Overview
Dan Mishin, founder and CEO of Manifest Law, an AI-native immigration law firm, talks about building a services business on top of AI. He started his first company at 11, turning his grandmother's apartment in Ukraine into a backpacker hostel, and later founded the rental platform June Homes. The conversation covers why this wave of AI services may avoid the failures of 2010s tech-enabled services, why lawyer adoption is the real bottleneck, and why brand matters more than most technical founders think.
Key Takeaways
Mishin argues the 2010s tech-enabled services companies could grow revenue but never got real productivity gains, so quality slipped and margins never expanded. AI changes that. If you believe AI raises output per worker and the market is big enough, he says, margins and growth follow.
He doesn't fear price compression. Lower prices are a way to win share, and his incumbents can't match them. Law firms are partnerships that pay out cash every year, can't fund R&D, and legally can't give equity to non-lawyers, so they struggle to hire engineers. Horizontal work like NDA review should collapse toward free. Defensibility comes from depth in a niche that looks small from outside. Manifest started with U.S. immigration, which he puts at a $5 billion total addressable market (TAM), and is expanding into global immigration for the same clients before adding one adjacent legal practice at a time.
Quality at scale is the hardest problem. Manifest uses AI agents that track client sentiment and missed deadlines in real time, check lawyers' drafts as they write them, and estimate approval odds throughout a case. The metric he watches most is cases per lawyer.
Of the four models for AI services, he ranks outsourcing work to contractors as the worst, because people who aren't bought in won't change how they work. Manifest also hires in-house, recruits partners who bring their client books with them, and uses mergers and acquisitions (M&A) selectively. Lawyers join to serve clients better and drop busywork, but also out of fear of being left behind.
On go-to-market, the activities that are hardest to measure have worked best. One example is Mission, a community for immigration managers that he says became the largest in the field. Another is a heavy in-person event schedule. He no longer trusts cold outreach in writing because so much of it is AI-generated.
Practical Steps
- Pick a narrow vertical where the rules are clear. Mishin likes immigration because there's no opposing side and the government publishes exactly what it wants.
- Launch one practice area at a time and get dense before adding the next, even if investors push for more.
- Earn the right to change professionals' workflows. Start with tools that mirror how they already work, ship requested features fast, and leave no support ticket unanswered.
- Put AI quality checks throughout the process, not at the end, so errors surface before they become bottlenecks.
- Replace "angriest email first" triage with an agent that ranks tasks by deadline, recent client contact, and sentiment.
- Use quarterly North Star goals mainly to decide what you're deprioritizing.
- Put budget into in-person events and community, and accept that you can't attribute ROI cleanly.
- As CEO, give smart teams clarity and fast decisions. Most choices are reversible.
Notable Quotes
"The real constraint to AI-native services, I don't think it's technology or product building. It's actually adoption." - Dan Mishin
"You get punched in the face enough times and you kind of learn the lesson. When you lose somebody extraordinary because you're a micromanager, and it happens four times in a row, you kind of learn it over time." - Dan Mishin
"The most common trait amongst every great founder is every year they're a different version of themselves." - Host
Full Transcript
We don't care about the number of use cases that we service. We do only one thing and one thing only, which is we help American companies to hire international talent through immigration. We are the company in the space. There's nobody else. In this episode, I'm excited to sit down with Dan Mishin, the CEO and founder of Manifest Law, an AI-native immigration law firm. For us, immigration was a perfect starting point. Five billion dollar TAM, real tech leverage. You don't have the opposing side. You're dealing with the government. Government tells you exactly what they want. Dan has been a founder for most of his life. He started his first company at 11 years old when he turned his grandmother's apartment in Ukraine into a hostel. He later moved to the U.S. and founded June Homes, a rental platform he scaled to more than 2,000 apartments in 11 cities. In our conversation, we dig into why Dan thinks AI-native services companies won't repeat the same mistakes as the last wave of tech-enabled services. The real constraint to, like, AI-native services, I don't think it's technology or product building. It's actually adoption, and how he convinces lawyers to join Manifest. We also discussed why succeeding as an AI company today hinges on the ability to build a strong brand, and what Dan's learned from founding multiple companies. You get punched in the face enough times and you kind of learn the lesson. When you lose somebody extraordinary because you're a micromanager, that happens four times in a row, you kind of learn it over time. Let's dive in. What was your path into entrepreneurship? I don't really remember, like, one specific moment in time when I was like, Hey, I'm going to become an entrepreneur. It just kind of, I just, I just, like, fell into it in so many different ways. Everything from constantly selling and buying things from everyone since, like, six years old, like antiques. I bought a bunch of antiques, resold a bunch of antiques. I was buying, selling souvenirs to tourists that are coming in town. It was, like, endless. Every day of my childhood, that's what I remember. It was like a non-going hustle. And then the first, like, real business was, like, at around 11 when I, like, just started launching the backpacker hostels. And I kept building that company for 11 years, 12 years. And that was, like, my most formative experience. It was a really hard business to run. It was very, it was very, like, operationally intense. And, you know, you're selling ten-dollar night beds to drunk students, you know, who are 20 years old. Like, a lot of crazy shit happens. How did you start doing that? I was 11, and my mom and I, we were—the family was saving money for—I'm from Ukraine, from, like, a small town. We were saving money to go to Europe, and for years we were saving, like two years. And then we finally could afford to go to Berlin, and we took a train ride, which took, like, probably, I don't know, three days or something. And then we spent time in Germany and Netherlands, and on our way back last night, where the Berlin train station is supposed to take a train back to Ukraine, my mom's wallet gets stolen right at the platform, and it has the tickets and the money. So mom is like, Oh, okay, like, we need to sleep somewhere overnight until our father wires us money via, like, Western Union or something, you know, that used to take, like, a full day, like, until the money arrives. And we have 10 euros. That's all we have. And we go to, like, the cheapest backpacker hostel in the outskirts of Berlin with, like, a bunch of graffiti on the walls and a bunch of people, like, smoking weed, and sleep in, like, a 12-people bedroom, like dorm rooms with bunk beds, until the train next day. And I remember sitting in the living room in that backpacker hostel in Berlin and seeing, like, a group of, like, 20-year-olds who are having the blast. They're drinking beer, they're having fun, they're speaking in language that I can't understand. Everyone was talking English. And I was like, Wow, this is so much fun. Why, like, what's going on? Everyone is happy. Everyone is, like, good-looking and stuff like this. So on my way back, I remember this precise moment when we were on the train back to Ukraine, I made two notes on a piece of paper. The first note was, I want to learn English, and the second note is, I want to own a backpacker hostel. So I get back home in my hometown in Ukraine, and my grandma lives in a two-bedroom apartment. So I make a pitch to my grandma why she should move in with us, with my parents. Like, after two months of whining, grandma moves out of the apartment, moves in with my parents, and I turn that apartment into the first backpacker hostel, and put it online. It was the first property, I think, on Airbnb in Ukraine. Really? Yeah, something like that. It was like back when the internet was like dial-up and shit. And ten minutes later, I get a reservation. And in three months, the family makes equivalent of like three years' worth of rent that this apartment would have ever generated. And then it was a seasonal business, so like started in May, ended in October. Then next year, I was like, okay, let me have two of those. And I hired a bunch of classmates, we rented an apartment, then we had two. And this is like when you were 13? 