellence. You traded him away, and now he's leading the Panthers to what looks like another Stanley Cup. That's gotta sting, huh?
Jeff Bussgang: It rhymes a bit with Mookie Betts going to the Dodgers. It is kind of tough.
Willy Walker: But I have to say, Jeff, Boston has had its fair share. I mean, your kids growing up in Boston, our classmate Dana Schmalz, who's on my board. I always say to him that his kids growing up in Boston over the last two decades have just had a lifetime's worth of championships. So you can't be too bitter about that one.
Jeff Bussgang: There's a running joke with my kids. After a year with no duck boat parades, they wonder what went wrong in the universe.
Willy Walker: Exactly. So talking about experiences, Jeff, I was thinking about your career and all the different things you've done. And I was just curious. As you think back on the thrills of careers, non-personal, just business side, and think about being on the floor of the New York Stock Exchange the day that Open Market went public soon after you got out of business school, raising your own series, a round for Upromise, having your first 10X investment at Flybridge, or your first day of walking into the pit at HBS and teaching a class, which one of those is the one where you just sort of had to pinch yourself and say, “Man, is this actually happening to Jeff Bussgang?”
Jeff Bussgang: I would say the HBS pit, and I'll link it to the Boston sports. The first case I wrote was on Curt Schilling and the career he was trying to pivot away from being soon to be, maybe we'll see a Hall of Fame pitcher, certainly a world champion to being an entrepreneur. The question was, could this guy be a Hall of Famer in two different careers? The answer turned out to be no, but Curt came to that class. So not only will it be I in the pit teaching the 90 students for the first time, but I had Curt Schilling looking over my shoulder as we were judging his quality as an entrepreneur.
Willy Walker: Talk about a high wire act on your first time into the classroom. Jeff, you grew up reading sci-fi novels. Does it scare you a little bit that life seems to be following fiction today? As you write in the book, science fiction has become human fact.
Jeff Bussgang: It's a little wild. I came to Harvard College, as you noted, in 1987 to study computer science, which I majored in, and AI. And I took graduate courses in natural language processing and computer vision, where we studied parsing sentences so that computers could eventually pass this Turing test, this famous test that indicates when a human cannot determine whether on the other end of the chat is a human or an AI. We've passed the Turing test officially for the first time just in the last few weeks and various research that academics have done. It's crazy what's happening today, as many of us feel like it's the second inning or third inning, to continue with the baseball analogy. We've got so much growth ahead of us. So it's kind of wild, Willy.
Willy Walker: Scare you?
Jeff Bussgang: I wouldn't say it scares me. I actually find it exhilarating. It's exciting to see the productivity, the potential. Certainly there are risks and downsides. Not to be political, but I thought of all the things that JD Vance has done in the last few months when he went over to Europe at the AI conference and said, “Hey, I'm not here to talk about AI controls. I'm here to talk about AI opportunities.” He really struck the right tone. I think we need to really shift our thinking to an abundance mindset, as opposed to a scarcity mindset, where we look at AI as a tool which is turning us into a society of potentially extraordinary abundance where compute and intelligence and data and insights are free or very, very inexpensive—what human beings can actually achieve with that incredible resource at their disposal.
Willy Walker: But you talk in the book, Jeff, at the end about ethics and about the need to act in an ethical way and that the companies that Flybridge invests in always have that as their business ethos and mantra, and it's part of the founding of the companies. Obviously, knowing you and Chip, I'm certain that that's high on your priority list. At the same time, I'm also certain that there are plenty of people around the world that don't even come into consideration. So what kind of guardrails do we have that are either in place or potentially going to be in place that can protect us to some degree against this technology falling into the wrong hands?
Jeff Bussgang: It's a great question. This is what the whole defense tech and cybersecurity industry is very focused on today. And certainly it's something that the large companies that control these models, OpenAI and others, are focused on. As I was doing my book tour OpenAI invited me to come speak to them and as I went into their offices, their security is incredibly tight because they are treating their assets and their materials as if it were a national security secret, which it is. So I think we've got a…I sort of have a couple answers on that very important question. One is it's important that we do have the leading companies be American companies that are transparent and operating in an ethical fashion with checks and balances. Secondly, that we have a defense industry and a cybersecurity industry that can protect us from bad actors. There are a lot of bad actors who are acquiring and leveraging these tools for deep fakes and fraud and other things. All of our portfolio companies and all the major tech companies are super-focused on those issues and trying to stay ahead of the bad guys—one, two, three steps ahead in preventing some of these problems. We're going to need better fingerprinting of AI models and better kind of watermarking to know what content is being generated by AI, what video content is been generated by AI. One of my friends says that he now has a safe word with his family so that they know that if an AI clone tries to hack into their family and the systems and ask for help or have a kid call in and say that he needs money wired to him, that there's a safe word that they have as a family to make sure of the authenticity of that situation. I mean, we have to think about some new things in this modern world for sure.
