Agents, Metaprompting and Tech-powered Parenting with HubSpot CTO Dharmesh Shah

AI still cannot write even a decent original dad joke despite my best prompt engineering attempts. You know, I love AI as much as anyone. I have a huge amount of respect for the technology and what it's going to be capable of. My money is still on the humans. I'm sorry.

Welcome to Prompted, a podcast about AI, people, and the creative spark. I'm Cameron Adams, and I am fascinated with leaders who are actually using AI to be more creative. whether that's building better products, coaching their teams, or harnessing AI for their personal creativity. In this show, we dig into the details of how they use AI that we all want to know. Today we are joined by the amazing Dharmesh Shah, co-founder and CTO of HubSpot. If you followed his career, you know that Dharmesh has always taken the unexpected path, from starting his first company with $10,000 on credit cards to writing a regularly updated culture code for HubSpot and publishing it for the whole world to see. He's got a refreshingly simple approach to leadership. He runs a massive company without a single direct report. And in true entrepreneurial spirit, he has recently started some exciting new projects focused on AI agents, which we'll get into on this episode. Dharmesh, welcome to the show.

Thanks for having me, Kevin.

It is so good to have you here, and I am keen to dive into AI agents and tech-powered parenting with you, but To help us get to know you a little better, let's kick off with the prompt. This is the part of the show where we get an idea out of your head and onto the page, and then we get one of our designers here at Canva to take it even further with AI. So to start, we would love you to draw something from your childhood that inspired you in your adulthood. While you draw that, you've got to tell us the story about it as well. You get 90 seconds to talk and draw. Do you think you can multitask?

Wow, okay. The multitasking is going to be hard, but I'll try it. So I'll tell you the story. When I was in my teens, my parents visited Singapore for the first time. I didn't go. And they brought back this electronic device, which ended up changing my life. Funny thing is, when I tell people this story, which I haven't told that many times, It's like they automatically assumed it was a computer, and it was not a computer because I didn't get a computer until I was in my 20s, much later. The device I did get though that they brought back, which I'll try to sketch out here.

I'm excited to see what it is.

Looks like this. So it's a tiny little Casio keyboard. This is probably like almost drawn to size. I had not shown any inclination for music. I'm not exactly sure why they got that for me as a gift, but I sort of became obsessed with it even though I didn't know how to play it at all, and I taught myself through brute force, which is randomly pressing keys and seeing if the notes sounded right or not, and then pressing one of the other keys until something sounded right. It would take me like weeks to learn a single song. And I've been doing that for about 30 years now, you know, ever since.

Amazing. I've never heard anyone apply brute force to a piano before, but it sounds like it worked for you.

It worked.

But if you want to see that image and you're listening to this on the audio, head to the link in the show notes. To see Dharmesh's picture. We're actually going to send this picture to Paolo, our designer in residence, who's going to team up with AI to whip it into something new. We'll reveal Dharmesh, Paolo, and AI's creation at the end of the episode, so stay tuned for that. But while Paolo works, let's go deep on AI with you, Dharmesh. Now I want to get into HubSpot a little bit later, but I first wanted to touch on agents. You've recently launched Agent.ai, a professional network for agents. What was the spark that led you to build this platform?

I've had this thesis that in the future, we're going to have agents and humans working together in what we would think of as modern hybrid teams. I'm like, OK, well, if that's true, if we're going to be putting AI agents on teams along with humans, then we're going to need all the things that we have in the analog human world. It's like, how will people find these agents? What would their resumes look like? How do we know if they've had experience and all these things? And so the original spark was there should be a professional network for agents because that's going to need to exist someday. So that was the original kind of genesis of the Agent.ai idea. We launched it in beta about a year ago. We had about 50,000 users back then. It's now up to 2 million. So there's been some resonance with the idea.

Wow.

What's happened now is people are sort of building their own agents and kind of creating them on the platform. And we have We have about 2,000 kind of publicly shared agents now on the platform. So it's been a lot of fun.

Many of the agents on agent.ai are actually ones you've built yourself. I figure that's a cold start problem. You needed to get some agents on there to start. But can you walk us through one of your favorite agents and what it does?

One of my favorite agents, one of the most widely used ones, is a company research agent where you can say, type in the company name or the website domain, And then we'll do this kind of deep research from all across the web, hitting a bunch of proprietary APIs. But what's cool about it is just like if you had a researcher on your behalf, you would say, OK, well, here's the parts that I'm actually interested in. Here's the business that I'm in. Here are the questions that I have. So I want you to answer these 5 questions, focus on these particular areas, and give me a customized report every time I ask you for a research report on a company. That's one of them. But I've done— Probably built a dozen or so agents, from relatively sophisticated ones to really simple ones. And I'll tell you the one that I've been trying to build, which is like a dad joke agent to be able to generate a dad joke, which one needs. Well, I guess one of the last remaining frontiers, despite all the reasoning models and all the advancements— and I played with all of them— AI still cannot write even a decent original dad joke. That's what I've discovered, despite my best Prompt engineering attempts.

That's gonna be the new Turing test. Can you pick whether this is a dad or an AI dad?

Yeah, there you go. Yeah.