12, 13. So I kept—this is what I was doing between 11 and like 20, 22. And then it became bigger, and I relocated to Moscow. We opened an office in Moscow. We had a bunch of buildings in Moscow. We did ground-up construction. We did redevelopment. It became like a real thing. And then we raised $100 million from private equity to fund the business. And then that private equity two weeks later went bust, and they don't have $100 million. It was such a shit show. It was like just the most— But it was the same business that you started at 11? Yeah, same company. Was there one big thing that you took from that whole experience, or was it the zillions of ways in which you were taught by the business? Definitely a zillion ways, because I skipped middle school, I skipped high school, I skipped college. Entirely? Yeah, I didn't do any of that. And it was like a thing, like that was obsession. I slept in the car. Like when I was doing my first construction, and we were building like a five-story building, redeveloping, like it had the bones, but we rebuilt the whole thing in-house. And it was one mile away from Moscow—sorry, from Kremlin, like in downtown Moscow. In order for me to do that deal, I had to sign a personal guarantee on a 99-year lease when I was 18, with a rent of, I think, a million dollars a year or two million dollars a year. So, and I was like, and I had a quarter of the budget that I needed to actually do construction. And I had 300 people on site working in shifts, like three shifts, eight hours each per day for nine months. And during that nine-month period, I was sleeping in the car outside of the construction zone. And, like, that was like, it was an all fucking in. There was just no opportunity to fail. And, like, everything possibly wrong could go wrong. Everything that could go wrong went wrong. Like what? Everything. It was like just, it was like, it was a shit show. Because, like, okay, like, I didn't know I had to hire an architect, as an example. Really? I just, like, started building it inside. Nobody knew, like, what we were building because there was no plans. You know what I mean? We were just putting things together. I didn't know you had to do that. I mean, ultimately, I did. It was just, your ignorance is your biggest opportunity and your biggest threat at the same time. Like, I was so ignorant that there was just no way to know what it's going to look like, and what do I need to do to get it done? But, like, look, I mean, the ultimate thing that I would say is, like, the main takeaway from that business was that I think I definitely have higher risk tolerance than most people. There is no doubt about that. Sounds like it. What seems very risky to people seems Tuesday to me. I have a higher tolerance towards uncertainty than most people, most founders, I would say, because everything is less risky. Like, I mean, at the end of the day, like, we're in venture-backed startups. You're going to be okay. You're going to be okay. Like, just don't do anything illegal, like, and everything is going to be fine. And even if you fail, and you fail for a good reason, and you tried your best, it is one of the possible outcomes, and everyone knows it. So, I don't know. That's probably, those are probably the main takeaways from that company. I was really excited to talk to you about your thoughts on AI-native service companies. It's one of the shapes of companies that has exploded in the past three years, and it's a remix of the 2010s tech-enabled services. And I think, like, one of the lessons of the 2010s is that tech-enabled services were more of a mirage. Everyone was like, Oh, this is going to be the next generation. We're going to take X services company, apply software, and then we're going to have a software company with software-like multiples. Maybe you can break apart what you think makes an AI-native service company have the chance to sort of compound in a much more unlimited way, and what are the bottlenecks or things you have to get around that will make this not just a replay of all the capital and enthusiasm of the 2010s, where on the other side you don't have these spectacular businesses that were generated. Amazing question. 2010, 2021, 2026, we as an industry kept stepping on the same landmines over and over and over again, and not just with services. Like, it happens across every type of software business, really. And sometimes I'm like, do we not just learn from our own mistakes? Okay, so if we look at it from the first principles, the way, I mean, obviously the value of the company is the function of its gross profit, its margins, and the speed at which it's adding the margins, right? Like the absolute dollars, like the speed at which you're growing. And market share, and can you dominate a market? Agreed. And the size of the TAM as a function. Agreed, those three. So if those are the three levers of the value of the company, the margin profile, if you cannot increase at least gross profit margins, but ideally all the way at the bottom line margins, as an AI-native service versus a service, then there is no multiple expansion. You should be traded at three times EBITDA, five times EBITDA, ten times EBITDA, whatever your specific industry commands. AI has changed the game in a sense where the productivity—there is a real productivity improvement. What 2010 didn't have, what 2021 didn't have, is productivity improvement. They were able to ramp revenue pretty well, especially early days before they started choking with, like, quality issues. But the productivity never came in. AI actually drives real productivity. You can feel it in so many different ways. And the only real thing that you need to believe if you're an investor in AI-native services in 2026, 2027, is whether or not AI will drive productivity. And if the answer is yes, I believe AI can drive productivity, and there is a large enough TAM, margins is going to follow, and the speed is going to follow, because a lot of those service businesses that were venture-backed in the past, they— They either plateaued or crashed on revenue at a certain point of their journey at the scale-up stage. What do you have to believe in terms of pricing power? Because you could have productivity improvement across a whole set of companies, but if everyone has it, you could have a unit that you used to charge a thousand, now you can only charge 700. So do you need some path where you're both driving some form of, like, operating leverage, but also not having a price collapse? Well, actually, I don't mind the price collapse because that's a tool to increase market share. Like Amazon playbook of, like, just sort of reducing prices to win the market share. I actually like that because your incumbents structurally cannot do this. Like, if you look at Manifest and the legal industry