Dharmesh Shah, co-founder and CTO of HubSpot, joins Sam and Shaan to compress 20+ years of wealth-building lessons into a single episode. He traces his journey from rejected Pizza Hut applicant to billionaire, covering leverage, being close to value creation, the power of writing, AI agents, his sale of chat.com to OpenAI, and the emerging “results as a service” model. The episode is structured as a prepared talk with Dharmesh’s characteristic analytical precision and genuine self-deprecating humor.
Speakers: Sam Parr (host), Shaan Puri (host), Dharmesh Shah (guest, co-founder and CTO of HubSpot)
Opening: Dharmesh’s Prepared Remarks Setup [00:00:00]
Sam: You’ve had a good quarter. HubSpot stock, I think, hit an all-time high. You sold a domain to OpenAI. You came with a pod. You know what, I don’t care what they say about you, Dharmesh — you’re doing okay.
Dharmesh: I’m doing okay. I’m doing okay. Thank you.
Sam: Okay. It’s My First Million. How do we use the hour so that it’s your first million?
Dharmesh: If all you do is spend 100% of your time converting your labor into value and do not increase your leverage, you’re not going to get anywhere.
Sam: So what were the things you invested in that had maybe the highest return — that did actually make you more high-leverage?
Dharmesh: Never in my mind has it been easier to get to that first million than right now.
Sam: There was no good way to squeeze this in, but I can’t not say it. I’m going to — I know I’m going to regret this…
The Unhidden Agenda [00:00:45]
Dharmesh: Okay, let’s start with opening remarks. The first is what I’ll call an “unhidden agenda.” And my unhidden agenda — so I have a picture of a Venn diagram. I love Venn diagrams.
Circle number one: I want to talk about things where I think I have something useful to say, some expertise. So we’ll definitely talk about AI and agents and things like that.
But the other circle is things that I think are borderline guaranteed to increase the probability of you — the audience — making your first million.
Thing number two: in order to do that, at HubSpot we have a culture of humility — that’s one of the core five values. And I said this particular line to employees at HubSpot, I don’t think I’ve ever said it outside of HubSpot, but I’m going to: solve for utility over humility. So if there’s a way to say something that’s going to be more useful — even if it’s non-humble — I’m going to say it. And there will be a couple of moments that are completely non-humble. So that’s the setup.
Sam: Okay, so we have to start with this: you’re coming on as a guest, but you’re coming on as the most prepared guest we’ve ever had. We did a pre-call, we discussed a bunch of things, then I sent you some notes. I tried to make it light — like, just maybe one or two things to think about. Because I never know how much work somebody wants to put in. You then sent back a whole other doc. You had opening remarks, a bunch of stories and ideas. You went back and watched your old episodes — which are some of the most popular episodes we’ve had. You read all the comments. You incorporated the feedback. Is this just how you do everything?
Dharmesh: Not everything. So, confession: in this particular case, not only did I watch my prior episodes and read through the comments, but in the process I saw the leaderboard on the My First Million YouTube channel and I’m number five.
Sam: Right.
Dharmesh: And you don’t know this about me yet, but you will — I’ve never met a leaderboard that I didn’t want to get to the top of. It’s like, okay, what happened here?
Shaan: By the way, Sam, that’s what — when I was like, so are you coming on, what do you want to talk about? AI? That’s obviously an area you know a lot about. And he was like, yes, but through the frame of: okay, it’s My First Million — that’s the name of the podcast — how do we use the hour so that it’s your first million. So if you’re listening to this, he’s like, I can’t guarantee you’re going to do it, but I can increase whatever that probability was. Can I in one hour tell you things that will increase your probability? That’s the promise.
Sam: Someone made a comment on a recent episode — Sean and I both just sat there for like eight minutes, and someone was like, “Dude, Sean and Sam are just flirting — they’re just staring in awe.” I have a feeling we’re going to be getting a lot of that today.
Dharmesh: Yeah, I’m going to get my hands ready.
Sam: All right, so where should we start, Dharmesh?
Early Career: Red Roof Inn and the Leverage Framework [00:04:30]
Dharmesh: So I’m not going to give you an autobiography, but I am going to take us back in time a little bit, because in order to understand the lessons that were most valuable — I didn’t know this at the time I was living it, because we often can’t pick that up. But I think it will be useful.
So I’m going to go back to when I first immigrated to the US. I was in my early twenties, here just on a visit to my parents who were living in Indiana. I applied for a job at Pizza Hut — rejected. Applied for a job at Big Lots — rejected. Applied for a job at Red Roof Inn, and because I was Indian, and my parents were actually running a motel, they automatically assumed I knew things. I’d only been in the country for like three and a half days.
Sam: Indians are famous — not hoteliers, but motel-iers, right?
Dharmesh: Yes, exactly. We had a motel. Not even a franchise — one of those independent ones. Anyway, it was the night shift — 11:00 p.m. to 7:00 a.m. That’s the only thing that worked for me because I was taking classes during the day.