What's the actual process for building an agent that you are using at the moment? Is it mostly just wiring up APIs or what does it look like when you're building from scratch?

There's a spectrum of them. When we started the platform, it was a, we had a low-code agent building platform, which primarily deterministic flows that says, here are the steps I wanna take in a multi-step flow. Here are the inputs. Here's a minimalist UI. And then it would call the LLM or call APIs. And we have a bunch of data connectors to get social media data and pull all that in. And now we also have the ability to have agents that are built off-platform. Pick your agentic or vibe coding tool of choice and build an agent that way and just wire it into the platform, which I've been spending a lot of time on, building these much more sophisticated UI agents that still plug into agent.ai.

We see a lot of data mining and research and going out to the web and fetching stuff. But do you think AIs can possibly be creative? Could you have a truly creative agent?

Yeah, for some definition of creativity, right? I use AI for what I think would be kind of classified as relatively creative use cases. They're not just data research. It's not just kind of writing or wordsmithing something. One of these tricks I sort of learned from my— well, not sort of. I learned from my son, who's 14, is around kind of creating what I'll call a simulation, for lack of a better term. So what he does is he's an aspiring fantasy author. So he'll do world building inside of ChatGPT. So he's got like a 2,000-word prompt to say, oh, here's the world. Here are the characters. Here's the power structure. Here are the things that people can do. And then he'll pressure test the ideas, like, what happens if I do this? And it'll come back. And it's this very iterative process. I do something similar with kind of business ideas. Like, oh, here's what I'm going to tell you. And I've got this kind of very long kind of knowledge base that I can pull information in from. and say, what would happen if HubSpot launched a product X, or we did this change to it, or whatever? Like, give me, uh, or if I were to, uh, ask 5 people across different disciplines to give me feedback on this thing, uh, from different, uh, regard, like, what would that be? What would that look like? In a way, creativity is sort of kind of combining primitives that already exist in novel and unique ways. It's not like we're kind of building things always from whole cloth. I don't use it for writing full articles or, you know, help with talks and things like that. But it's really good at connecting dots that might be hard to connect and doing, like, transitions or, like, oh, I'm trying to wire this up, but I can't quite find a way to get to this thing that I wanna get to from where I am right now. Can you give me some options? It's a great critic. It's a great brainstorming tool. It's a great way to kind of just spark some ideas and add some dots to this kind of mix of dots that you're trying to wire up.

Yeah, I love that possible concept of what creativity actually means in terms of connecting dots and bringing ideas that maybe you didn't think should exist together into proximity to one another and seeing what comes out of it. In my life, a lot of creativity has come from that. And I can truly see the value of doing that through an AI model that has all these connections in it that you mightn't possibly actually think of.

One of the neat things you can do, it does it really well, is you can— so imagine you had a spreadsheet or just a 2-axis thing, whatever. And you have your kind of products or product features. down one side and the same product features along the top, right? So it's the exact same thing along x and y dimension. And then what you do is you run the simulations like, I want you to combine this product with this product or this feature with this product in novel ways and come up with a description of like, what would that look like? What might be like, oh, we happen to have this and we also have that. Is there some unique blend of things? And some of them or a lot of them are not going to make any sense. But sometimes it comes up with things that are like actually useful and novel. It's like, oh yeah, I hadn't thought about how we could take— This feature that's in our kind of sales product and connect it with this customer service product. And that would be kind of cool, actually.

And that's kind of an emerging skill, isn't it, in the age of AI? It's like picking out the gold, you being that editor that is taking in everything that the AI is spitting at you, which might be terrible, might be brilliant, but you need to figure out which one's which.

Yes. Yeah. It's weird, back on the dad joke thing, not only is it not able to write dad jokes well, It actually struggles with even when you tell— so I have a database, which is what I call this, a database of curated jokes that I've collected over the years. And so it can actually explain to you what makes a joke a joke and why it's funny. So it actually does a very good job at that. But then if you have it rate the jokes on a scale of 0 to 10, that says, I'm going to give you 100 jokes. You tell me which ones and give me a quantitative score based on the joke. It does actually not that good at that. And I'm not exactly sure why, even though it can tell you what makes a joke funny. funny, like the underlying mechanics of it, it's not good at actually quantifiably measuring the quality of something, which is— I don't know if that's just a function of the way they work or something like that. Anyway, I haven't been able to figure that out yet.

Do you think there actually is an objective measurement of humor, though?

That's a good question. Maybe I'm asking the wrong question. Maybe the question I should ask is, I want you to simulate an audience of 1,000 people and estimate how many people would laugh at this joke.

Well, you've gone deep on this before, haven't you? Because you have optimized some of your talks for laughs per minute.

I have, but I actually use real-life humans and human audiences to kind of test that prior to actually delivering it on stage. But that's some interesting ideas, like maybe just run a mock simulation of an audience with a certain makeup and say, okay, well, I'm gonna test this joke or this humor and see if it gets a reaction or not.

So you've talked about you know, AI agents possibly being teammates, and you're actually all in on hybrid human-AI teams at HubSpot. Can you actually share a real moment at HubSpot or one of the companies you advise where that human-AI collaboration has created something unexpected, something that neither humans nor AI could have pulled off alone?