specifically, okay, who are we competing against? Our youngest competitor is 40 years old. Our oldest competitor is 120 years old. We are not competing against tech startups. Every RFP that we are on, we are against behemoth law firms with 500 to 2 billion dollars worth of revenue per year, with thousands of lawyers and absolute structural disadvantages that essentially prevent them from being able to compete with us. They're a partnership, which means they have to distribute all the free cash flow every year to their partners. They don't have the money to actually invest in R&D. They cannot give stock options to engineers because law firms are legally not allowed to give equity to non-lawyers. They cannot attract the right tech talent. So they're structurally unable to compete with us, and they're structurally unable to compete without price cost structure, which means we can come in and win on price while retaining margins. So I don't think of this as a bad thing. But you don't think that the form factor of this technology means that there's going to be 20 Manifests coming and competing against you? I don't think so because, well, it depends on the business. If you're going horizontal, let's say you do NDA reviews. I don't know how you can retain pricing power in that market. That should have never been expensive. It should collapse. Correct. Should be zero. Like, you should use OpenAI for that or Claude. If you're doing contract review, it depends on the size of the contract. If it's a simple vendor agreement that today AI native service companies charge 500 bucks for, I don't think that's a defensible business, to be fully honest. What manifests the choice that we've made is we said, we don't care about the number of use cases that we service. We care about the value creation. Let's go deep into one industry, and we do only one thing and one thing only, which is we help American companies to hire international talent through immigration. That's the only thing that we do. And because we went so deep in one industry, we established we are the company in the space. And this density, this focus, this expertise that you've built, the insights that you get at the most granular level, that's what makes this defensible. So in my opinion, the types of businesses I would invest in would be AI native services in niches that look too small to the outside world, but that actually have the depth to build a meaningful-sized company as a wedge. I'm not saying that's a forever story, but that's a starting point. And for us, immigration was a perfect starting point. Five billion dollar TAM, real tech leverage. You don't have the opposing side. You're dealing with the government. Government tells you exactly what they want. And from American immigration, we now expanded into global immigration, where the market size went from five billion to 20 billion, and we're servicing the same clients. It's pure account expansion motion. Because most of your clients have— Everyone has. Yeah. I mean, our clients have offices around the world and everything. And it's the same 60 countries where all of them have offices. So by the end of next year, we'll have immigration in 60 different countries where we can service them. We can continue growing revenue that way. Sometime mid next year, we're going to take the same accounts and we're going to expand them into another use case, because the same exact buyer who already went through the RFP with us, who already trusts us, has another legal use case where they allocate capital. Immediately adjoining to the immigration, where we will expand. And all of those points of entry become wedges on its own, because U.S. immigration clients then need global. Somebody might enter us through the UK office, and then they need U.S. immigration. And then the next practice area will be a wedge into immigration, immigration is a wedge into this. All of these things, like, feed into each other. And what we will never do is we're not going to launch three at a time. We're going to launch one at a time, build density, build expertise, hire the best lawyers, have a reputation of the company in the space, and then go to the next one. It's contradictory maybe to how VCs want—like, some VCs push me on this, and they're like, Dude, like, launch three practice areas at a time. And I'm like, That's not the point. You don't get measured on the number of use cases that you serve. You get measured on the ultimate value that you create. You mentioned something in passing that I think is really important, which is if you look at the last generation of AI, or I should say tech-enabled services, almost all of them had this quality where as they got bigger, quality went down, which is the opposite of, ideally, all great companies, right? That as you grow, the whole business gets better. In that case, when you were sub-scale, you delivered a magical experience to customers, and at scale, you churn customers. Talk a little bit about how that applies in this next generation of companies, or how you think about it and what you guys are doing at Manifest. That is the hardest problem to solve with any AI-native services across any industry, because you still have a higher human dependency than pure software businesses. And there's no one bullet that you can fire that's going to solve this. For the first time, we have a unifying technology that actually unifies quality and makes it scalable. If you have the same quality—like, what AI does is it allows you to control quality in fundamentally different ways. For example, instead of having surveys that nobody cares about, you have, in our case, AI case health tracker, which in real time—it's, I mean, it's basic, like sentiment analysis, broken promises analysis. Like, AI agent just runs a continuous quality assurance work to understand, like, where does each case stand. And proactively, before there is an escalation, a complaint, or an angry email, it flags that the sentiment has been deteriorating. And it's based on transcripts of calls, any written communications, any planned deadlines that were violated. With real-time flags into the quality team. Now, that is a fundamentally different way to treat quality than once-a-month survey. You have real-time signal. That was not possible pre-AI, really. Or the unification of the outcomes of the case. We have AI QA built throughout the process at three or four different touchpoints. If you're an immigrant going through an H-1B and your lawyer makes a typo in your case or a grammatical error or just doesn't do a good job, that pisses you off. We've built AI QA that, in real time, analyzes any work that lawyer produces, flags to them areas of improvement, suggests what they can do better. We have AI agents that act as a government assessor of cases, that evaluate the probability of approval at any point of time in the case, from the first touchpoint all the way to the filing. For the first time, there is tech that can actually do this. It doesn't solve all the problems. You still have the human inertia and lawyer adoption, and it's a battle, and it's not an easy battle. It's every single day you have to be dealing with this. Talk a little bit about actually how AI and software has evolved in the business and how it provides leverage over time. And maybe one way to think about it would be if you were to go back and look at the average lawyer at Manifest and look at an hour, day, or week, what were they doing last year? How is that different today? And how do you think that's different in 12 months from now? Some of it is very basic low-hanging fruits. Imagine you're a client going through the immigration process, and the lawyer sends you an email and says, I need these 10 documents from you, and one of the documents, let's say, is a birth certificate. And they send you a Google Drive link to upload the birth certificate. That was the use case before. Now you upload the birth certificate as a JPEG, and the lawyer is like, Shit, I really