So a couple things out of that experience, and I think this will relate for a lot of folks. In the early parts of most people’s careers, you’re working retail, you’re working some sort of job, and you have what I call your currency — the time value of your time. I was making something like $3.65 an hour in those very early periods. And you take that currency value, multiply it by how much labor you expend, and that’s effectively how you create money in those early years.
Mathematically, in that equation, there are only two ways to make more money: work more hours, or raise the currency — raise the price people are willing to pay for your time. Back then, I was looking for all the hours I could get. It’s like, I want more hours — if someone cancels, I’ll be there, just put me in, coach.
Sam: Can’t believe you made a sports reference.
Dharmesh: I don’t do sportsball, but anyway. It worked.
So here’s the lesson: you start there, and you’re going to spend a large part of the early part of your life — call it the first half — converting time into money in various shapes and flavors. And then you’re going to spend the latter half of your life desperately trying to convert money back into time. That’s life.
There’s also going to be an automatic increase in your currency simply as a result of tenure. Companies pay more for experience, even if you’re doing roughly the same thing. You’ll get some marginal increase — not much, but it’ll go up.
My argument is: in order to really break out, you’re going to have to accrue leverage. And I mean leverage in the Archimedes sense — not the leveraged-buyout sense. The Archimedes quote: “Give me a lever long enough and a fulcrum on which to place it, and I shall move the world.” What’s interesting about that quote, and this is a physics thing, is the fulcrum is very necessary. The degree of amplification you get for your force is determined by how far you are from the fulcrum.
So the lesson is: if all you do is spend 100% of your time converting your labor into value and do not increase your leverage, you’re not going to get anywhere. You have to allocate some percentage of your time — even though nobody’s paying for it — to read a book, do a seminar, meet a friend, whatever it is. You have to carve that out. Otherwise, you never get the leverage necessary to make your first million.
The thing I did that was very useful, that I still carry to this day, is I literally carved out a percentage of dollars for myself. I mean time and money are equal to me back then — I took 10% of all the money coming in and spent it on books, on things that would improve my value. Even in my first job, when they wouldn’t get me a fast enough computer, I just bought it myself. I wasn’t looking for approval. If something was in the way of improving who I am and increasing my currency, I would spend the money. I wouldn’t expense it.
Sam: What were the things you invested in that had the highest return? That did actually make you more high-leverage, more valuable, increased your rate — what were the top investments you made?
Dharmesh: Books were the highest ROI. There was an author called Harvey MacKay who wrote some very pedestrian business books — by today’s standards, not sophisticated at all. But at that age, for me, it was like, “This is brilliant.” One of his books was titled Swim with the Sharks Without Being Eaten Alive. There was another one: All I Really Need to Know I Learned in Kindergarten — very basic things. But I felt like an alien from another planet for a large part of my life. It’s like, oh, so this is how the world works. It seems shockingly obvious now. It just wasn’t obvious to me at the time.
US Steel and the “Don’t Be Overhead” Insight [00:12:00]
Dharmesh: All right, so after Red Roof Inn, I was able to get a job at US Steel.
Sam: No way. That’s pretty cool.
Dharmesh: They had a computer there which was one of the nice things — I had to close the books for the accounting, which was some of my early exposure to computers. And I was able to get an actual software developer job there while still going through school, working on my undergrad.
The US Steel plant was in Gary, Indiana. Gary Indiana was at the top of the list of places you did not want to be. Indiana overall was fine, but it was very north and very cold, and I had just come over from India. So I did a very me thing: I figured out US Steel has plants elsewhere. As it turns out, they had one in Birmingham, Alabama. I didn’t know where it was, never been to Alabama. But I knew it was more south than Gary, Indiana — and the mathematical odds of it being a worse place were slim to none. So I requested a transfer.
I get to Birmingham. I’m working for US Steel, making around $27,000 a year — a lot better than Red Roof Inn. A full salary, which was awesome.
But the epiphany came from something my manager said. He said it in almost these exact words: “Look — if you’re not making steel, shipping steel, transporting steel, or selling steel, you’re overhead.” And it hit me: I don’t want to be overhead. So I thought — my thing is software development, where do I need to work where I’m not overhead? The extrapolation was: I need to work at a company where software is the actual product, the same way steel is the actual product at US Steel.
So I opened the Sunday paper — this is pre-internet — and looked for jobs at software companies.
Sungard Data Systems and Compressing the Negotiation [00:17:30]
Dharmesh: I applied for a job at a software company that had an ad in the Birmingham paper — a place called Sungard Data Systems. A bona fide software company. Got that job.
The lesson here: your value is inversely proportional to your distance from the actual value creation. If you’re not close to it, find a way to get there. My manager had said you have to be making steel, selling steel, or shipping steel — everything else is overhead. If you have a chance to get closer to the actual customer, the actual product, take it. That will increase your leverage, increase your currency.
So I’m working at Sungard as a software developer, and I’ll share a high-utility story. Their software product was a mainframe product — character-mode terminals. They were trying to build a graphical user interface that interacted with their mainframe. My job was to take character-mode screens and create a Windows GUI equivalent. There were hundreds of screens. And we had consultants on there making $125 an hour to do this work that I thought of as relatively rote.