I'll be honest, the answer is not yet. I think we're very early in this notion of agents being more than just tools. So there sophisticated tools now as AI is, right? Because they can actually carry on conversations. We have a customer support agent. We have a prospecting agent. The thing we haven't quite figured out yet, and I think we're right on the brink, is I think the really interesting things start to happen not when you have an agent that's purpose-built for a specific thing, which is what our agents currently do, is when agents have knowledge of each other and can actually bring in— it's like, OK, I'm trying to resolve this customer support issue. I'm going to bring in the prospecting agent to try and resolve this issue for whatever reason, like bring in some other agent that's disconnected from the core mission of the agent that's currently running. And I think part of this is going to get resolved. We have things like the agent-to-agent protocol from Google and MCP where the large language models can be exposed to tools and other agents. And we have some sort of discoverability. But that dynamic of unexpected things— now, it will do unexpected things and be able to answer questions we didn't think it would be able to answer. But it doesn't really draw outside the lines, right? It's like, OK, well, sort of within the system instructions and the RAG-based knowledge that we provided, it will do this, mostly stay within the lines. And we see this with humans as well. It's like very, very interesting things happen when the number of humans exceeds 1, right? It's like, OK, there's a certain number of creativity I can draw in a brainstorming session with myself. But then when I put other people in there is when the truly interesting things start to happen. And the other thing I think This is more on the technology side. And AI is getting better and better. And that's great. The protocols will get better. But I think we need a better way to weave agents into all the same tools that we use, be it email or Slack or whatever it is, right? They need to be sort of part of the same systems that we use for doing the regular work that we do. And I think that'll probably happen probably next year sometime.

Yeah, I think sitting around waiting for a prompt to come in is probably the least interesting thing you can do. Getting a WhatsApp notification or a Slack notification is definitely going to be the future.

Yeah.

Do you see agents really being part of a team? You kind of mentioned they're at the tool stage now, or do you think they'll always be tools? Will they bridge that gap to truly being a teammate that you trust, that you converse with, that you're happy to delegate to?

Yeah, I think so. And it's not just the optimist in me. It's, you know, from a just a raw IQ perspective, you know, models already have the intelligence to do a lot of the things that we do that we would ascribe and say those are human-oriented tasks. There's a couple of things where we can continue to do work. One is around EQ, around understanding, despite its lack of human experiences, can it actually develop true— at least a proxy for empathy? And it's starting to get better at that because it sort of fakes it pretty well. There have been some studies that are done. But the thing I think will sort of really unlock it is not the technology's problem. It's our problem as organizations. The analogy I use for agents is like, so let's say you were hiring what I think of as a very adept intern. That's the way to think about an AI agent. And let's say this adept intern was really adept, had a PhD in everything, was very, very knowledgeable, super smart. But when they show up for their first day of work, they don't know about your business. They don't know about your policies. They don't know about their teammates. They don't know anything other than— Mm-hmm. All the external training that they've had. We're like, oh, we hired this really smart agent to do x. And we sort of unleash it on the world, unleash it within our organization, and expect it to really perform. And I think that's an unreasonable expectation. So once we start truly treating them as potential teammates, well, they're going to need onboarding. They're going to need some training. They're going to need some feedback. They're going to need some objective, measurable way to know whether they're making progress. They're going to get some things right. They're going to get some things wrong. And then over time, I think we'll have much more feels like a teammate kind of person that just so happens that they're digital, but they're sitting in my Slack, they're responding to emails, they're kind of watching what's going on, and they don't have to always be prompted to do something, just like— and the distance between how much you have to sort of look over their shoulder, it sort of grows over time, just like we have with a regular person. The trust builds. They get more awareness and more context about the organization. I'm hopeful we'll get there.

I wonder what those relationships are going to be like, because you've talked about onboarding, you talked about feedback, you know, possible coaching of AI agents, which are all very human things. You have conversations about performance, you identify growth areas, you have growth plans that kind of roll out over 6 months, a year, all very human things. And are we going to treat AI agents in the same way, or is it going to almost be a servant relationship?

I still think of it as a tool. So I don't like personifying tools. It's like, I love AI as much as anyone. I have a huge amount of respect for the technology and what it's going to be capable of. I don't think it actually replaces a human in the literal sense. I think it replaces individual tasks that humans may or may not— should have been investing their calories in because they're capable of so much more. And the reason we did those things. And we have the same thing with the industrial age. It's like we sort of, as we invented machines and sort of automated things, like, OK, now we can sort of work at a higher level as humans, work at a higher level of abstraction. I think these will still continue to be tools for a while. But that doesn't mean that they can't be accountable and responsible and overtake things with a high degree of trust. But I think ultimately, like, the tastemaker, the thing that's going to hold it all together, the glue, is going to still be carbon-based life forms. I think there's still something— particularly special about humans. I may be biased being a human, but—

I've seen you described as a creative CTO. To what extent does technology unlock your creativity, and how do you think this might be changing in the AI era?