wanted that as a PDF, or I want that, like, you uploaded it in a French language. I need it in English with the certified translation. And the lawyer says, Go and certified translate. Then the client is like, Okay, where do I find a certified translator? And it becomes a— 13 email thread. That's a huge time suck. At Manifest, you're a client, you're invited to the portal. The first thing that you see is a shoebox, one evergreen drop zone. You drop a document, an agent in real time analyzes the document, compares it to requirements, immediately tells you, like, areas of improvement. Hey, the quality is not good. The format is not good. Translation, language, whatever. Lawyer hasn't even touched the case at that point. So, client with an AI agent collaborate on all the documents. All the documents information is automatically extracted, pre-populated into the questionnaires and government forms. So by the time the lawyer logs in, the case is ready. They just need to review it. So what used to be a 20, 30-hour exercise now becomes, you know, a 20-minute exercise. Now this is a future-looking state. Like, the product is there, but, like, the productivity is a function of what technology do you have and what's the adoption that you have. And we're still, you know, it's work in progress, of course. But, yeah, I mean, like, it shouldn't take 20 hours. It should take 20 minutes. The other thing is, like, traditional law firms, because they're so risk-averse, they have three, four rounds of lawyer reviews, which creates immediate human bottlenecks. The Eliyahu Goldratt kind of approach of, like, eliminate the most important constraint at any point of time. Just, like, the processes are so broken. If you have AI QA not at the end of the process when the case is already, like, ready to file and now you're delaying because you found a mistake. If you have AI QA going through the process at any point of time, 24/7, 365, you flag mistakes earlier and you eliminate human bottlenecks. So the productivity levers are completely real. The number one metric that we look at is the number of cases per lawyer before and after, and we're seeing major, major improvements there. What are your thoughts on On change management and in the world of AI, what's going on for humans? One of the interesting things in AI-native services is there's right now two dominant strategies, and at points maybe they sort of intersect. But one is we're going to create an AI-native services firm. We're going to hire humans. We're going to build our software operating system, et cetera, et cetera. The other is there's billions and billions of dollars that is going into some version of AI plus private equity, holdcos. You see this in accounting. You see this in a bunch of these. We're going to take these firms that have been around for a long time. We're going to acquire them. We're going to— We're gonna change manage them in some way. We're gonna apply software. And in a bunch of the strategies, the thing I've been kind of most interested in is what's going on for the individual lawyer or accountant or what have you. And it seems really hard, particularly with somebody who has a lot of experience. And ideally, you would want somebody with experience to get anybody to do anything differently. And I think there's going to be a bunch of the buyout strategies that work and a bunch that don't. And a lot of the ones that don't, they just can't get a thousand people to go and rewire the entire way that they're working. I fully agree, and I think you nailed it with, like, that is the real constraint to, like, AI-native services. I don't think it's technology or product building; it's actually adoption. So, I mean, look, sort of three separate strategies that—and by the way, we do all three. So we have all three parts of our business that do all three of them, so I can kind of see it across the board. Sorry, there is four, actually. So if you think about it, one model that maybe you haven't mentioned is I'm an AI-native company, and then I outsource the work. So I act as almost like a lead gen plus software, and then there is like a 1099 somewhere doing something. We've started there. It was so hard. If you're trying to have change management and whoever is your supplier, in our case a lawyer, is not bought into the mission and is not working for you full time, adoption becomes a nightmare. So that, in my opinion, is the worst of all the models. Then you have the model of, like, I'm an AI-native accounting firm. I'm going to hire a bunch of accountants who are super experienced. They're all in-house, and then I'm going to manage them. You got to gamify it. You got to make it interesting. You got to build the credibility and the trust to do it. You got to earn the right to change their workflows. You cannot sort of circumvent this part of earning the right to do this. We started with building more traditional tech that was more rigid, sort of like more reflective of how they work today. And every single little thing was building trust with them. You know, somebody posts a message in Slack and says, Hey, like, can I have this feature built? And if next day you can ship it to them, that's a trust-building moment. If there is no unanswered tech support ticket, if there is no feature where we didn't explain why we are not building it, like those are trust-building moments. And you got to do that for a while before you earn the right to change their workflows. Like, it's all the basic stuff, the same way that, like, people used to manage sales teams, like celebrations, Oscars, awards, drinks, in-person events. All of that goes into this, like, one bucket of, like, earn the right, earn the credibility, earn the trust, and get a buy-in to adopt the technology. The third model is a model that we internally call laterals, where you're a partner at a law firm, you have your own book of business, you have 20-year relationships with these five hospitals in Nebraska, and you do immigration for them for the last 20 years. Historically, you've been a partner at a law firm. We tell you, like, here's a better way to run your book of business. You can grow it, and you can make more money, and you're going to have a much more meaningful work. Come to Manifest and they carry their book of business with them. Have you found that each of those value props resonates in different ways with different people, or most of the time it's, I want to make more money, or most of the time it's, there's so much drudgery in serving these customers. If I can do the most important client-centric work, and AI can take that off my plate, I'm going to be so much happier. Like, there is the reasons that they articulate, and then there is the underlying reason that I just realized a couple of weeks ago. So the reasons that they articulate is anything between, Lawyers are good people. And lawyers become lawyers because they genuinely want to help people in the vast majority of cases. And especially immigration lawyers, because, like, think about it. Like, you go to law school, if you can do a M&A and make, like, a ton of money, and then you choose to do immigration, that speaks to, like, your values and your personality. You're here to help people. A big factor is, like, how can I better serve my clients? A real factor. It absolutely clicks and works for them. The second factor is, hey, like, I don't have to do work that I find meaningless. I can focus on a more meaningful work. I can spend—if 90% of my work is eliminated and the 10% of the work is left, I can spend 100% of my time on 10% of work, advising clients, providing exceptional white-glove service, having a phone call with my client every single week pre-scheduled on my calendar. They historically can't do that at law firms. They don't have the time. Proactive updates that I send every Friday to all my clients, giving, like, a case update on what happened in their case. A strategy that I can think through for my enterprise clients of how