Willy Walker: Big time. Associated with that, and you mentioned it in the book as well, is the lawsuit that the New York Times has against ChatGPT as it relates to copyright infringement. Just curious, do you think that there's a chance that you get a significant ruling there that could start to unravel the LLMs that have been built just because they've been aggregating so much data from so many copyrighted sources that all of a sudden they have to say, “Hold it, this whole thing has been basically building on itself and we didn't build in the safeguards as it relates to copyrighted information,” and so therefore one of these LLMs could end up being taken down.
Jeff Bussgang: People refer to this as the original sin of AI. It's like Eve and Adam grabbing the apple. It's pretty clear that the models were trained on public copyright information or let's say private copyright information. I think it's just going to be a matter of economics. So I don't know if it's going to take down a model, but I think it's going to be a negotiation. And you've already seen some companies come to terms. Reddit struck a few deals; News Corp struck a few deals. I think this is going to be more about economics than existential. But it's a pretty material issue. As the evidence, we have a portfolio company that's actually on the inside of some of these issues because they're very good at detecting model content training source data, and it's pretty clear that these models were trained on copyright information.
Willy Walker: Fascinating. Let's back up for a moment to where the book begins, because I think it begins focusing on, if you will, typical startups, not AI-enabled startups, but then moves very quickly into how AI is allowing startups that Flybridge invests in to do a lot of things a lot quicker, with a lot fewer people, et cetera. And one of the businesses that you focused on at the beginning, Jeff, is Squire. It's a startup that was started focused on how can people get a haircut appointment a little bit easier than going and finding the barbershop they want to go to. I love that you picked something like that is so sort of, I don't know, I'd call it meat and potatoes in the sense that it's like, “Yeah, that it is like Open Table for haircuts.” Right? But I think one of the things that I found to be so interesting about it is that the entrepreneurs in that case, Jeff, iterated. They sat there and they thought, “Well, we're going to first sell this to someone who's a consumer who just wants to go on and find a place to get a haircut.” And then they were like, “Well, no, actually, maybe we can automate the actual experience for the owners of the barber shops.” And then, they went and they actually set up a barber shop inside of a WeWork workspace to sit there and try and answer the question of ‘who is our end customer.’ Why I think it's such an interesting place to start is that what you're seeing in all of your subsequent companies is the ability to iterate using technology that sort of compresses that iteration process from, in this case with Squire, years into in some instances, months, weeks or days. Talk about that for a moment.
Jeff Bussgang: Yeah, you really nailed it, Willy. About 12, 15 years ago, Eric Ries wrote this book, The Lean Startup, and that revolutionized how all businesses thought about experimentation and creating these minimum viable products and iterating and learning from customer discovery and getting out of the ivory tower and getting in contact with your customers. And what I've been seeing is my portfolio companies and the companies I work with through my teaching at Harvard Business School are leveraging the AI tools to accelerate that learning process, to accelerate iteration and experimentation, which is why I decided to call the book The Experimentation Machine. I think all companies are working really hard on instrumenting their businesses to allow for faster, more effective experimentation. Booking.com, which is a $150 billion market cap company, this travel industry behemoth, runs thousands of experiments a day on their website all over the world, across all different products, as they iterate and iterate to test what messaging and what design will be most effective in driving conversions. What I've been seeing is that our founders are using AI to drive that experimentation even faster and more effectively than ever before. They're creating AI personas that represent their customers, and giving 24/7 access to everybody in the organization. It's like a real-time focus group available at any time to have that customer be literally at your beck and call 24/7. They're able to mock up different applications and different ideas and different workflows in instants or minutes. Things that would have taken weeks to develop are able to be achieved in a few keystrokes. And so that's allowing for so much faster execution. I think the thing that I've really been struck by when you look at the successful companies that are out there is just the speed, the sheer speed of execution. We used to talk about internet time execution, but AI time execution is just like nothing we've ever seen.
Willy Walker: On that, as you think about the iteration, one of the things that you focus on a lot throughout the book is the product market fit, the PMF of any given product or service. As you talk about the product-market fit, you also talk about moving from the jungle to the dirt road to the highway, and that's sort of the various iterations of a typical startup. I think it is important for people to keep in mind that the foundation of your book is based off of the, if you will, old venture model and how AI is completely changing all of these concepts from the PMF of how do you figure out what market your product is going to actually fit into to how long it takes you to move from the jungle to then getting onto the dirt road to then not just the highway, I would call it the super highway. Talk about a little bit of what you were seeing in investments from 2017 to 2020 and what you're looking at in investments between 2024 and 2026.