I understand the arguments that argue that AI kind of squashes or suppresses creativity because it allows, you know, it sort of replaces the need for that kind of creativity. And I think the exact Opposite. Creativity is like, OK, you have some concept, some idea in your brain. And the challenge that many of us have— and I'll talk like my son has. So he's this aspiring fantasy writer someday. But when he was 9 and 10, and even like now, his skill set in terms of being able to put words together into sentences and actually write and do, it was lacking, right? So it's going to take him years to kind of develop that to even a rudimentary level to where he can kind of—

Yeah.

craft a story, uh, you know, for his, uh, target audience. But now with AI, it's like, okay, well, he still has the ideas. He's always had the ideas in his head, right? It's like, and I'll, I'll, you know, from a design perspective, since, uh, we are here on the Canva podcast, it's like, okay, well, Canva did this for me, um, you know, back when you folks first came out, and now AI does this for me. It's like, I've never thought my, of myself because I'm not like a designer, but it's not because I don't have visual ideas, right? It's like I have things that are in my head that I want to accompany a blog post or something like that. And now with AI, it sort of reduces, kind of lowers the bar in terms of the kind of entry level for people that do a particular thing. And so I think it amplifies creativity because now it takes the ideas that you have in your head, the creative ideas, and helps you manifest them, creative, uh, express them, be it music, be it visual design, be it writing, whatever it happens to be, as your kind of skills catch up. On the whole, I think people are going to be more creative. We're gonna see more output, more creative things, uh, from people because the bar is just lower. More people can do it.

Yeah, I love getting that insight into how the next generation's using AI. You mentioned your son is building worlds to express his fantasy writing, and he's putting in 2,000-word prompts, which is like amazing. I think you're pretty experienced with tools like ChatGPT. Is there a way that you use it that would surprise people?

Yeah, a couple of things. I just launched a very simple kind of agent, you know, so we've heard of like prompt engineering, which is the craft of writing a prompt in order for you to get good results from AI back. But there's this kind of step up from that. And it's called meta-prompting. And the idea behind meta-prompting is using AI to actually improve your prompt. And the agent that I have is metaprompt.com. It's completely free. And so what you do is you type in the prompt that you have been using on a recurring basis. And what it will do is say, OK, here's what you're trying to do. It'll come back and ask you questions. It's like, oh, you're doing this, but you haven't told me this. You haven't told me this. You haven't told me this. Answer these questions. And when you do, it comes back with a well-structured prompt that's actually written to— and it's like a one-time investment to take your prompt that you've been using all the time into something that will demonstrably prove to give you better results, right? And I use that all the time. And this is one of the nice things about AI now and agentic or vibe coding is that these are the kinds of things I would build for myself back in the day. And it wasn't the calories necessary to be able to kind of launch something and put it out there for the world to use. I could just never make that kind of trade-off. But now it's like, okay, not that hard to build, not that hard to share. And it's, uh, I can just put it out there.

Metaprompt.com, you heard it here. It's actually one of my favorite pieces of advice. It's turned into my let me Google that for you, just telling people to go Metaprompt. That is like my automatic response when they're asking for how to deal with AI.

Totally.

So you actually write simple.ai, which is a newsletter with more than a million subscribers now. I'm gonna guess that you actually use AI to help get this out the door. So how do you personally use AI in your writing workflow?

I tend not to use it for writing per se. What I use it for is editing. I use it for visuals. I use it for what I think of as less about writing and more about copywriting. So I'll use it for headline ideas or subject line ideas, and I'll use it for coming up with like punchy ways to kind of capture something. It's like, okay, I've got this thing. I'm like right around here on this particular concept. I need a pithy way, uh, that maybe uses alliteration or a figure of speech that makes it more memorable, more remarkable, and sort of improve and take it up a notch. And I've made multiple attempts at this, and I think, I think it's pretty good, but it still can't capture my voice. And it's not good at that yet. Um, and I say yet, maybe it will be someday.

But I'm sure you are very data forward and you've got an amazing amount of content that you can feed it. Surely it could find your voice somewhere in there.

It can replicate my voice. What it is likely unable to do— and maybe this will change— the thing that it doesn't have, though, is it doesn't actually have human experience, right? So it's not had the day that I just had. Kind of going back to comedy, which I'm not a comedian either, but I've heard comedians say it's like their best material just comes from just observation of like, I had this day, and it was nothing particularly remarkable, but that was kind of quirky. I had this or whatever. And then they cut. pull on that thread, and that turns into, you know, a skit or a joke or something like that. And I think it's similar for writing. It's like, okay, I was solving this particular problem, or I had this thing where I had this debate on a call, and it's like, oh, then I thought about it. It's like, ah, there's something broken or something kind of noteworthy about that. I just put it out there. So, yeah.

So what I'm hearing is that the creative spark always needs to come from your brain, from a person.

Yeah, for the kinds of things that I'm doing, right? It's like if I'm doing a LinkedIn post, the kind of purpose of it is to kind of put my brain on speakerphone in a way, right? That says, OK, well, that's, I think, what people are there following me for. It's like, ah, what's going on in Dharmesh's head? And if you're not curious about what's going on in my head, then you're likely not following me. That's because I have a little grab bag of things that I post out there. And it's like, OK, well, that's sort of why the audience is here. So like, and my kind of purpose is to kind of serve, you know, serve that audience. And so, yeah.

