Ep. 37 - Platform First with Karen Ng | Devs:After Dark

In the early innings, it was like, oh, is AI gonna take your job? Is AI gonna replace you? And like, I don't think so, you know, 'cause it's not replacing the thinking, but it is gonna replace—

Hello and welcome. This is Developers After Dark, the show that prompt injects joy. Into your LLM of life. Today we have an amazing show for you, but before we get started, I want to do a little bit of housekeeping, a little bit of cleaning here, and I just want to tell you that the only website you're going to ever need is devsafterdarkpod.com. That's devsafterdarkpod.com. That's where you're going to be able to see some of our back episodes and see a little bit about any news or any information about the show itself, a little bit about what it's about, And then also, if you want to see all of our episodes, you can go to the HubSpot Developers channel on YouTube. It is amazing. It's not just us, so I don't want to brag and say that's the only thing there because it's definitely not. We are a small part of a larger kind of gathering of just information, newsletter summaries, all kinds of tips and tricks. So you can get lost there for a good week just watching videos. I recommend you do. And while you're there, also like and subscribe everything that you're watching. So you're definitely going to want to do that. Now, uh, today is going to be something very, very amazing because we're going to be talking about platforms, we're going to be talking about AI. And one of the things I want you to do before we get started as viewers is I want you to go to developers.hubspot.com. And I'm gonna throw this up here real quick for everybody. This is going to be the place you're going to go to create tomorrow's solutions today. This is the platform that you are going to want to be at. This is the platform you're going to want to participate in. You can create an account if you just see up in the top right there. And then there's also our survey you can take too for our 2026 roadmap. So I encourage you to go and do that because that is instrumental in what we do in the future. And plus, you're going to be ahead of the game as far as developing on one of the premier platforms. So for an episode like this, we thought there would be no better person to talk about both unified data, AI, and more than the person who actually said that quote. So, without further ado, I would like to introduce Karen. Karen, why don't you tell everyone who you are and what you do?

Hey everyone, it is, it is really exciting to be on the show. It's so fun to see Developers After Dark. I'm Karen Ng, and I'm an EVP of Product at HubSpot, so I run product here. I've worked on platform teams for a long time. I've run programming languages, C#, Visual Basic, and F#, at Microsoft. I've worked on Kotlin and the Android developer platform at Google, but I love developers, platforms, and building ecosystems.

Love it. One of the big things, just to get us started, you know, one of the big stories that came out of Inbound was the AI connectors that we have created, right? So that would be either ChatGPT, you've got Claude, Gemini. You know, as one of the people leading the charge in the creation of these things, can you give us a little bit of like a I guess you'd call it behind the scenes or insider baseball. There's a million different ways you can phrase it, but kind of a look on how that kind of worked with regards to like the building and the design.

Yeah, I'll give you 2 contexts first, which is, you know, if you think about like this cloud era with microservices and services, APIs powered all of that. And that was the rise that we saw in REST APIs. When I looked kind of ahead to what does an AI ecosystem look like? What are, how are agents and assistants going to talk to each other? I, we kind of knew, or one of the things I really looked at was like, there must be a standard or protocol at some point. And this was predating even MCP, or maybe very early days MCP before it was adopted by everyone. And so that was one key thing that we were like looking at, which behind the scenes is MCP, when it starts getting adopted by all major players, and the one we were waiting for was kind of OpenAI. We, I've seen programming languages and standards come along a lot, and We knew that was going to be a thing. So that was one thing, which is like, that's going to be important. Second thing is what's happening in this LLM or AI OS world. And we call it internally AI OS because if you think about all of the major technology shifts that have happened kind of in our lives, lifetimes, they were cloud with browser-based wars. They were PC with PC-based wars. There were mobile with mobile platforms. And I believe that AI is going to be the same, where it's going to be an entry point into how we do work. So, if that's true, then you kind of want to be an app or how things are done kind of in that ecosystem. And so, that was the beginning of why we thought it was important to do contacts, because HubSpot has a lot of unified context about companies, contacts, businesses, the history of deals that people have worked on. And we wanted that to be accessible to all of the different connectors. So that was the beginning of it. And I combined it with MCP because as we thought about building it, we said, how is this gonna— how is this gonna work? So, you know, we built the connector to ChatGPT with HubSpot. It's based off of a remote MCP server. We looked at how to build that quickly. That was built in about 2.5, 3 weeks, you know, OpenAI engineering team.