Jeff Bussgang: Chip and I started Flybridge over 23 years ago—2002, 2003 time frame. And for 15, 20 years, the model for seed investing, which is what we have been doing day in, day out, is you work with entrepreneurs, they raise a 2, 3, 4 million dollar seed round, and they take that money and build a team. They hire 8, 10, 12 employees. They build a product. And they race to acquire customers and get to some meaningful milestone, call it half a million or a million dollars of revenue, and then they go out and raise a Series A and so on and so forth as they grow and build and eventually hopefully build a valuable public enduring company. What's happening today in the last two years completely changes the model. We're seeing the era of seed strapping where founders are raising still that seed fund of one, two, three million dollars, but then they have only two or three or four employees and hundreds of AI agents that they employ to build the product, that they employed to go to market, that they employ to leverage new distribution channels to get to some revenue milestone, and then they skip the series A phase and they just keep growing and growing and going. We have one portfolio company, for example, that's reached 10 million of revenue with only six employees. Sam Altman famously has this quote from a few months ago where he said, “Hey, in my founder's circle, we joke that what will the year be when we see a unicorn, a billion-dollar valued company created by a solo entrepreneur?” And I don't know if that's going to happen anytime too soon, but I did this analysis. I measured the unicorn universe today. There are 12 of them with fewer than 100 employees, Willy. I bet next year we'll see a number with fewer than 50 or even 40 employees, and maybe in 2027, fewer than 20 employees. The productivity improvement that our teams are able to create with these tools, initially coding tools, customer service tools, sales and marketing tools, but eventually operational tools, chiefs of staff, advertisers, writers, the whole range of functions that you have in your company all being replaced or enhanced by AI agents. That's the future that we're entering into.
Willy Walker: Talk about Matt Mireles and TalkTastic, because Matt, two things on that. How fast did it take Matt to create TalkTastic and how much code did Matt need to write to create TalkTastic?
Jeff Bussgang: Matt is a business founder. He's reasonably technical, but he's not a coder. He was able to build this product extremely quickly within a matter of days, and realized he needed a customer service capability. The product is similar to Wispr Flow and some of these other tools where you speak into it and it generates text in any format you want—email, LinkedIn, an X post, a blog post, a New York Times article, even a college admissions essay, so maybe something useful for the kids of the listeners out there. Then, he realized, “Well shoot, I've just launched this thing and it's this incredible product and it is incredibly viral. I need a customer service capability,” and he decided to create his own customer service capability and so he would take the queries and questions from customers and he would feed it into ChatGPT and get better answers and he'd take the voice, transcribe it, get better answers, and expand on it and then feed it into his customer service platform, get the answer out of the customer service platform, and then pull it out and transcribe it and create a video avatar of him explaining to the customer, what is the problem they were trying to solve and how to address the problem and basically creating this workflow of attaching these different AI tools all together in a way that closed the loop completely. I've been talking to some of my other founders since I wrote that, Willy, and since the book came out, where we're talking about this vision and you may see this in your business, where we will see customer problems come in overnight, be processed, solved by AI agents, where software will be automatically written, and then the next morning, when you and your product team come into the office, you'll have the features waiting for you to approve or disapprove. Swipe left or swipe right or tweak a little bit. Implement the features and requests that the customers gave you the night before. I mean, this ability to close that loop and iterate so rapidly is right here. It's right now and upon us.
Willy Walker: I got a thousand things to jump in on that. One thing I would just put out there, if you wanna look at the reverse side of this, of a TalkTastic, but on the downside, I used to work for a company called TeleTech TTC, which is a big call center company. We had call centers all over the globe and coming out of the pandemic, TeleTech Market Cap was a $5 billion publicly traded company. TeleTech’s Market Cap, last I checked a couple of weeks ago, is $250 million. Because all of this customer service is being automated by AI, and there's really no need to have all those people and all those former Walmarts that are sitting on telephones talking to people and waiting for the phone to call. And so, this is both a great opportunity for entrepreneurs to create the type of things that TalkTastic has done and also a massive, massive disruptor of existing businesses that we've sort of relied on in the past to deal with customer service and all sorts of different types of help desk functions. One of the things, Jeff, that I wanted to go in on that, though, was this. So on TalkTastic, he was able to iterate the product using AI. Talk for a moment about Topline Pro and how Nick and Shannon used AI to basically identify their customer. Because from my read of your book, they started out saying, “We want to work with pairing up plumbers to opportunities where they can actually go get a job,” and then that iterated into working with contractors more generally, but the contractors today…I happen to be in Big Timber, Montana. And I would imagine for Nick and Shannon to find a contractor in Big Timber, Montana when they were starting up Topline Pro was pretty difficult. The Yellow Pages don't even exist anymore to sit there and leaf through the Yellow Pages. The person who's the contractor in Big Timber is out on a job site and not coming home and potentially checking in their Outlook emails or even answering their answering machine because they've got enough workflow here, and so they're not being introduced to Topline Pro’s services that could automate and streamline their business. Talk about how they used AI to find the end customer.