All right. Well, AI agents might be teammates, they might not be, they could be tools, but you also famously took on the culture project at HubSpot despite feeling that you were absolutely the worst person to do it. So I'm interested in a truly hybrid team where maybe we've got AI agents sitting next to people. How do you instill culture in the AI half?

I'll summarize, uh, my dozen-plus years in kind of working in culture, uh, involuntarily. I'll just distill it down. Like the, the big lesson, big analog I've come— I've only said like maybe 3 smart things in my life. I think this is one of them, which is, uh, culture is a product. And what I mean by that is, as, as, as companies, we build a product for our customers. Everybody knows that whatever your offering happens to be, culture is the product you build for your team. I'm just talking about human teams right now. Well, you would never build a product without asking your customers how the product was doing. And so we get product feedback. So we do an NPS for our employees in the same way. So we sort of carry as much as we can from our kind of product sensibilities over into building this kind of second product, which is what culture is. And the other kind of big lesson on culture was, in the same way that a product is never really done, culture is never really done. I think the mistake a lot of founders and entrepreneurs make is like, oh, my job is to preserve our startup culture. It's like what we had when we were like 5 people, 10 people, 50 people. As long as I can preserve and maintain that, then, like, we're good. And it's like, no, that's not the answer because the product that the company needs and the people need is going to be different at 50 people, 500 people, 5,000 people.

Yeah, totally.

And so now I'm gonna continue to pull on that thread. And maybe there's a point at which the analogy breaks. But it's like, OK, well, if we continue to think about culture as a product, what's changed now is who the customers are. So now we have a combination of humans and AI agents. And if the goal for the product is to deliver for those customers, now we have 2 constituents of customers, just like you might have for your regular product. And so then how would we get feedback from the other constituent of customers? Like, is the culture serving AI agents and making the most? Is it helping them live their best life, so to speak, right? Not that they're conscious. But that's the way I've been thinking about it is like, how do we replicate the things that have worked for us in thinking about culture into this hybrid agentic world? Yeah.

How far do you take it though? Does the AI get a vote on what the values of the team are?

It's a good question. I don't, I don't think so. And the reason is, is I'm trying to think off the top of my head right now, just like we would have, so we may have multiple product constituents for our regular product, but there's sort of like the constituents that we're focused on. It's like, oh yeah, this is like, for instance, HubSpot serves SMBs. That's our kind of target. That's what we're passionate about. We've been passionate about that for 19 years. But we have segments of enterprise. We have other kind of adjacent customers. But the ones we solve for are those kind of key, the SMB constituent. Real-time evolving thesis here is that humans will continue to be the kind of primary customer. But we have this other constituent that we need to be mindful of, but that's not who we're building the culture for. And so they may get— they may have a voice in terms of we want to know what they're sort of thinking and how they're responding.

Yeah.

But they may not get a vote in the way— in terms of shaping what the product looks like. We're not solving for them in that way.

Do you think leadership, maybe management, is going to radically shift in a world where AI can take on so much of that role and some of those management tasks and what a leader normally does?

Yeah, I think so. I think the managers and leaders of the future are going to be the ones that embrace this idea of hybrid teams because I think that's going to be inevitable, whether Wherever on the spectrum you place the agents in terms of their capability, whether it's a tool or, you know, to what degree do they become like a, uh, a full teammate, but they're in there, right? Just like managers sort of have to understand technology today, right? So you have to sort of understand that we have HR systems and CRMs and things like that in order to like be a good manager. I think what agents will start to kind of show up and manifest is like great managers will know how to manage a diverse team, which includes both carbon-based life forms and agents, and they'll just know That's like, here's what agents, how to get the most out of them, and here's what people expect and how to get the most out of them. It's gonna be a combination. I think there's going to be a whole new generation of managers and leaders that are, that have that skillset.

Yeah, it's gonna be an interesting dual skillset of, of managing 2 different types of organisms, really. Yeah.

Yeah.

I'm gonna take us down a slightly different path. Your LinkedIn lists one of your favorite topics as tech-powered parenting. Does this mean you just put your kids in front of Canva for 12 hours a day?

I don't. And even if I wanted to, my wife would never let me. But I don't think that's a good idea. So I've exposed my son to technology very, very early, including— so he had access to GPT-2 before ChatGPT was even launched because I got an early access beta.

Early adopter.

Yeah, very, very early adopter. What I want him to do is I want him to sort of understand kind of the technology That's available and use it as a tool because it's going to be in there. It's going to be part of his life. So I don't want to keep him from that. But I want him to, over the fullness of time, understand what responsible use looks like. And the schools will help out here as well because he's getting this message from all over. But I will sit down and it's like, oh, here's how I use AI. And here's why I use it for this. I don't use it for that. And What I love about collaborating with him on that front is one thing kids are really good at is coming in with a blank slate. They assume the technology will actually work. That's their default assumption. He will type things into ChatGPT that you would never— it's like, why would you even think it could do that? And he's like, why wouldn't it be able to do that? It's supposed to be able to do everything. They don't know of a pre-internet world. It's like, OK, well, yep, there are dark uses and ways you don't want to use the internet. I think we're gonna have the same thing with AI. We just have to use it responsibly. But I don't think it's a responsible thing to kind of keep our kids from it because they're going to be out in a world where it's going to be there. That's just the kind of reality of life, whether we like it or not. So yeah.