they can launch a Singapore office more efficiently. Like, there is, like, all this, like, really cool work that they love to do that you, like, give them the time for. But also there is fear tailwind. We are all, as an industry of AI companies, experiencing a combination of some form of fear. Like, lawyers in immigration industry genuinely see how Manifest is winning every single RFP. All of their accounts are switching over to us, and the noise is so loud for them that at this point they see that if they don't jump on this train, it might be too late in the year. And that is a major driving factor for a lot of people, because if they can sort of see, you know, put the trend line into the future, it is obvious that in three years, Manifest Law will be the largest immigration law firm in the world. We will be basically the market dominant. And if they can join today, they're going to be on the right side of the history. And if they don't, they might become irrelevant. So going back, you were going to explain the fourth, which I assume is some form of M&A. M&A. It depends on, like, who you are. If you're a New York-based private equity fund with billions of dollars of assets under management, and your main goal is how do I allocate capital efficiently on scale, M&A might be the right thing to do. But that's not our DNA. We are not flush with cash. We are, like, we've raised a fairly good amount of money, but we haven't raised an unlimited amount of money. And for us, that's not our core strength. If you're a private equity fund and you just do M&A, I don't believe in that. I don't like that. That's not—people genuinely don't want to sell to private equity unless they have to. This is not your most desired buyer, you know what I mean? Like, they're just capital. They don't add anything to the table. Now, there is a form of being a holdco where you have both the private equity dynamic and you have real tech leverage and you have in-house engineers. Thrive is doing an incredible job in that, and you can prove actual tools that you've built for your clients. I believe in that model more than traditional private equity model. But I think my favorite model of all is just you're an actual operator in the space. You know the space. You understand the pain points really well. You've built the best software in the space. You have the best brand in the space. And you use M&A as one of the possible levers. In my opinion, that's my favorite, because you add so much to the table. You bring them exposure to accounts that they can't have on their own. You give them global footprint that they can't have on their own. You give them tools that they cannot buy anywhere else. You give them community that people crave that they can't get anywhere else. And that's kind of what we see. Flipping back over to the technology that you've been building. Maybe could you share a little bit about where you found leverage recently that wasn't possible before? Like, one of the interesting things I think about building your business today is that every month and every quarter, the ecosystem is giving you kind of new magical powers. And sort of in some of that, I would assume it causes you to have to throw entire parts of the product away and kind of rebuild it. And in other ways, it allows leverage you wanted to provide to your attorneys that wasn't possible a year ago is now very easy. Maybe you could talk about some of the big unlocks of the last year, and you shared a little bit about this, but like what the actual product surface area looks like now and kind of how that's evolved. Every week you have a hard technical decision to make. I mean, with Muse, I'm like, okay, like I'm back to like ground zero where we just have a complete breakthrough again. It's just, it's such a fascinating time to build. And all the computer use stuff that underpins something like Muse seems like a huge opportunity for you guys. Yeah, exactly, exactly. It's so interesting. Look, I would say, what is the product surface? Well, three different stakeholders: a company that hires us, an employee that gets the service, and an attorney that provides the service. There are sort of more traditional elements to, like more traditional tech of like all the dashboards and reporting and compliance tracking. There is like a lot of compliance that goes into this, which is still like interface designs and UI/UX. They're wired up, they're agentic in a sense where you can like customize however you see the data and display it however you want by talking to an agent. But that's, I don't think that's like that breakthrough anymore. If I'm a company, what do I care about? I care about full visibility into my program. I care about super fast case initiation. For example, historically, law firms took anywhere between three to five business days to evaluate a prospective candidate that a company wants to hire. That's crazy. Why do you need three days? Like, the market moves so fast. And with AI, you can do it in 20 minutes and just, like, revert back with an analysis. And if it's too complex, then, you know, push it to a lawyer, have a lawyer review it, and then give it back to them in three hours. That's an interesting one. So, like, I want to hire candidates fast. I want all the data flows between my HR systems, applicant tracking systems, and my law firm, everything to flow smoothly. I want to navigate all the policy and regulatory landscape changes. So, in addition to the AI craziness, we're also dealing with regulatory craziness in our business. The visa rules are literally changing every single week right now, which means we're, like, we wired the product a certain way, but a week later we have to rewire it a different way. And traditional law firms are just, like, drowning because they're, like, they cannot keep up with the pace at which this government is changing. Pivoting back to what we were talking about a second ago, you were talking about sort of the technical underpinnings and kind of how it's changed over the last year. Maybe you could sort of build on that a little more. The main evolution is going from rigidity to flexibility and going from human-led workflows to agentic workflows. That has downstream effects everywhere in how lawyers' days are structured. We tell lawyers, you know, AI agents help lawyers navigate hundreds or thousands of different deadlines and conflicting priorities across a massive caseload in a very efficient way and recommends to them, like, specific actions that they should be taking in the next 20 minutes, 30 minutes, 40 minutes. And, like, it's so much better at it than a human. And lawyers love it because when you're operating based on the signal and not based on data, it's very hard to stay afloat, and lawyers are drowning all the time in traditional law firms. They have 1,600 cases, and the way they used to prioritize which case to tackle was which customer writes the angriest email. I swear to God, they literally said that to us in interviews. That was the prioritization. I log into my inbox; whoever is the angriest, I'm going to deal with that. I mean, what kind of prioritization system is that? How do you plan your capacity that way? And we've—AI has just been magical. Just to make it more tangible, if I'm a lawyer, I'm getting a notification, or like, what is the system doing for me? You log in, you see one dashboard, and this is your single source of truth. And this dashboard, in real time, AI agent recommends to you the next step, looking across all your caseload and saying, for this caseload, the deadline is three days away. The last client communications was 24 minutes ago. The client sentiment is decreasing. You should action this case before you action the next case. And it prioritizes your day in real time. In your model, what's the role of a PM? Well, first of all, planning. Planning is definitely like a role of, like, a product manager, of like, how do we—what are the bets that we're making? What are the OKRs that we