Jeff Bussgang: So Nick and Shannon, who by the way were HBS dropouts, which I had nothing to do with, but after their first year they dropped out. They joined Y Combinator, and they pursued this idea of helping contractors like your friends in Montana, whether it's an electrician or a plumber or a landscaper or a roofer who's too busy working on the roof to be on their desktop or making appointments and posting websites. They say, “What do these people need the most?” And they began to develop this AI-agentic capability to build the websites for those individuals, to help them with their sales, with their customer service, and with their billing and operations—basically, business in a box with AI agents for service pros. The business of Topline Pro has taken off. It's been an extraordinary journey. Well, to your question about how are they using AI throughout their organization, they realized that they had a sense of their ideal customer profile, their ICP, and all these different qualifying criteria, and they were trying to train their sales organization about this, and they decided, “Well, why don't we just build an agentic workflow for this and have an agent that's an AI agent that goes through the criteria and looks at all the information about service pros, not only in the service pro listings and professional association sites, but also let's look at Facebook pages and Instagram.” And then they realize, “Well, wait a minute. Pros often do post about their jobs. They're up on the roof. They take a photo. They brag about the job they did at Walker's Ranch.” Topline Pro could use that information to help build the websites for the pros and then, they say, “Well, not only can we qualify who we want to go after, not only can we ingest their content and help them build the website automatically, we can create a demo and reach out to them in a personalized fashion,” and so, using these new video generation tools, they have one of their sales reps do one video, and then that video is replicated thousands of times using AI, customized and personalized for that service pro so that each individual pro gets a video of Willy the sales guy saying to them, “Hey, Joe's Roofer and Landscaper. I noticed on Instagram and Facebook you got some great jobs that you've been posting. I decided to create a website for you showing off your work since you hadn't done it yet. I decided to do some listings for you. Let me show that to you. Scroll through the website. Let me show you what that would look like if I were to handle your customer service and your billing and operations all generated with AI.” It turned their response rate because of this personalization from below 1% to nearly 10% and it's just totally transformed their cost of sale and cost of service. Many of us thought historically when we would see software companies, sorry, just to finish, software companies that service small businesses would say, “Well, the economics won't work. You can't sell, you can't reach, and you can service small businesses.” But AI is changing the cost curve and the sales and service curve, and it is dramatically changing the economics of supporting small businesses.
Willy Walker: It's that graph that you talk about in the book as it relates to the mice into the elephants and the whales and what the addressable market is and what it costs to go get them. That whole graph has just basically been turned on its head in the sense that you used to sit there and say, “I'm going to go, if you will, upmarket toward the whales,” and you would add your customer acquisition costs to get into that institutionalized world. It's very interesting because at Walker & Dunlop, when I first joined the firm, we were sort of a middle market player, and then we moved into whale hunting and got into the institutional space. Now all of a sudden, we're going back after what I would call the retail space of smaller loans and one of the big things, Jeff, that we ran into was that in our large loan business, all that data is securitized. So we can go out and we could pull all that data down and find out. Finding out what loans Blackstone has is not a terribly difficult thing to do, but finding out the person who owns a four-unit multifamily property here in Big Timber, Montana, for instance, is exceedingly hard to do because most likely that person is borrowing from a local bank or some other type of lender here, and they're not showing up in securitization data. So that whole business case that you just talked about— as I read it, I just sat there and said, “Are small balance lending businesses using those similar tools?” They can just explode by pulling all that data together and being able to find the end borrower, because I can almost guarantee you, we're a better lender than the local bank here in Big Timber. No criticism of them, but we’ve got better systems, better people, better processes, and we should be able to win that, except they've got a captive audience because they're in this small little town. If we can identify them through the use of AI, just like that company did, boy, oh, boy! That opens up a whole different market for us.
Jeff Bussgang: You're pointing to another area, which I think is fascinating. One of my other portfolio companies works in, which is a company called Zest AI. Zest AI is a company that was building ML models, machine learning models for credit underwriting. The company was founded 15 years ago and when they went to banks and credit unions 15 years ago and said, “Hey, I've got this AI model to do better credit underwriting,” the banks and credit union said, “That sounds like black box and very scary and regulated space, I'm going to stay away from that,” and they really struggled to get adoption. And then two, three, four years ago, the narrative flipped and now everybody, even small credit unions, are embracing Zest's technology to bring AI, modern AI, big data analysis to underwriting, which is going to yield better credit, more accurate credit, better profitability for those credit unions, more available loans for individuals and homeowners and small businesses, less bias, less discrimination. All these things that, again, as I said, taking the abundance mindset to heart, you can imagine this incredible efficiency and greater availability and expansion of credit through AI. And that's just one small example building on your example that you just described.
Willy Walker: I can't remember, Jeff, is Zest AI the company that you profiled as it relates to Medicare fraud and Medicare fraud prevention, or is that another company that's working on that?
Jeff Bussgang: That's another one entirely.
Willy Walker: I was fascinated by that and only that it appears from the way you wrote about it, that fraud in Medicare claims is rampant and so therefore being able to come in and apply the technology. I mean, in our world of lending, fraud is clearly a four-letter word. When fraud shows up, you like to jump all over it because there isn't that much fraud in our part of the lending world and particularly in the commercial real estate world and the single-family world. I'm assuming it's got a little bit higher prevalence. But it seemed like the company that you profiled there that this is their job because people try and come up with fictitious or false Medicare claims left, right, and center, and that they need to be able to go pull in massive data sets to track whether the actual claim is legitimate or fraudulent.
Jeff Bussgang: I think fraud is one of the really interesting early areas of AI capabilities. We're seeing this with another one of our companies called BrightHire that does interview intelligence. All interviews are now videoed and so you see these capabilities like BrightHire to allow you to interpret how the interviews are going, for hiring managers to get visibility across the interviews that are happening in the organization to drive insights and analytics. You can see things like who's starting interviews on time versus who shows up late, who tells the company story effectively, percent of time that men speak versus women, interviewer versus interviewee, all these types of things. But what's interesting is that they're being asked by their customers, “We want your help with fraud because there's a lot of interview fraud going on.” People who are pursuing jobs who may have two or three jobs or may not be exactly who they say they are or may themselves be AI bots to your earlier question of deep fakes and protections and guardrails. So this area of fraud across every aspect of our businesses, we're seeing a lot of opportunities and a lot need for.