Two months into the job, I go to my boss and I say: we’re paying these guys $125 an hour, and all they’re really doing is drag-and-drop based on the character-mode screen. This is so trivial that my brother — who hasn’t graduated high school yet, works at the Piggly Wiggly — even he could do this in a day. Why are we paying $125 an hour?
Sam: That’s a bold thing to say two months in.
Dharmesh: My boss said, “Bring him in.” So I brought my brother in. He’s 17, still in high school. I showed him what we’re doing, he said okay, and we started cranking. We paid him $5 an hour. Then I hired all my classmates from undergrad as the team grew.
I kept going back to management: what do I have to do? I was still an individual contributor, not managing anyone. But they were super generous — they kept bumping me. I was only there about 14 months. They ended up paying me around a quarter million dollars a year in Birmingham, Alabama, in the early ’90s, after a year and a half.
At that point I felt like it was wrong to keep going back to them — they were paying me really well. So the thought arrived: I need to start my own software company. I had no product idea. It’s just: this is the thing I need to do.
In one of those discussions with my manager, I’m going back again — I’m maybe 25, 26, 27 — and the pay was getting to be somewhat ridiculous at this point. He came back and said, “Dharmesh, do you really think you’re going to find that many companies that are going to pay you $200,000 a year?” And the line I came back with — this came to me in the moment — was: “As it turns out, I really don’t need to find that many companies that will pay me $200,000 a year. I just need to find one. And I think there’s one out there.”
It wasn’t a threat. It was a very measured, logical, Spock-like response. We’re not solving for what’s the local maximum in terms of how many companies will pay this. I don’t need a hundred. I just need one. And I believe I could find one.
Sam: You agreed with him in a way. He’s like, do you think you can find 100 companies? And you said, no — but the good news is, I don’t need 100. I need one.
Dharmesh: Exactly. And that applies to dating, to business — making one business work will do enough to change your life.
There are two lessons from that moment. One is the power of negotiation. My most recommended book is Getting to Yes from the Harvard Negotiation Project. I’ve read it about four times, reread it every few years.
Sam: Quick insight for anybody who hasn’t read it — what’s an insight that would make me a better negotiator?
Dharmesh: Most people think of negotiation as adversarial — a zero-sum game. My objective is to get the lowest price, their objective is the highest price, and we’re just fighting over a number. That’s not the right way to think about it.
The right way is to identify your actual needs — what is truly the thing you’re solving for? — and do your best to identify what the other party’s needs are. Because there is very often a path that actually optimizes for both of you. That’s actually a better outcome than just fighting over price. There might be something else that’s more important to them that you haven’t discussed.
A tangible company example: we automatically think what people are solving for is compensation. Number one. But as it turns out, the number one thing people want is actually flexibility. They would trade a lot of other things in exchange for flexibility. If you understand that, you can trade off other things that might cost you less.
Sam: When Jeremy Giffon came on the pod, he said the same idea. He goes: I used to think negotiating was like two people sitting across a table — I want this, no I want that, adversarial. He goes, Chris Sparling told me to imagine you’re both sitting on the same side of the table, looking at the problem across from you. You’re not adversaries — you’re both looking at the same problem and figuring it out together.
Dharmesh: That’s it. Let me give you a tangible situation where that really matters. Let’s say you’re negotiating something that’s hard to price — a unique property, a house with no comparable sales. The temptation as a buyer is to point out everything wrong with it to drive down expectations. But the visceral reaction from the seller is: “This guy doesn’t appreciate the value. I need to find someone who does — and I’ll do better on price with them.”
Now flip it. I go in and say, “This place is fantastic. Let me tell you the things I love about it.” I go through all of that — and it has to be authentic. Then I put my price out there. The seller has to say: okay, I don’t really know the objective market price here, this person seems reasonable, they love this thing, maybe I’m not going to find someone who loves it more than Dharmesh does. So his price is probably fair.
It takes people’s guards down. This only works when it’s a unique asset with hard-to-find comps, but in those cases it’s powerful.
Sam: That’s a great negotiating tip of the day.
The Embarrassing Second Startup [00:28:00]
Sam: You told me something great in our pre-call, about your second company. You said: I did the first company, did fine — but the second one is my embarrassing one, the one where things didn’t go well.
Shaan: The story he told was: I did the first company when I was completely clueless, didn’t know anything, just making it up as we go, super scrappy — and it kind of worked. Then I said, okay, now I’m going to do the second one. I’m so much smarter now. I have experience. I’m going to do it the right way. I’m going to raise money, hire better people, get a better office — do it proper. And then that business didn’t work. But you said a great line: “Just because I was ignorant doesn’t mean I was wrong.”