want to hit with the product this quarter? We set quarterly North Stars, and we say, this quarter, this is the five North Stars that we want to achieve with our product. They're closely tied to business OKRs. That's a very heavy exercise for PMs, like just thinking it through. Then we get into the sprints, and they're not real sprints because we ship like 60 times a day, but like a two-week kind of planning cycle, and within the two weeks, here's the specific projects and initiatives that we're solving. PMs are very involved in planning that. And then let's say there is a feature. Let me give you an example. What's our business? Our business is to enable companies to hire the best people through immigration, right? How do we do this? We need visas and green cards for employees of those companies as companies recruit those people, right? That's the core business. What is a visa? Well, a visa is basically a government tells you, here's the set of rules of, like, how you can bring this specific individual from outside of the U.S. to the U.S. Government publishes their evals, they publish their instructions, they publish their manuals, and they tell you exactly what they want from you if you want to get the product from the government. Historically, lawyers operated in the world of checklists and templates and sort of tasks, and everything was very rigid and kind of brittle, and everything was, like, extremely standardized. And H-1B is like this six things. Now, reality is, like, U.S. immigration system is so complex that every case is different, and the rules for every single case are different, and there is a lot of personalization. So we switched from the rigid systems, which we kind of, like, replicated by following the market, to a fully agentic workflow system, where an agent knows what the end state looks like, because agent is educated on the government standards and templates and requirements. And at any point of time, agent takes all the case context that exists, your personalized specific facts: you were born in Mexico, you lived in the UK, you make $200,000 a year as a software engineer in the Bay Area. And all of those facts in real time define what the next task for an attorney is. And then an agent also acts as an orchestrator across 500 different cases that a lawyer is running in parallel, defining which task for which case should be prioritized right this minute. And the project was called Moonshot. And the Moonshot essentially was just fully switching from rigidity to, like, agentic workflows, where agents define the next step. That was a PM project because there was so much discovery going into, like, how that should work, and there is so much evals of, like, launching this. That we were involved at PMM. Do you think sort of this setting OKRs, planning, you know, having a broad body of work over the next 90 days, is there a path where that gets appended and in a couple of years you'll look at it like that's like the way boomers ran a company in 2019? Or do you think, like, there is something to planning cadence even in kind of an agentic world? That's a good question. I mean, we just reintroduced it. So we had it, then we killed it, and now we brought it back. So why do we do this? Well, reality is, like, if you look at how, like, special forces operate, special forces, the commander of platoon says, This is the objective. This is the goal. This is the why. And then a lower-level commander decides the how. The North Star is meant to be the why, because you can't micromanage the solutions at every single level today. And if there is no one North Star, then, like, teams are sporadically running in a whole bunch of different directions. So to me, this is meant to be a tool for me as the CEO to control the direction of where the product is moving. Do I think that it can be done by an agent? I mean, it is agentic planning anyway, because all the context, everything lives in Linear from every single data source, from Gong recordings of account executives to discovery sessions with existing clients to every support ticket to every email. Everything is fed into the centralized harness, and effectively all of that data is used to plan. But I still have to say a very clear direction: where are we going, and what are the clear trade-offs? I found the exercise of North Star planning less about what are we doing and more about what are we clearly deprioritizing. You mentioned this a little while ago, but— How is sort of this AI-native service that you're building, in what ways is it sold very differently to your enterprise customers? And in what ways is it sort of similar to sort of classic selling any enterprise service? When you're talking to a customer in a sales engagement, they are literally shipping software, right? And instead of this idea of everything has to be a unified roadmap, if one of my customers is Procter & Gamble, I could just have a Procter & Gamble roadmap, which was never possible before. Never possible. Is there anything else that, like, you've figured out as you've been selling an AI-native service to enterprise that's been a real unlock or kind of interesting learnings? Maybe this applies only to the high-trust product, and that might not be universally applicable for horizontal software, for example. But for us, trust is the most important factor. So the things that have actually moved the needle from the go-to-market perspective are the ones that had the most difficult—like, they were the most difficult in terms of measuring the ROI on them. One of the things that we've done, purely out of curiosity, without a real strategy behind it, was we launched the first community for immigration and global mobility managers in this industry, and it just took off and it became the number one community in the industry. It's called Mission. And, you know, every week there is, like, a webinar, every month there is a meetup. This month we're hosting Mission Summit in Washington, D.C., where we're going to have heads of every government agency that regulates immigration in the same room with 50 largest enterprise programs, and, like, director of U.S. immigration, deputy director of ICE, undersecretary of Homeland Security, undersecretary of Labor. We have all these, like, massive government officials who are, like, right now in the administration, sitting in the same room with Mission members. How do I measure ROI on that? Like, I don't know, but I know that it's going to work. But I can't assign a dollar value to it. I just trust my gut that this is what I would want as an enterprise client. In-person motion, game changer. Like, I think selling AI in person is the best way to sell AI. Right now, we've built a real muscle around events. Anywhere between this 2027, we just started planning, but the preliminary number I've seen was 350 events next year. It's a real motion. And it goes back to this sort of trust building. Yep. Get on the plane, meet with them in person, show real proof points. That's the way to sell it. I mean, we have so much noise coming at us, all the email blurbs. I mean, I can't even imagine what your inbox looks like. Like, you probably AI slop coming in every two seconds for you. I don't trust anything that comes in writing to me anymore. I just don't. Like, it's all AI written. All the cold calls, like, obviously don't work. I think, like, any GTM leader that is still like, okay, let me make—I'm going to hire a bunch of BDRs, make a thousand calls, I'm going to have 2% conversion rate, I'm going to send 2,000 emails, have 1% conversion rate. I just, I don't like marketing that way. I just find it stupid. Like, make actions that build trust, create brand visibility, do something that nobody else is doing in your industry, something that's going to have the right psychological effect of establishing you as a winner. And this is things that are hard to measure, that are difficult to execute on, but they're the ones that are actually moving the needle, in our