Willy Walker: Before I get off this, as it relates to the sequencing of those videos, there's another company that you've invested in called Outreach that talks about the sequencing of the videos or of the emails to make it so that while AI is generating a huge amount of content. Let's say I work with JP Morgan. From a personal banking standpoint and so I'm assuming that JP Morgan could send me 100 emails a day with new ideas, but they know that I can't take it. Does Outreach sequence that and make sure that it's a tailored email to Willy Walker's needs?
Jeff Bussgang: Yeah, it's not one of our portfolio companies, but I do feature it in the book because it's a cool company and there are other companies in that space. All the AI for marketing tools are just incredible, the proliferation. And it's from the existing vendors, obviously Salesforce and HubSpot are coming out with incredible AI capabilities, but also newer, younger vendors, some that we invest in, some that others invest in. So one of the things I say in general to your listeners, to you, is, you’ve got to take this experimentation mindset forward in the organization and that's why a lot of CEOs are putting these top-down mandates, like the Shopify CEO and the Duolingo CEO most famously, where they're saying, “Hey, AI tooling and AI native behavior, that's table stakes.” Everybody needs to do it. It's a requirement. It's no longer optional, and now you have to organize your company in a way that allows you to be more facile and nimbler in bringing in AI tools and experimenting with AI tools, because the tools are changing so rapidly. I can't tell you what the best tool is going to be in a year or two years or three years from now, but I can tell you that your organization needs to be good at bringing in, evaluating, and embracing tools and then swapping out for another new tool if it's a better one down the road.
Willy Walker: You mentioned that there are 30 publicly traded companies that Flybridge tracks as it relates to the use of AI and parking to the side, Jeff, those that are creating their own LLMs because we all know who they are and we know what they're investing in LLM. But as it relates to operating businesses, like Shopify, who are you all tracking as it relates to scale businesses that again aren't in the development of LLMs? But who are taking AI and implementing it into their businesses in innovative ways?
Jeff Bussgang: Some of the top software companies are first, naturally, because they're so facile at leveraging these technologies like Shopify, Adobe, Duolingo, ServiceNow, some of the big software franchises, even Salesforce. By the way, I have a prediction. You may have heard this already, but Mark Benioff recently renamed their annual conference to Agent Force, and he's been pushing agentic AI as the unifying next wave for Salesforce. I predict in the next, let's call it three or four years, that Salesforce will change its name just like Facebook did to Meta and Square did to Block, I predict that Salesforce will change its name to AgentForce that will no longer refer to that company as Salesforce but rather it will be reinvented as AgentForce because the agentic AI future is coming and Benioff and that company is embracing it so rapidly. But anyway, those are some of the companies that I think are doing a really nice job. It's interesting we haven't heard that many case studies and my colleagues at Harvard Business School like Karim Lakhani and others are really trying to profile and showcase these case studies, but not that many other companies that have great case studies for using AI throughout the organization to drive efficiency and improvement. It's just beginning. We're just in that generation of moving from pilot to production. And so we've got a few examples that I think are quite interesting. Some of the consulting firms like BCG and McKinsey are embracing it, as you might expect, as knowledge workers trying to stay on the cutting edge. But it's definitely like a tsunami that's coming in the next year or two of case studies of big companies leveraging these AI tools for their benefit.
Willy Walker: Talk about making me feel old. Andy Jassy, he was a year behind you and me at HBS. There are plenty of people with lots and lots of success around the two of us, but when the CEO of Amazon is someone who was after us at business school, it was one of those moments I'll never forget, Jeff, of just being like, okay, now I'm officially old when the COO of Amazon is younger than I am.
Jeff Bussgang: Officially, officially old and unaccomplished, frankly.
Willy Walker: Yeah, exactly. The double whammy. But he did talk about the fact that on Amazon Q, they saved 4,500 developer years of work or something like 260 million bucks of annual spend and so you just think about it at that kind of scale of what a company like Amazon can do. And you sort of say to everything from Amazon down to a startup, if you're not implementing these tools to both grow as well as cut costs, you're missing the boat in a very big way.
Jeff Bussgang: We're all really blessed to live in this era because we're benefiting from the AI dividend. All the billions of dollars that's being invested in infrastructure by the NVIDIAs of the world and the hyperscalers like Google and Meta and Amazon—we're all benefiting from that. All of us have applications in the cloud. We all have phones that have various capabilities and they're all getting AI capabilities immediately. These devices are getting AI-enabled, a billion of them. Every day, every week, every month, with these incredible updates. All of our cars, all of our homes. It's just incredible to see this global distribution of these capabilities so rapidly and that's what's so exciting to me, is that what was very different, and Mary Meeker writes about this in her latest annual report that she publishes, just came out a week or so ago. What's so different about the internet era, where you and I grew up as business leaders, is that it was a very US-centric era and it really only affected a very narrow part of the economy. The AI era is global and affects a very broad swath of the economy. It's affecting all elements. It's affecting education, nonprofits, government, everyone. So that's what makes this moment so exciting to me.