Giving our kids access to technology is one part of it. But with tech-powered parenting, is there something that you particularly do to help them come to grips with it or ways that you use the tools to Give your kid a better life.

I think what AI is really good at is taking in context. And you only have the one child. This is our first time, first-time parents, which lots of people, obviously billions of people, have gone through that exercise. But it's actually really good at providing advice. And we will collaboratively sometimes like, OK, well, here's the situation. How do we think through this? And sometimes it's just me and my wife. Sometimes it's just me. But in terms of, kind of helping me navigate or helping us as a family navigate what are common issues, but they're uncommon for us because we haven't encountered them yet. And obviously, every child is different, but there are definitely patterns. And it turns out AI is actually very good at that. It's like having kind of read all that there is to read out there on the internet is actually quite helpful. Even before the reasoning models came out, they were actually really good. But now the reasoning models that they can do, kind of deep research, can do inference time compute. The quality of the responses you get on very, very deep, sometimes philosophical questions. And this may be typical— atypical because my son is wired a certain way, but he sort of likes the AI as arbiter kind of objective third party. Also because it's like, you know, kids sometimes don't trust their parents. Like, ah, you're just saying that because you're my parents. Like, no, we're not. It's not just us. So let's, you know, Let's see what AI has to say. And it's— if nothing else, it may not give us the answer, but it does often at least give us kind of spark of what the dialogue should be, at least possibly a next step or something like that. So it's been very useful on that front.

You need to give him access to the special LLM that's only trained on great parenting tips from Dharmesh. So anything he asks, it all just agrees with you.

I'll maybe put that in my will and pass it down to him when he has his child or something like that, yeah.

You have a fairly well-known philosophy that you've written about called Sorry Must Pass. And for those who haven't read it, and you should, it's essentially a way of gently letting down all the people who ask you for your time, but who would distract you from the things that are important to you. You've also paired it with a quote from Derek Sivers saying, if you are not saying hell yeah about something, say no. But as AI amplifies capabilities, has your definition of a hell yeah shifted?

I'll say it's shifted a little bit. A more accurate thing might be it's changed. I have this kind of defined set of 3 things that I'm working on. And the things will change out, but there will be 3 things. It helps me kind of narrow my focus. What has shifted, though, is that within those 3 things, the number of things I can accomplish now that I will say yes to has actually grown. I can take on more now as a result of having these power tools and the assistance of AI than— I would've likely— so I can say yes to more things, but I still don't say yes to more than the things outside of 3. So I haven't gone from 3 to 5 things as a result of AI. It's like still those 3 things, but more things will sort of filter in in terms of number of those things within those buckets that I'll, that I'll work on.

That's awesome. You're living the AI dream, getting more done.

Yes, I am actually literally getting more done. I really am.

When it comes to creative or strategic work, the kind that's, you know, it's kind of subjective and hard to measure, What new ways do you see emerging to value and monetize AI's contribution?

Yeah, this is a tough one. I think there's been lots of discussion within the business circles around how you charge for software and how you build products and charge for them. And we've gone from kind of user-based, receipt-based things to consumption-based things to now kind of outcomes, like AI does the actual work and you charge for the work that was done or the outcome that was achieved. My sense is that, uh, although people are excited about this, uh, especially the outcome-based thing, I think partly the reason they're excited is you look at the early use cases. Uh, for instance, uh, customer service is a really good example. This is one of the kind of most common use cases for which, like, oh, if AI can resolve a ticket, uh, you know, the software can charge Y dollars per ticket as long as Y is considerably less than, uh, what the, the human cost, uh, or the manual cost would've been. My sense is that, um, there's a reason why that particular use case, the outcome-based pricing worked, and it's because You sort of have an objective measure. It's like, did it resolve the ticket or did it not? You have an objective measure to some degree of quality. Was the CSAT at or above the levels it was with, uh, human resolution? So we sort of know. And you also had the economics. It's like, oh, we've been solving tickets for, you know, months and years. So we know what the actual cost of a ticket is. So we have all those 3 things. Most of those things have to be true in order to have kind of outcome-based pricing. I don't think that works, uh, for logo design or for— it's like, I— because it's like, okay, well, What's a good versus a bad logo? Do you create 100 iterations versus 5? What does that mean? This kind of classic approach to how do you value outcomes, I think that's still going to be a little bit human-driven, which is going to limit the degree to which you can just charge products based on outcomes. And sometimes, maybe it's OK for a tool to be charged as if it were a tool. It's like, OK, well, you have a human that's spending x amount of time doing whatever it is they're doing. If this can increase that productivity— and this is age-old—

—productivity.

pitch for, you know, software products. But it happens to still be true. It's like, if we can deliver a product that, you know, cuts down costs or makes the humans that are super valuable even more valuable, that allows them to accomplish much, you know, much more, I think that's great.

But we've seen data come into all sorts of decision-making, and you can put a metric behind anything. So logo design, yes, you can put it in front of 100 people, and the one that gets a 95% response is the one that you go with. So you can Measure things like that and then feed that to the AI and say, job done, and then pay it for that output?