case. Yeah, makes total sense. What are you seeing in terms of, like, how your portfolio companies are handling go-to-market right now? One is that I think that the marketing function has never been more important in technology. I'd argue that if you look at sort of vertical AI companies that are selling to enterprise, all of the leaders have cracked marketing. They both invested a lot. They've, to the point that you made a second ago, they've found weird values-aligned, mission-aligned ways to market themselves. They've invested in marketing at subscale like you would imagine a Fortune 100 would do marketing. Like they are taking people and/or sponsoring the U.S. Open. They're doing all of these things that you would imagine you do in year eight or nine, in year two or three. In any one of these categories, there's a brief window where you can be the trusted brand that your customer goes to you to figure out what is happening as this kind of technology is changing. And I continue to think the most valuable place to be is when your customer has a problem or opportunity around your area of business. They come to you and they say, Hey Dan, how do I solve this? And I think one of the ways you do that is by building the trusted brand. And there's many ways of doing it. To your point, I think in-person is incredibly important. Events are incredibly important. But generally, just the marketing function and building the brand that everyone looks to very, very, very early on is one of the most important things. I think hiring marketing talent has never been harder. If you look at now, Ramp started sort of, call it pre-ChatGPT, and they've kind of transformed their company. But if you say what's one of the things they do better, they are in their category, there is no one anywhere as close in marketing ability as them. And I'd argue when you think about every category leader right now, they are all the best at marketing. And I feel like 10 years ago it would be founders-led. Would be founders-led sales. Some sort of demand gen function has to be built. But when you think about real mature marketing, telling our story, getting in every channel, being weird and authentic to who we are, that wasn't a thing as much 10 years ago in B2B. In consumer, it was. Right. And I'd say that's probably the biggest, on the go-to-market side, sort of the biggest change. I mean, that makes sense because those marketing channels just got saturated to a degree where just there's too much noise. And I don't think you can just do email marketing to the point that it's all clogged now. And so you have to find a way to cut through. And I think you have to find a way to cement yourself as the trusted place. As technology has changed in a post-AI world, there's even more the opportunity to be the trusted sort of consultant to your company, to your customer. Because what you can provide to that customer today and tomorrow and in six months and in six years is so much more expansive. And I think that even exists in mid-market. Like, I think it used to be that if you have a Fortune 20 account, you could have three people on your team that have this very consultative sales motion, these deep relationships. But now a mid-market product with an ACV of 75K or 200K can do so much sort of consultative work for that customer that there's a whole sort of dynamic where it's always been that the best selling, I think, was consultative selling. It wasn't, I need to hit my commission, let me sell this thing to you. It's, let me understand what's going on for you. Let me solve problems for you. I think that's like all of good sales now. I have a split mind on this. Part of me listens to Alex Karp and says, if you build the best freaking product, you don't need sales. And I kind of like, it really clicks for me because it's just like being a true product-led organization with, like, exceptional product. But the second part, I totally agree with that too, which is like, Arnold Schwarzenegger might not have been the best bodybuilder. He was the best promoter. He was the best speaker. He knew how to get media's attention. And that applies universally to, like, any type of company. And the question I ask myself, I asked myself that question last night at 1 a.m. when I was, like, just finishing work. I was like, how much time should I be as a CEO spending on, like, the sort of outward-facing work, inward-facing work? How do I reconcile the desire to build the best product in the world with the desire to be the best-known brand in the world in my category? You want a great product and you want to be the trusted brand. But the trusted brand is really, really, really hard to create. And once you have it, if you think about how power is developed in any business, scale economies, you know, you go through all those things. The most misunderstood one, I think, is brand. And one of the most powerful ones, if not the singular most powerful, a la Coke, is brand. And so it has me—not that the only thing that matters is brand and marketing and we're going to have crap product. My general worldview is that wherever you are operating on the brand distribution trust versus obsessing over product, more units of attention resources need to be put into... Need to be put into brand and trust. Maybe you can make the case that the product could be 25x better. And in that case, maybe it really needs to go into product. In most cases, if you think about the next five years, I think the best product in the category and the second best, maybe it's 2x better, maybe it's 50% better. I don't think it's literally 25x better. And so if that's the case, I think you build your compounding advantage around trust and brand, and you become Coke. You become, in the last generation, what would have been IBM in 1975, you know. And that, for this generation of founders, is much less natural. This generation of founders, like the last one, are product and technology oriented. But I don't know. I mean, if you think about the last generation of companies that were built, one of the best ones is objectively Stripe. And I think one of the most misunderstood things about Stripe from a distance and all the people that I know there is they built an extraordinary brand, and they obsessed over it from the early days. I think that insight is just going to be blown out 10 times as valuable over the next few years. Because AI cannot automate trust, human relationships, and building brands. No, I totally agree with you. And even, by the way, I think AI is going to be very good at creating marketing assets and all of this. We're getting there. But when you think about extraordinary companies that do marketing and sales and distribution the best, it's so much more than the asset creation. A hundred percent. You've been building companies since you were a kid. Your last company was an eight- or nine-year stretch in the 2010s. Now you're building a company that's built on top of this technology transformation in AI. When you reflect on what is required for you to be successful, or what being an effective founder and CEO in this environment was versus being an effective CEO in 2017, in what ways do you think it's most similar and what ways do you think it's really different? And in what ways have you had to grow and stretch and kind of reinvent yourself? Look, I mean, I don't think of myself as the most effective CEO, to be fully honest. And I'm certainly not the best manager in the world. That's the part that I hate the most. The part that I love the most is the vision, the product, and going hyper deep into specific problems. That's what gives me the most energy. And the least energy is doing one-on-ones with people. I hate that part. Which they probably don't like either. If you don't like the one-on-one, they probably don't. Yeah, like if you have a question, just let's have a five-minute conversation. Like, we don't need to block 30 minutes on my calendar. Like, you know what I mean? I think relearning skills and