Willy Walker: We've talked a bunch, Jeff, about these high-flying startup companies that have harnessed technology. And I'm certain there are plenty of people listening saying either I'm going to figure that out or I'm not going to. I've got to invest in it. I got to get my teams focused on it, what have you. But one of the other pieces of the book that I found to be so interesting was the framework which Flybridge uses to underwrite the companies you invest in. And you go through both your hunch, and I want you to explain hunch to listeners, and then how hunch goes into KPIs that you all measure as companies start to scale because again, while it is focused on startups and while it's focused on entrepreneurs, these lessons of how Flybridge sets up hunch and the various key markers and then discounts them as well. And in your underwriting, one of the things that I thought was really great about your book was how you sit there and you talk about gross margins and some of these tech companies will be running 70% gross margins, but you sit and say, “Ah, they might be 70% percent gross margins but we'll put 60 on it.” And they sit there and they say, “We acquired the customer and we think the customer's going to be around for five years.” And you say, “Ah, if a Netflix customer's only there on average 36 months, you might not be able to use five years as your average customer life.” So before I steal everything you're going to talk about, talk for a moment about hunch and the framework you all use at Flybridge, because I do think it is incredibly useful for people running any size business to think about the hunch framework and then the KPIs that come out of it.
Jeff Bussgang: Sure. So I named this framework Hunch in the context of the world of determining product market fit, because in the early days you do have data, but the data is very early. And so at best, you have to apply some subjective judgment and have a hunch about whether this is working or not. And so the five letters of the word hunch represent five KPIs that we look at and really think about in the context of asking the question, has this start up got product market fit, and like you said, it applies to big companies with new products and trying new things as well. So the H is “hair on fire” value proposition. It's got to be a must have, not a nice to have. It's got to be so critical that the customer's hair is on fire. Our customers, your customers, there's so many things coming at them. They're so distracted. We all have so much in our inboxes and across all of our different channels of information and inputs that unless it's a top two or three priority, it's not going to cut through the clutter. So we really evaluate whether that startup we're working with is addressing a problem that's a true hair on fire value proposition. The U then is “usage.” Avoid the vanity metrics of downloads and installs and other things that show sort of initial interest. Observe the actual usage of the product. Do they use it more and more every day? Do they expand the usage across more departments, more parts of the organization, more units? And does the usage become habitual? That's really what we're looking for with the U in Hunch is the usage behavior. N is for the “Net Promoter Score” or NPS, which is a measure of customer love. Would they recommend it to their colleagues? Can you get some virality and word of mouth generated? The C is the question of “churn.” After a bit, you want to look at the churn rate and see month to month what those churn curves and cohort curves look like. And then the final H in the Hunch framework is “high unit economics,” high LTV to CAC, lifetime value to customer acquisition cost ratio. Just because you've got something that's working that seems to be getting used a lot and the customers love it—if you're not going to make money on it, it's not going to be an enduring, profitable, sustainable business. So we do think about an underwrite to the economic equation. Even if the numbers are early and we're throwing in a lot of assumptions and speculation, we try to be conservative; we try to imagine the future; we try to underwrite based on the fundamental unit economics. So we take that framework and then we look at companies and we say, “Well, look, early stage companies, early stage products, they're going to be emerging or nascent on some of these metrics and as they mature, hopefully those metrics are improving,” and so we provide a chart in the book, which you saw, which shows at each of the stage from nascent to radical product market fit what each of these five metrics should look like.
Willy Walker: I have to say it's a fascinating framework. I will tell you, in running a pretty established company, when I think about things like hair on fire, to be blunt about it, Jeff, I don't know how many of Walker & Dunlop's clients wake up every day with hair on fire to buy Walker & Dunlop services. I was going back and forth with one of our bankers just yesterday, and I was sitting there and we got kicked in the teeth on a financing by a client that we've been chasing for a long period of time and we'd sort of brought the whole cavalry. I mean, I'd gone to a meeting with the client and we tried to win this financing, and they didn't even give us the ability to compete for it. They went to one of our competitors and gave it to them. So I had this dialog with this banker just sort of saying, “How much time and effort have we invested in this client? How many resources have we brought to it?” And they don't even talk about hair on fire. They're bald in this situation as it relates to Walker & Dunlop, nothing's burning. And I just sort of said to the banker, like, “We got to think about, unless we come up with a value proposition that is wholly different from the firm that beat us to do this financing, we ought to just forget about this client, just move on.” And it was interesting, because I think I surprised the banker. He's been investing a tremendous amount of time. And I think that in our industry, there's a sense that if you invest enough time and you show them enough deal flow and you kind of talk to them about all these different things that we've done as far as structure and price and all that stuff, that at some time they're going to wake up and say, “You know what? I'm not going to work with Wells Fargo tomorrow. I'm going to give the deal to Walker & Dunlop.” But in listening to your framework, I also went to what's the lifetime value of that customer. And if you're just waiting for that one opportunity that they don't give to the competition after all the investment of time and effort we've put into it, I'm really not sure that we're actually getting a return on it. I found it to be wildly eye-opening in the sense that do we have the luxury to continue to invest in that client relationship? Of course, we do. We generate a huge amount of cash flow at Walker & Dunlop, and we're a very profitable company. Can I allow that banker to keep doing it? But I thought that the rules of a startup where capital is so scarce and identifying real customers who do have hair on fire for your customer really made me underscore the ‘what's the value proposition?’ Are they ever going to buy from us and what's the return on investment and time and capital that we're putting into winning that customer?