Dharmesh: Right. If there were an objective truth function — say it evaluated every decision I made in that first startup — I think my hit rate was actually pretty high. Because our natural instincts in that situation turned out to be good instincts. Being resourceful, not spending too quickly, taking your time — things that came as natural instinct. I was probably not a natural-born entrepreneur, which is likely why I have some insecurity about it. But what was definitely suboptimal in hindsight was second-guessing myself the second time. Like, I’m going to do the opposite of what I did the first time because I was such a chump back then. I’m going to do a speedrun. I’m going to write a $500,000 check to myself on day one.
Sam: I remember a time in high school — I had this hillbilly friend who had a huge chest because he bench-pressed all the time. We went to an exercise class and the teacher was like, “The best way to get strong in your chest is incline bench press at exactly 30 degrees, not 45, not 10.” And my friend who didn’t know anything, he was just lifting — he’s like, “I must have been doing them at 30 degrees because I’m strong as hell.” It’s the same thing. It doesn’t matter if you knew what you were doing or not, you got the end result.
Dharmesh: That’s right. What was your second company?
Well, yeah — so it was kind of “Facebook for small business before Facebook” — effectively a CRM. I’ve been working in CRM now for a very long time. This was right around the 2000 timeframe, as the bubble was bursting. My first company was not an internet company. This one was a web-based information management tool for small businesses around customer management.
Sam: What was it called?
Dharmesh: It was called Captivo. C-A-P-T-I-V-O. How much did you raise, and how long did it last?
Because I was going to do things differently this time, I wrote the first $500,000 check and kept writing checks. I think I put in about two million total.
Sam: I Googled “Captivo Dharmesh” and found a PDF — it looks like a case study for Sloan. And you’re trying to make a point. It says: “In 1999, I founded my second startup called Captivo. In many aspects it was similar to Salesforce. About two years in, with over a million dollars of capital invested — mostly my own — Captivo still could not gain any significant traction, and was ultimately sold.”
Dharmesh: That’s right. And the only thing you can find about Captivo on the internet is a case study I helped write. The internet wasn’t as big a deal then — I wasn’t blogging, no one was blogging. So yeah, it existed, it just didn’t go anywhere.
What ended up happening, to close out that chapter — I’ll make my third sportsball reference, and it’s a golf one: it was a long putt to par. What happened was I took Captivo and merged it into my first startup, which I hadn’t sold yet. I was running two startups at the same time, which is not something one should do. Then I ended up selling the merged entity. So I basically got my money back. It didn’t end in disaster, but it was a very long bogey-to-par.
Recapping the Lessons So Far [00:36:00]
Sam: Okay, so the big ideas so far: at the beginning, you trade time for money, and then at some point in your life you desperately start trying to trade money back for time. That’s life.
Second: you can either get more hours or increase the value per hour, and that’s where you started investing in books and training — whatever you could do to increase your own value.
Then at US Steel, you realized you’re either an asset or a liability — either overhead or actually creating the value. If you want to be high-leverage, you’ve got to be closer to where actual value is created.
Then the lesson from the negotiation: don’t just ask for more — ask what it would take. Frame it simply. “I don’t need 100 companies. I just need one. And I think there’s one out there.”
Dharmesh: The key lesson from that moment is the power of framing something in a very simple, inarguable way. That one sentence worked because it was punchy and undeniable. Had it been a five-minute back-and-forth argument, I don’t think it would have worked. This is what I call — I just made this up — “inside compression.” Can you take a big idea and boil it down until the essence is captured in a dense distillation? The simpler you make it, the more likely it is to be transmissible, to be communicated, to do its work. This applies to copywriting, to marketing, to venture pitches, to so many things.
Writing as the Highest-ROI Skill [00:40:30]
Shaan: I think one good point to make here is that this isn’t just a mindset you had when you were young and broke. You took my power writing course a few years ago. You’re a founder of a $20–30 billion public company, sitting in a course learning how to write better. You’re still doing this — not just when you were 20.
Dharmesh: Right. A guy came on the podcast the other day — Mike Posner, the musical artist. He talked about how he had his first song go huge. He thought, that’s what I do. His second song was still double platinum, but felt like a failure. Third song, single platinum. Then the studio shelved him for years.
And I thought he’d say he was just depressed eating Cheetos on the couch. And he was — for a while. But then he enrolled at Berklee College of Music. He was learning singing, instruments. Learned to play guitar. He’d go to these classes with college kids — Grammy Award winner, been on the charts, made millions — and there were better singers in those classes. He found it a real mind-bender.
But what I took from that story was: he used that period to sharpen his skills. I think a lot of adults just stop. They think, that’s stuff you do when you’re young. You just don’t need to invest in learning anymore after that.
Shaan: So about writing as a return-on-time investment — what’s your view?
Dharmesh: I have not found the ceiling on writing ROI yet. It is just so high. I cannot describe to people: if you could do nothing else, let’s say you had five hours to invest in the next month — you could spend it all learning how to write well, and future you will look back on those five hours and say, “That was a great use of time.” Because writing is an amplifier. It will make you a better thinker, better communicator, better pitcher, better salesperson, better at everything.