just, like, beginner's mindset. I mean, how do I say this so it doesn't sound cliche? But I think having a beginner's mindset. Like, I am constantly feeling like my work is not good enough. I feel that way every single day when I wake up in the morning and when I go to bed. I'm constantly putting myself in a situation where I'm like, I'm challenging myself so much every single day. So I think, like, just this beginner's mindset of, like, discovering new parts of the company. Like, I've never been technical. Like, I've been a business person in every company that I've built. And, like, this is the first company where I became more technical than I used to be. And, like, I'm loving it. The learning curve for me, the amount of insights that I bring. And I realize that it's not that, like, a lot of the work that I do today... And I don't write code, but like it's more like management of the product and the technology. It's not that hard. They make it sound hard, but like once you actually go deep, like it's not rocket science. Like it's just basic human dynamic and like planning and like making decisions. I push myself for mental clarity. I push myself to be calm more than I used to be. I push myself to eliminate noise because the amount of noise just became so great in the last two years. Everything is exploding. Everything is urgent. There is like noise everywhere. But I don't know if I can attribute it to the age of AI or to just like my personal evolution. The kind of people that I've hired at Manifest are very different than the people that I had at June Homes. They're autonomous. Most of them were founders or entrepreneurs in some shape or form. And they need to be managed very differently. If they need to be managed at all, I don't think they need to. What's different about the people you used to hire and the people that are working at Manifest? Well, I didn't know how to hire in my previous companies. I was just sort of making mistake after mistake. I still make mistakes, but here at Manifest, like the caliber of talent is just unbelievable. I walk in the room and like I remind myself that like these people are so freaking smart that like they are probably smarter than me in a lot of different ways. And then I ask myself a question: what is my role in this room? Like if I have four really smart people working on the problem, what is my role? And I find my role to be mental clarity and simplicity and sort of full alignment on the vision and decisiveness. A lot of very smart people are very indecisive, is what I'm finding. They're overthinking everything. They're spending, they're going in weird rabbit holes. And maybe because I'm not as smart as they are, I just don't get to go into those rabbit holes. I like yes, no, like quick decision making. And I found that clarity and sort of decisiveness. Even if you make the wrong decision, 90% of them are two-way doors anyway. So you're like, I just like walk in the room and I'm like, how can I bring clarity to this group? And how can I make decisions to unblock everything? And I pushed myself to do this. What was it like when you were running your last company in contrast to that? A lot of micromanagement. It was just doing everything yourself? Or micromanaging people, not trusting people, not giving them autonomy to execute. Why is that such a natural thing, do you think, for first-time founders? Like I'd say it's the number one thing. You manage everything. Everything's dictatorial. Everything is, this is what you're going to do. This is how you're going to do it. This isn't good enough. I'm going to go redo it myself. And then you see people that are in the role a lot, and they're in the details. They're not like a Fortune 500 CEO up in the clouds. Right. Like that. Founders can't behave that way, generally speaking. But they spend a lot more time on alignment. They spend a lot, a lot of time on clarity, clarity, clarity, clarity. They spend a lot of time getting the right people in the right seats. They spend a lot of time on making sure the company's focused on the right problems. But they're not doing everyone's job. They're not, you know. Look, I mean, you get punched in the face enough times and you kind of learn the lesson, you know what I mean? Like when you lose somebody extraordinary because you're a micromanager, like, and it happens four times in a row, like you kind of learn it over time. I mean, that was my case. Look, I thought, but I'm also maybe very different than a lot of, like, people that went to Stanford and they're super smart and they maybe they don't have the same learning curve as I do, where they have to make all these mistakes to learn. No, I think they do make a lot of the mistakes. I mean, what are the common traits of people that start companies? You know, like most of my friends are founders. Most of my friends are founders and, like, you know, often founders can be a little narcissistic, and we tend to think that we're smarter than everyone else, and we have a very big ego, and we have some unresolved trauma that makes us do what we do and eat glass for breakfast, lunch, and dinner every day. I think a lot of it stems from there. Like, I know in my case, I was just thinking that I'm smarter than everyone else for a very, very long time, and I wasn't being a complete jerk in a lot of ways in many of my companies that I've built before. Did you evolve and grow out of that gradually over time, or was there some critical moment for you that caused you to sort of phase shift? Crises and a lot of mistakes and a lot of, like, just—look, and I'm still learning. Like, I'm 34. Like, I'm still fairly young in my career. If I'm still building companies in 30 years, what I'm like, and what am I going to learn between now and then about, like, how to do this? I would assume it would be very sad for you if there wasn't much to learn between now and then, and you were just kind of doing the same thing. Okay, that's a very good way to put it. Like, it's a very different perspective, because, like, every time I make a mistake, I'm like, What an idiot I am. Like, why did I make that mistake? I could have foreseen this. But you're right. It is part of the journey, right? Yeah, there's this really good Sam Harris quote a while ago, which is like, he was talking to a friend about all these problems that he has, and this didn't go well, and this and this. And the person's point was that that's it. That's it. That's life. Like, imagine right now all your problems are gone. You know, it's one of the issues, I think, when people talk about UBI and where we're going with AI, it's like— Imagine you're just on vacation all the time. Most people, you're on vacation, there's not a single problem. The flight's not delayed. Everything's perfect. You have no problems. Everything is perfect in your life. People might say they want that life. Nobody actually wants that life. And so the essence of life is not the absence of problems. The essence of life is working through problems. And then once you've finished this problem, the next one comes. And that's the whole thing. I guess. I guess so. The most common trait amongst every great founder is every year they're a different version of themselves. And one of the bottlenecks of great founding talent is most people are not like that. Most people you meet on January 1st and you have dinner with them on December 31st, and they've grown and evolved, but they're not like an entirely different. And that's a big part of, I think, what it is to be a founder in a lot of ways. Maybe what's different for—there's many things that are different between founders and managerial CEOs, but I think that personal rate of improvement is one of the biggest ones. That's it forever. It's just that compounding, you know. Well, that's a good place to end. Thank you for joining and coming on. This was so much fun. I really enjoyed it. It was really good. Cool. It was so much fun. Thank you so much. I really liked the conversation.