You could. Um, so this is another kind of dimension of it, right? There are things where the outcomes are, can be reduced to, like, I'll call 'em semi-fungible things. It's like, okay, let's say the fungible thing is, in this example, is a, uh, logo design that gets, uh, 95% approval rating among some, uh, uh, you know, breadth of people. Yes. But my, my sense is there is a large amount of human activity where the outcome is nonlinear that says, OK, if you produce the one thing that happens to actually have that breakthrough, breakout outcome, that pays for a million of the mediocre, modest, average ones or whatever. And you're not going to actually hit those by simply just solving for the mean or the average or something like that. You look at the publishing industry and look at writing in books and things like that. It's like, OK, well, yes, can AI produce a book? And maybe it will sell. But will it produce something that's actually going to be this thing that, like, that you could not have measured, uh, in a, in a test case or whatever? It's like, it's, I don't know, maybe it will. Um, but I think it's going to be really hard to know a priori that this set of outcomes, these 100 logos that were produced were actually worth $1,000. They're $10. That's the actual economic value of those logos. It's like, okay, well, Yeah. but for a brilliant one, for the right brand, for whatever, it's like, my money is still on the humans. I'm sorry. Uh, humans powered by AI, if they want to use AI, Fine, but I want someone that's got the taste, that sort of have that has that certain je ne sais quoi that yeah, that's well. I think that's the takeaway.

My money is still on the humans. That's everything we need to know. Is there anything that keeps you up at night when it comes to AI? What's one thing you think everyone needs to be doing more of now to prepare for the new age?

Yeah, there's two things that keep me up at night: one on the light side and one on the dark side. We'll do the dark side first. I'm not a like a. Doomsday thinker at all around AI in terms of like it's going to overtake the world and we're going to cease to exist as a species. That I don't like worry about all that much. What I do worry about is like on the kind of consumer side of the fence, uh, AI is going to become sort of indistinguishable in almost every modality that there ever was. It'll be able to write like a human, speak like a human, produce videos, produce images that are indistinguishable, uh, from real life. In that kind of world— so I'm not worried about AI taking over the world or doing really bad things. I think that's possible but relatively unlikely. What is possible and more likely is you have malintent humans, as we've always had, that now have power tools that they didn't have access to before. And they can actually wreak more havoc than they would have been able to do. And I think the answer to this is just more education around Yeah, just how we consume AI. Like, for my parents and everyone's parents to sort of know, it's like when you get that call or something like that saying, you know, your son's been in an accident, whatever, like, okay, you sort of need to know what's possible now with this technology, whatever. So, you know, we don't get taken advantage of and things like that. So I think we just need to raise the level of kind of education. And people need to understand how it works, what it's capable of. On the lighter side of it, what keeps me up at night is that This feels a lot like the internet back when I was first starting HubSpot, which is back then it was like, oh, there's this thing called the internet. Millions of businesses and business owners and entrepreneurs should be benefiting from this thing called the internet. And very few were. And the problem was they just didn't understand it and the tools weren't there. And it's like they just couldn't kind of bring it together. And that was sort of the genesis for HubSpot. I'm having that same sort of kind of feeling in a positive way. It's like there's this new thing called AI that millions of people and businesses and entrepreneurs should be benefiting from. Not enough people actually are benefiting. Yes, they'll use ChatGPT, they'll type in a prompt, but they really are not using even 2% or, you know, 5% of what AI is capable of. And I think that needs to change. So that's partly the motivation behind simple.ai is like, okay, demystify all these things going on, try to put some simple tools out there. HubSpot's doing a bunch of good work in terms of kind of democratizing access to AI for hopefully millions. That's the part.

Yeah.

I'm optimistic about, but it does keep me up at night, but in a good way.

I love the light side, and thank you for going deep on me. I want to change pace just a little bit. I would love to know who you would have on your all-star creative team. You get 4 seats. You can pick anyone from any era, any universe, dead or alive. Put together your dream creative team.

So I'm going to pick people that are alive because I like constraints, um, and also because it's like, oh, well, maybe I have a shot of working with these people someday. Maybe they'll hear this. Uh, so I'll put— since we've talked about, uh, music before, uh, Jacob Collier, who's a singer-songwriter, multi-instrumentalist. Oh, nice. He's amazing. I think what Jacob Collier would be really great at is, like, deep down inside, he's a product guy. I'll say it right here. Uh, like, if you watch his— he's, you know, won lots of Grammys, produced lots of albums, done live shows. He solves for the customer, he solves for the audience, but he brings himself into it. I think parachute him into your company, to my company, to pretty much anything, and I think he would be like a top-deck performer, whatever the thing happens to be. I'll say Jack Butcher, who's a visual designer. He does Visualize Value. There's a really good tweet. But what I love about Jack Butcher, you should look him up, he's one of the best that I know of what I think of as like visual abstraction. And like, my life is about abstractions in differing ways. in differing levels. That's what software engineering is actually all about. And he does that with visuals, be able to take a concept and reduce it down to a monochromatic, like just lines and circles kind of image that sort of captures the essence of that idea, right? And he is world-class at that. So I think he would be awesome. I'll say Jason Fried from 37signals/Basecamp.