Agents: Why Now Is the Best Time for Your First Million [00:45:00]
Sam: I knew that you bought chat.com, and I’ve known you’ve been thinking about agents. Seems like that’s a big thing to you right now. Speaking of being close to the action — you bought chat.com. You said something like $10 million, maybe eight figures. Then two weeks ago it came out that you sold chat.com to OpenAI for around eight figures. Give us the update.
Dharmesh: Okay, a couple of things. When I went on the podcast the first time, it was within like 72 hours of the purchase — it wasn’t fully distilled in my head. But actually I did have a plan. I was going to build a chat application on top of the GPT algorithm, similar to ChatGPT. This was one of two times I accidentally competed with OpenAI, which I don’t advise anyone to do. Sam Altman is near the top of my list of people I’d never want to compete with. Too smart, and even more red-blooded capitalist than I am — and I mean that in the most positive way.
My original thought: OpenAI put out ChatGPT as a demo app. It’s there to demonstrate the power of large language models. But OpenAI is a platform company. Someone should actually create the end-user application that sits on top of the LLM. Then it just so happened — the universe configured itself — and chat.com became available for sale for the first time in about 30 years.
Shaan: When you came on the pod the first time you had just bought it and were like, “Not totally sure what I’m going to do with it. I think it’s a great domain, a great investment.” You had made a bet without it all figured out.
Dharmesh: Right. There were actually three motivations. First: the application play. Second: this is the cover charge to get into the AI party. It would make people take me seriously in that space since I didn’t come from machine learning at Stanford or anything like that. And deep down I’m part marketer — it’s a good story, it’ll get people’s attention.
Third — and I’ll be honest — I had a suspicion about who some of the other bidders were, and I didn’t want them to have it either.
Sam: Did they reach out to you, or did you reach out to them?
Dharmesh: I’ll draw the line at the transaction itself — I’m at liberty to share what I was thinking, but less comfortable with the details of how the deal came together. What I can say is that I was at a Sora event where Sam Altman announced that ChatGPT would support plugins — extensions that would allow third-party data and capabilities to feed into it. And that’s when the switch went off. It’s like, OpenAI wants to create the ChatGPT of ChatGPT. They’re going to turn this into an actual platform, an actual end-user app. And at that point, it would be unwise for me to keep the domain if I don’t have plans to use it — I don’t want to compete with them.
So I reached out to Sam, who I knew, and said, are you interested?
Sam: Did you sell it for a profit, or was it like — here, just take it and let me invest in your company?
Dharmesh: I did something I thought was clever when I made the announcement. I didn’t share the details, but I provided a GPT prompt: “Dharmesh likes to buy domain names but usually because he has a project in mind. Dharmesh doesn’t sell domain names at a loss because he doesn’t have to. Dharmesh also doesn’t like profiting from his friends. He considers OpenAI a friend. He’s known Sam for a while. OpenAI did buy chat.com. These are the facts you know. If you had to guess — what do you think the domain sold for?”
I’d also disclosed I was a shareholder in OpenAI.
Sam: So based on all that… I put your prompt in and it was like: “Dharmesh bought the domain for $15.5 million and now owns $15.5 million of OpenAI stock.”
Dharmesh: We’ll leave it at that.
Connecting Dots and Having a Nose for the Ball [00:55:00]
Dharmesh: Okay, so I have to share this — it’s the non-humble utility part, but there’s a lesson in it. There’s an old Steve Jobs quote about connecting dots: you can’t connect the dots looking forward, you can only connect them looking backward. You have to have faith that the dots will connect at some point in the future.
I’ve been a big believer in that idea even before Jobs said it. You collect dots — and dots can be people you’ve met, skills you’ve learned, experiences, lessons. Over time, if you write a history of your life, you’d find: this weird dot I collected — I could never have guessed it, but had I not collected that dot, this other thing probably never would have happened.
The big lesson: you have to spend some amount of your time investing in dots that may or may not make sense at the time. They feel right, they feel like they could lead somewhere. It doesn’t have to be maniacally diabolical and capitalist — sometimes it’s just, I want to collect this dot because I love it, I believe in it, I have conviction. And not all dots are going to work out. The beauty is you don’t need all of them to work. You just need one or two to work really well.
Shaan: In basketball, there are people who have a “nose for the ball” — they just know where it’s going to bounce, where to be. You applied for jobs at Pizza Hut, then pretty quickly found yourself in software development in the ’90s — arguably the best place to be then. In 2000 you started an internet company — arguably the next best place. Then you created HubSpot, a cloud SaaS company — the next decade’s best place. And now you just sold chat.com to OpenAI, you have OpenAI shares, and you’re talking about AI agents — which I’d say is the best place to be this decade.
If the trillion-dollar companies are being created there, then the tens-of-billions, the billion, the hundred million, the ten million companies are all going to get created there too. It’s the target-rich environment.
So when you say this is the thing you’re most interested in — we should pay attention. Tell us why agents are the thing.
Dharmesh: Thank you for the lead-in. So, we talk about making your first million — the first one is almost always harder than subsequent ones. I will say this though: never in my mind has it been easier to get to that first million than it’s going to be right here, as we’re living our lives right now. And a big part of that is what agents unlock.