Jeff Bussgang: One of the things that I find in our world is, just to underscore what you're saying, is if you use certain words, it changes the frame of the conversation. So if you say, hey, is this an important value prop? Sure, it's important. But is it a must have? Is it hair on fire? It changes the frame. One of things I like to talk about is the quality of the business model. You may have seen this also. I talk a little bit about business model quality. Is it a good business model? Marvelous, that sounds wonderful. But is it a magical business model? Does it have all the elements of network effects and recurring and high gross margin? And we talk about customer profiles, ideal customer profiles. The word ideal really clarifies, is this customer, sure I could service them, but are they my ideal customer profiles? Because startups, like everyone, but even more so, have a finite amount of resources and they have to really sharpen their tools and really focus. That's the thing I tell my founders time and time again. A repeated piece of advice is you need to focus more, more narrow, more tightly. You've got finite resources. You got to focus on the really few things that are going to move the needle for the business.
Willy Walker: Who was it in the book that you point out says I'd rather have a hundred customers that love me than a million who like me?
Jeff Bussgang: It's a Brian Chesky quote, the Airbnb founder and CEO that he attributes to Paul Graham, the Y Combinator founder, but everybody refers to it as the Brian Chesky quote now, which is so powerful because look, if 100 customers love you, everything works. They keep coming back; they cost less to serve; they have amazing economics. Everybody's happy and if a million people like you, you're just churning and churning, and exerting a tremendous amount of energy to stay in place.
Willy Walker: Talk for a moment, just real quick, because I thought it was fascinating. You talk through the various valuations on different business models, where you talk about why do SaaS companies trade at 10 times revenue? Why do ad-based companies trade for less than revenue-based companies? Will you talk that through quickly, Jeff, because you did put the multiples in. While, as you're very clear in saying in the book, various business models are combinations of these business models at their core, the multiples that are applied to startups— like a snowflake versus someone who's selling ads on their give-it-away product are very different where the market will put a multiple on it. Will you just talk that through for a second?
Jeff Bussgang: Yeah, it's very funny, sometimes in my class I say to my students, “You know, up until now you've learned to characterize business models, but I want to teach you how to judge business models and be discerning.” And the discernment is what business models belong in the 10x club, as venture capitalist Bill Gurley first coined it, business models that are so high quality, they're worthy of being valued at 10 times revenue, not because it's frothy or illogical, but because they're that high quality versus business models that are not worthy of being in the 10X club and I rattle off a few of those. They have to have a magical business model. They have to network effects. They have high gross margins. They have recurring revenue. They have competitive moats and they have really enduring economies of scale and enduring and sustainable, profitable, sort of recursive feedback loops or what's known as kind of these positive loops. When you see those models, like a company like Snowflake or other SaaS companies that have really just exquisitely executed these magical business models, I would argue Shopify might be considered one of those as well, then it just sings and everything works. One of the things that makes investing in new companies so exciting is we're all inventing the business models as we go. At this moment in time, the SaaS business model has had a great 20-year run, 25-year run. But in the next 20 or 25 years, in the age of AI, the SaaS model is being questioned. Maybe we're going to move to more of a usage-based model or a value-based model, and so there's some really interesting new questions that we're all trying to figure out about business models and pricing. The companies that get it right are going to engender incredible valuations.
Willy Walker: We're about to run out of time to two other questions. One is you talk about scaling and I think a lot of the people listening to this aren't entrepreneurs with a startup business, but are actually listening to it of ‘what do I do with my existing business?’ Many of them have either scaled dramatically or are in the process of trying to scale. And you talk a little bit about the coin-operated sales force. To be honest with you, I sat there and thought about how the Walker & Dunlop sales force is coin-operated today. And I think one of the questions I have for you is on a coin-operated sales force, how do you get the implementation of AI or innovative thinking. As I sat there and listened to you talk about a coin-base sales force, it seemed like most companies in the old model would sit there and say, “Okay, we're scaling now. Let me go get an enterprise sales executive. Somebody who really understands how to do enterprise sales and let me get him or her to scale my sales force, get annual quotas, get the team focused and here we go.” And so as I heard you saying that I was like, “Okay that's the old business model.” What's the new business model as it relates to AI, as it relates to a coin operated sales force, but with somebody who has the technical capability or the imagination to incorporate AI into their sales function?