Okay.

Awesome entrepreneur. I've known Jason for a long time, but I've never worked with him. But his One of his many superpowers is he does idea distillation, idea abstraction really well, right? It's like he can take the concept, express it visually in terms of product design, but almost more importantly, he can reduce it into words. He's an exceptionally good writer in terms of conveying the concepts and ideas in his head. He's written lots of really good books, all of which are worth reading. I think he'd be awesome. And the last one I'll say, because his name doesn't start with J as the first 3 didn't break the pattern.

Sure.

I'll say Andrej Karpathy, who works in the AI space, and he does what I think of as technology distillation or complexity distillation on the technology front. So he will describe complicated AI concepts in ways that mere mortals like me can actually understand what the kind of mechanics of a large language model is. And obviously, he's a practicing software engineer, deep on AI research or whatever, The way his creativity really shows up is that ability to kind of make things simple that are actually really, really complicated. So those are my picks.

That is a fascinating team. We're now turning this podcast into Tinder for product people. So let's put the call out. If anyone knows Jacob and can hook them up with Dharmesh to make the most amazing music/marketing product all brought together, please make it happen. Okay, I think Paolo has had enough time to feed your sketch into AI and work with it just a little bit. So let's bring him back to see the final results of this Paolo-Dharmesh-AI-human creative triangle. Paolo, are you there?

Hello. I've been like so excited to share like the new generation.

We are excited to see it.

Awesome. Hey Dharmesh, I love the story because I could relate to it like a lot. Like my grandmother gave me like a guitar when I was younger and I also just had to like learn From scratch and stuff. I've also been obsessed with old-timey cartoons lately. I started generating like these old-timey cartoons. I just ran with it and had so much fun, like listening to your stories and even like hearing about your son who's a writer. So I wanted to incorporate like those little like fantastical elements in it. I ended up with like this, like short, a bunch of clips like just put together of you like learning how to play like the keyboard. And just getting older and learning how to play more. But yeah, that's, that's pretty much it.

That's brilliant, my friend. That is awesome.

I love it, Paolo. I love the development of the narrative and seeing your brain in action being fueled with AI and what you ended up with, just telling a great story about Dharmesh's experience flowing into the next generation. It was really beautiful. Thanks for doing that.

Thank you so much. Thank you. And thank you for having me.

Thank you for joining us. Well, Dharmesh, it's almost time to close out the episode. And we've got some rapid-fire questions. Are you ready?

I think so. I'm not fast on my feet, but yes.

We normally put 60 seconds, just for you, I'll put 65 seconds on the clock. Let's go. What are you reading at the moment?

So right now I'm reading Superagency by Reid Hoffman, which is a book about AI and like and what it's actually capable of. Like, what would happen if everything went right kind of thing?

Of course you are. What's one prompt you overuse?

Tell me about XYZ person. I do it in Perplexity. That's my kind of tool of choice for those kind of search-based prompts.

What is the first thing you built with AI?

The first thing with what we would now call AI, generative AI, was that ChatGPT before ChatGPT, which is I had GPT-2 access and built a little chat app that I use and my son used to just run little chat things through.

Always a pioneer, Dharmesh. What's your favorite underrated AI tool?

The one I use the most right now is— it's a new tool. It's called Willow, Willow Voice. And effectively what it does, it's like voice input for everything on your desktop. So it's a desktop tool. You can just like hold a key down. You just say things, whatever app you happen to be in. And it goes beyond just like transcription. It will actually be app-aware and sort of format things like, oh, well, you said this thing. But you're writing an email, so I'm going to format the text in this way. Anyway, so Willow Voice is the answer.

Sum up AI in one word.

Power tool.

You and AI are making an album. What's the genre?

Bollywood meets the West, so we'll call it BollyWest.

Oh, I can't wait to hear that. What's one job AI should never do?

Lead. So our jobs.

And finish this sentence. A surprising thing AI taught me about myself is—

I have the same limitations as LLMs, which is I have limited context window, I hallucinate, and I still struggle with writing dad jokes.

I'm hoping you hallucinate legally. So that's it. That's all we have time for. Thank you so much for diving deep with me on AI agents and creativity, Dharmesh. It has been an absolute pleasure to chat with you.

Thanks for having me. This was a lot of fun.

Thank you for tuning in to today's episode of Prompted. That was an absolutely engrossing conversation with Dharmesh, and some of the things that stood out to me were firstly, how agents are already becoming powerful tools, though the jury is still out as to whether they'll become fully fledged teammates like many people are predicting. I was also really interested to hear about Dharmesh's idea that culture is a product you build for your team, which really emphasizes its role in creating a successful work environment. Finally, what we're all here to learn about— creativity. This notion that creativity is about connecting the dots and the powerful way that AI can serve as a valuable partner in making those connections was one of the best insights. To keep up with future episodes, please subscribe wherever you love to listen to podcasts or head on over to YouTube. And if you're enjoying the show, please leave us a rating or a review. We read every single one and we always love hearing your thoughts. Until next time, keep chasing that creative spark.