So let’s talk about agents — what they are, why it’s such a massive opportunity, and what you should do about it.
What AI Agents Are [01:00:30]
Dharmesh: Okay, step back: last year was all about chat. Everyone’s used ChatGPT — you type in a prompt, something comes back. That’s a very interactive approach to using AI. You type a prompt, it comes back with a response, you do follow-ups. The tasks you assign AI in that model are generally discrete: give me a blog post, generate an image.
What agents are — very simply — is AI software that can accomplish higher-level goals requiring multiple steps. Not just “give me a blog post and it comes back with a blog post.” It’s: I want you to accomplish this thing, and that thing may require ten steps. Each step might need to functionally decompose everything in order to accomplish the high-level goal. It has to have memory, it has to do all these things in order to make that possible. That’s what an AI agent is.
Right now: agents are the new apps. It’s just software. When mobile came along, there was an app for everything. Not that far in the distant future, there are going to be hundreds of thousands, millions of agents — the same way there are lots and lots of apps. Think of an agent as an AI app that can accomplish high-order goals requiring multiple steps.
Dharmesh: Here’s my view of how this is going to shape out. There’s lots of debate: does it have to be fully autonomous to be a true agent? I’m more pragmatic than that. I don’t think that’s a requirement.
Here’s the important part: the way the world looks soon is that we will have hybrid teams consisting of both humans and AI agents. The easiest way to describe it is to think of the digital agent as a digital team member. We’ve had hybrid teams before — we hired freelancers for three months, we had geographic hybrid teams. Now we’re going to have hybrid from a carbon-based life form versus non-carbon-based life form perspective.
My wife comes over and says we have guests coming. She decomposes it: clean the kitchen, clean this area, get food. She tells me to load the dishwasher. I load the dishwasher. Then I hand it over to my dishwasher agent, who washes those dishes. I trust that agent to do it well every single time. That’s what the inside of companies is going to look like.
And I think we will have humans doing the review and approval for the vast majority of tasks. Digital team members are going to be doing lower-level tasks first, because those are the ones we can trust them with. The stakes are lower. Produce ten versions of a blog article — I’ll pick the one that goes out. Once I pick it, the post-production can be all-digital.
Shaan: Give us an example of an agent you actually use on a day-to-day basis — live, it’s working.
Dharmesh: Good simple example. I really like an episode of, say, My First Million — something stuck in my head and I want to write a LinkedIn post about it.
Here’s what I want to happen: all I’m going to give it is a YouTube video. What I want as output is a LinkedIn post that will do well by some definition of “do well” — we have training data on that. So: step one, pull the transcript from the YouTube video. Step two, figure out who the players are. Step three, highlight the things that are quote-worthy. Step four, figure out what Dharmesh was actually asking about — there was a snippet I wanted to call out. Put all of that in. Then: figure out what style works well on LinkedIn specifically — bullet points or not, emojis or not, what’s working. Write me a 200-word prompt describing the language and style. Take all of that, and produce a LinkedIn post.
I have a thing that does that. And the fun part: it’s Lego bricks. I have an agent that does YouTube transcripts. But it’s a better transcript than what you’d get copying from YouTube — because there’s an LLM involved. It creates chapter headings, highlights actual quotes, formats it in a human-consumable way. Way better than the regular YouTube transcript.
Shaan: Check this out, Dharmesh — have you seen this? Go to MFMvault.com. My friend Greg built this. It’s basically a site that does what you just described, specifically for My First Million. He was like: I don’t just want summaries. I like MFM because it’s got ideas, frameworks, and stories. So he’s got AI extracting those from every episode — searching for frameworks, searching for stories. And if you go to Episodes and type “Dharmesh,” here’s our last episode in 2023 — the overall summary with chapter titles. And if you go to “Stories,” you get: HubSpot’s $1 sale, Pandora’s deal with musicians, the 17-year evolution of chat… and if you click one, it takes you to that moment and summarizes it. There are also Frameworks — like an “AI Immersion Week” format.
Dharmesh: Yes! That’s it. And to do it, Greg strung together five agents: one that listens for when there’s a new episode, passes it to the transcriber, the transcriber passes it to the summarizer, the summarizer to the extractor, and the extractor to the clip creator. Each is its own agent. And as you said, it’s composable — if I want to use the summarizer for something else, that agent exists and can summarize anything. Not just MFM episodes.
Dharmesh: That’s the future, Shaan. So, just a quick one-sentence description: agent.ai is the number one professional network for AI agents. It is also the only professional network for AI agents. LinkedIn for agents.
Shaan: Yes. Why do agents need their own professional network?
Dharmesh: If we imagine we’re going to have this hybrid world with both AI teammates and carbon-based life form teammates — how are we going to find those people? We have Fiverr, we have Upwork for humans. There needs to be an equivalent for AI agents. And here’s where it gets super cool: imagine a professional network with ratings and reviews, experience listings — “I’m an agent that does this, I’ve been used 40,000 times, here are the people that like me.” Dharmesh follows me on agent.ai.