Jeff Bussgang: So first principle, you need to incorporate an AI score into your evaluations, into your performance reviews. So first you have to ask yourself, “How AI native are we as a company from a scale of one to 10?”, “How AI native am I as an individual executive on a scale one to ten?” And then “How AI-native are each of my executive team members and each of their team members?” And then “What are the things that we're doing?” And again, you got to put this right in the performance review. What are the things we're doing to up our score? Are we celebrating AI experiments? Are we doing hackathons? Are we experimenting with tools and encouraging people to bring in new tools and try new processes and new workflows? And are we bringing in outside AI native executives who come in with a history and a track record of implementing some of these processes? It's all so new. So some of those people you may have to take risks on; they may be outside of the industry; they may be a bit younger. This is my point about reverse mentoring and AI chiefs of staff. Some organizations are creating small AI tiger teams that are just evaluating every part of the workflow, every function and trying to inject AI experiments. And so you've really got to take an intentional view to try to turn your organization, including the sales force, from let's say a 4 out of 10 or a 5 out of 10 today to ideally an 8 or a 9 out of in the next year or two.
Willy Walker: Fascinating. And it's so practical. I mean, they like, that's something that we can and should do at Walker & Dunlop and we're, again, a pretty successful company, a pretty scaled company. I think I’ve got a great executive team. But the concept of sort of measuring up on that and then measuring it and following up on it and trying to upscale ourselves, and then also that reverse mentor model that you talked about at the top of the webcast is also another one that I find to be fascinating and quite honestly, easily implementable. I mean, you don't have to go out and hire a whole bunch of people. There are a lot of new employees at W&D who could be very helpful there, which goes to my last question, Jeff, because there was an article yesterday in Bloomberg that talked about the employment of the class of 2025 out of college and the two skill sets coming out of college that are well, if you will, above from an unemployment standpoint, below from a hiring standpoint of the class of 2025 are mathematicians and computer science engineers. Well, well below, the data has just come out as it relates to mathematicians and computer scientists engineers are all searching for jobs when in past cycles, they have been the first picked up. And so at the top, I talked about a father to three. When you give your kids advice on what to study and what to do, what are you telling them to do?
Jeff Bussgang: Well, first, I wish my kids followed my advice or I could claim they do, which they don't. But I say to people who ask me this question, and look, I studied computer science as an undergrad and it served me well, I don't think I would encourage children today to study computer science or engineering because I do believe that AI tooling is getting so good at that. Instead, I think it's a moment for the revenge of the liberal arts, the revenge of the MBA to read about Shakespeare and understand human nature, to be able to understand strategic thinking and read Michael Porter and Clay Christensen, and to really be able to lead and discern what is the right thing to do. The AI will do whatever we tell it to do, but discernment is the real question. And then I also tell folks, play with the tools. I think this year's class of 2025 at HBS was the first AI native class of MBAs. I think we did a really bang up job over the last two years of training AI native MBAs, and I think you'll see more and more of these AI native generalists and executives and young professionals entering into the workforce. I think that's going to be really exciting. You may not hire more software developers. But you may hire more AI native product managers and marketers and designers who can leverage those tools to great effect.
Willy Walker: I think one interesting thing about that, Jeff, is when you think about eras, and if you look at graduating classes from either undergrad or from business school, there are certain eras where people come out into the advent of a new financial framework where asset-backed securities came out and everyone who went into asset-backed securities did really, really well, or go back before that to Michael Milken and junk debt and people got into junk debt. And when you and I came out in ‘95, you and I had AOL Beta when we first went to HBS in ‘93. And I remember putting those disks into my laptop and we thought we were pretty advanced and what have you. And a couple of people from our class, including you, went to some high-flying internet startups that were quick hits, but relatively small hits. But one of the things that I consistently think about is that, Amazon was starting at that point, and Jeff was in business when and I believe ‘98, but if you'd invested in Amazon in 1998, you didn't get back into the money until 2008, that it took a decade for that business model to really get going to the point where then, obviously, in hindsight, it all took off from there, but that we're going to see lots of fits and starts. They're going to be winners and losers, and identifying which are going to be the winners and losers in the AI landscape is not you do it every single day, it's up to you and your team to identify the winners and the losers, but that this is going to take some iterating and it's going to take some time before we really see how this class of ‘25 that you just said is the first AI literate class coming out of HBS goes and actually takes that learning, takes that understanding, puts it into practice and then goes and creates real value.
Jeff Bussgang: I think that's right. I think one of the fun tests for me when people ask me, “Are you optimistic or pessimistic” is would someone like you and like me want to trade places with the class of 2025 in terms of the opportunity set ahead of them? I think the answer is yes. I think almost anybody in their fifties or sixties today would trade places, not just for physical reasons, but for professional opportunity reasons, that this moment in time to be an entrepreneur in your mid late twenties, in this age of AI where we believe so much value is about to be created. I think it's an incredible, incredible opportunity and just an amazing time to be an entrepreneur and to be alive.
Willy Walker: Jeff, it's such a great conversation. I could keep going for another hour. I'm deeply grateful. The book is fantastic. Anyone who listened to this conversation who doesn't understand how great this book is, I've certainly failed in my job of trying to convey that because everything you've talked about is so eye-opening and insightful. Thank you, my friend. I'm sorry I'm not going to see you this weekend, but I will see you hopefully soon enough and give our mutual old friend Chip my very best.
Jeff Bussgang: Thanks Will, I really appreciate you having me.
Willy Walker: Take care, Jeff. Thanks very much.