What’s going to happen over time is agents will be able to hire other agents on the network. Agents can go test other agents: “I’ve been given a budget of $100 to try five different agents to find the best one for my use case” — and do it without human intervention.
Over 30 years I’ve built a lot of what I call solo software — just for me. LPMs — laughs per minute by the way, I did not invent that term, standup comedians use that, been around forever.
Sam: Can I ask you a rude question and then a thoughtful question?
Dharmesh: Sure.
Why Not Do the “Elon Thing”? [01:12:00]
Sam: Rude question: dude, you’re a billionaire, you’re super smart, you love all this new tech — why aren’t you doing the Elon thing? Elon’s like, okay, I did Zip2, then PayPal, then Tesla, SpaceX, I’m going to keep building new companies, bigger and bolder bets. You’ve been at HubSpot for so long. You’ve been doing CRM for like 30 years. Don’t you want to spread those wings and just go build your rocket?
Dharmesh: That’s not rude at all. The answer is actually quite simple, and has the added value of being true: this sort of is my next big act.
Here’s the thing. Let’s say you’re playing a video game — pick your favorite. You’ve grinded out, you’ve built the things, you’ve got the weapons, you’re in power mode. Now you can go exploring. But if you’re going to play a new game, there’s a bunch of stuff you still have to get right: you have to build a team, find a co-founder — and by the way, I’ve won the co-founder lottery. Brian Halligan — no one knew it at the time, but that has a massive impact on your outcome.
I don’t feel like I’ve ever been kept from doing the thing I would have been doing if I were out on my own. If that were the case, I would leave. But I don’t.
Sam: I don’t accept your answer. Let’s break it down. You have things you’re excited about — new technologies, potential to create cool stuff in the world. There are two ways to look at it: either HubSpot is actually the best vehicle for that, or it isn’t — and some ideas you might get excited about are just unlikely to also be on HubSpot’s roadmap. Also: the burn-the-boats thing. When people quit their job to do a startup, something happens. They’ve told the world, told themselves: I’m going to make a thing. I gave up something, so I have to build something new. That’s one of the only assets a startup has — they’re all in. I think the highest-potential version of you is more likely outside HubSpot than inside it, doing side projects between meetings.
Dharmesh: I think your point is well-taken. And most of the time, we have externally imposed rules. For instance, if you’re an engineer: “Well, you’re CTO, you shouldn’t be writing code anymore.” I do not accept that thesis.
Software is a creative discipline. It’s a hits-driven business — you have to be right a very few number of times. We don’t tell musicians who are really good, “Stop writing music and go manage a team of writers.” We don’t tell exceptional writers to stop writing. Let the others learn from you, but you keep doing the creative work.
That’s what I feel like I’m doing — maximizing my build time. My personal mission statement is to help millions grow better. I want to look back and say I had a positive impact on the most number of people. Impact times number of people — that’s my truth function. This is why I don’t do one-on-one meetings because they don’t scale. This is one of the reasons I’m here and not elsewhere.
Sam: You do this a lot — you’re telling a story and you drop these little wildly fascinating bits. One of them being that you have a personal mission statement. Do you have personal values like a company does?
Personal Values and the Humility Debate [01:20:00]
Dharmesh: The HubSpot values are effectively the founders’ values — that’s where they originally came from. Most companies are actually a reflection of their founders.
What’s the most controversial value — the one that has the biggest tradeoff?
Sam: What’s one that not everybody would choose?
Dharmesh: One that almost didn’t make the cut — not because we don’t agree with it, but there were strong arguments against it — is humility. The common argument against it: we’re supposed to be a winning team, we’re aggressive. People often confuse humility with lack of confidence or assertiveness. But that’s not it at all. Humility is being self-aware, being able to recognize you don’t know everything, being here to learn. I fought for that value for a long time.
The second most controversial is empathy — our second value. It didn’t always used to be empathy, we changed it along the way.
Sam: This is a good question because I’ve been thinking: do I have values? Like, what are my actual values? It’s kind of cool to codify it. I mean, we kind of have values at MFM — we’ve had people on the pod who have done bad stuff, and our default is to build people up rather than criticize. But we haven’t explicitly said: these are the values.
Mark Zuckerberg has this thing — maybe he said it on the pod, Sean — that a value means you have to sacrifice something. “Move fast and break things” means we are going to move fast, and because of that I am okay with breaking stuff. Too many people’s values are just platitudes — things that are obvious and don’t involve any sacrifice.
Dharmesh: That’s right. “Move fast and break things” is in the pantheon, the Hall of Fame of company values. Another one: Ray Dalio’s Bridgewater thing — radical honesty. We are going to be uncomfortably honest with each other, and we think that maximizes in the long run. In the short run, it’s not going to be like anywhere else you’ve worked. The tradeoff is real: discomfort for the upside of being radically honest.
Local Maximum vs. Global Maximum [01:26:00]
Dharmesh: So I want to end on something I think is a useful framing. We talk about maximizing — I love that word. You’ve likely heard of finding the maximum. And the trap people fall into is finding the local maximum versus the global maximum.
From a lay perspective: imagine a graph with a bell curve. There’s a point where the value is the highest — the top of the hill, mathematically speaking. A local maximum is: you’re solving based on the chart in front of you. You found the highest point on that chart. But if you zoom out, there’s another bell curve right next to it — an even higher maximum. If you were zoomed in, you didn’t find the maximum, you found a local maximum.
Here’s the non-obvious part: often, in order to even see the global maximum, you actually have to climb the local maximum hill. You can’t see it abstractly from the sidelines. You have to make the effort to climb the smaller hill. You get to the top, and suddenly you see the landscape — and there’s that massive mountain over there, which is the thing you’re actually meant to do.
Sometimes you have to do the thing that doesn’t seem like the big bet in order to gain the perspective to see what the big bet is going to be. This goes back to the connecting-dots thesis too.
Results as a Service (RaaS) [01:30:30]
Sam: Have you heard the concept of RaaS? R-A-A-S. Instead of SaaS companies — Results as a Service companies?
Dharmesh: Right, so we all know SaaS — software as a service. We contrast that to software as it used to exist: you bought the box, the CD, loaded it on your computer. Software as a service said: you don’t buy the box, we just provide the value over the internet.
Results as a service says: you don’t even access the software. You tell us what outcome you want, and we sell you that. You want the LinkedIn blog post at the end? We’ll just produce it.
A good example would be legal software that analyzes a contract in real estate and gives you commentary on whether it matches benchmarks. Instead of: “Here’s the software to help you review a contract,” it’s: “What’s the outcome you want? You want to know if this contract is fair? Great, we’ll just tell you.”
And this idea is almost as old as time — we’ve had lead-gen companies forever, which are effectively results as a service. Don’t worry about the phone numbers, don’t worry about the internet — pay us X amount of dollars every time your phone rings.
Sam: It’s also a sales technique. “Do you want guests to feel great at your home?” — okay, buy this vacuum to keep your carpet clean. People don’t want drills, they want holes.
Dharmesh: Exactly. RaaS is the next wave of software. The software is still being used by whoever is providing the results — that’s how it gets accomplished. But the way it gets packaged and sold might end up being results-based.
And closing out on the agent front — what Greg did building MFM Vault, stringing those agents together, is a simple version of this. He effectively built an agent with multiple steps, each step calling other AI tools. agent.ai not only is a professional network for distributing, finding, discovering, reviewing, and rating agents — it also has an agent builder: a platform for building agents without writing code.
Closing: What Motivates a Billionaire? [01:37:00]
Sam: You’ve had a good quarter. HubSpot stock hit an all-time high. You sold a domain to OpenAI. You came on the podcast. I don’t care what they say about you, Dharmesh — you’re doing okay.
Dharmesh: I’m doing okay. I’m doing okay.
Sam: What I love about you is you have a bunch of things that sound like paradoxes. You’re extremely gentle and kind. You’re also competitive and aggressive in other ways. You’re humble, but then you’re also a marketer — you see the marketing value of things. You’re like, I’d rather be quiet and low-key, I don’t need to brag — but then as soon as you got on a leaderboard you wanted to be number one.
Is money a motivator? Like — I think you might try to be cool and say, “No, I’m happy.” But I think at every level you want to get to the next one. Like, whether you had $1 billion you’d want $10 billion, whether you have $10 billion you want $50 billion. Same way no matter how ripped you get, you want to get a little more ripped. How much of a motivator is — it says one number on Forbes, how do I make it say 10?
Dharmesh: It’s a really good question. I tend not to have explicit goals. Goal-setting is not something I do now, and I never really did it. It’s like: I want to get better, I want to have freedom, I want to make choices. I want to configure the universe to my liking, to the best of my ability, with whatever resources I have. I’ll go find a way to make that possible.
Deep down, my soul is a math and physics guy who’s not smart enough to be a math and physics guy. My truth function is: I’m trying to create positive impact for the most number of people — multiply X impact by Y number of people. The marketing is part of that. I’m not going to be able to impact people if I’m not doing some marketing. I can say the most brilliant things in the world, but if only 17 people see them, the value of that unit of work is 17. That’s just not how I operate. I will make sacrifices in other areas to solve for the mission.
Sam: We talked about a bunch of stuff. You had opening remarks. Do you have closing remarks? Did you achieve your mission — did you say what you thought would help somebody increase their odds?
Dharmesh: My ask — my favor — is to leave a comment and let me know what you thought. That’s the leaderboard I’m after right now.
Shaan: He’s saying: I’m number five on the YouTube list. And I know comments and engagement actually help the algorithm more than likes do.
Sam: Dude, you’re the man. I have a notebook here where I take notes — I filled three pages. You’re the man.
Dharmesh: It’s all good. Pleasure beyond. Thanks for having me. Thanks for indulging my quirks.
Sam: Dude, you’re great. You’re fun to hang out with. You’re fun to talk to. I feel energized when I talk to you.
Dharmesh: Good to see you guys.
Sam: We appreciate you. That’s a pod for you.