Crazy.

And we had to look at different options of do you build versus buy? Do you use something turnkey? It has to work with our infrastructure and system. It was really important that user permissions are respected and that we were able to create an OAuth, like, kind of handshake on top of that. So we built our own. But that was kind of the context for, if we want to be— if we have such great unified data, private and public context, we want our customers to be able to use that in all places. And if there really is an AI operating system coming about, we want to be in the heart of that. And that was the beginning of Connector.

One of the things you mentioned with the unified data, there was a statistic that kind of blew my mind when you had mentioned that. I think it was on a LinkedIn post, but that it was the, what, 20% of customer data flowing through their business. They only see 20% of it. And that to me was like absolutely kind of, it was kind of shocking at first because you would think with all of the, and even before like the the AI push, you would think with the technologies that we had that that number wouldn't be so low. Um, now the unlocking the insights from the data is definitely important. Um, now that especially with the ability to do so now, the question I would have is like, how are we thinking about next year? How are we thinking about the future of that? Because we're, like I said, we're in that era now where the technology and the ability to unify it is in HubSpot now. So what's What is going to be the next step? Because that 20% is like, nobody should have to be running a business not knowing where 80% of their data is.

Yeah, yeah, I agree. Like, you know, that stat of like 80% of your data is kind of like trapped or locked. Like, think about a CRM today. You're collecting email. That email is like a blob of text. That email might have a PDF of a sales quote that you sent that's like not cracked into. You might have a call like this, and those transcripts and recordings Like some of those are kind of unlocked, but they're just really, it's just a pile of gold. And so the first part of that was how do you unlock all of those trapped insights? And the way I think about it is, um, CRMs are traditionally incredibly good at structured data, you know, objects, models, systems, contacts, companies, tickets. Those are all just objects.

Mm-hmm.

Unstructured data doesn't look like that. It's more like a JSON blob, or like as humans, we look at a screen and it's like we can process all this visual data, but we're really breaking it down into structured objects that we then commit to memory and context. So that's how I thought about our unified data and unstructured work, where you're really taking the unstructured data that could be blobs of text, videos, streams, audio, and then you're actually extracting it out and you're converting it to structured. So what we do with unstructured context at HubSpot is we're actually pulling out transcripts and saving it back into, okay, for a sales call, If I knew that the best methodologies were MEDDIC or BANT, that's like number one ways that you might be able to do a sales call and run it, then I know I need to know who the champion is, who the economic buyer is, who the, what the risks are. And those are structured things. So, the first set of what we did was pull out unstructured data into structured properties and contexts that allows automations to work off of it, events to work off of it, notifications to work off of it. But where we're heading with that is once you have all that data, like kind of in a way parsed into structured data, what are you going to do with it? And so those became very use case specific, which is if I wanted to build an AI sales assistant, it has to be the best in class at being a sales rep. If I wanted to build a customer service, you know, insights, I want to know about retention and churn. If I want to know about retention and churn, I now need product analytics. So that's where we're heading with it. And I call that context, which is the first part was breaking unstructured to structured. And the second part is truly building a context layer, a context layer that keeps things up to date. So I don't just pull it out once, it's updating live. A context layer that is proactive. So it's notifying me when people change jobs, when somebody had a new Series A raise or raise, proactive context. And then that context layer powers everything else. So, if someone comes into a website and asks a specific support question, that salesperson who's working on that account should know that that happens. So, those are the 3 areas I think about with context. Unstructured, turning from unstructured to structured, and then really a true context layer. that is unified and is proactive, that stays up to date, and then it powers a bunch of different scenarios across.

Would that context layer stay scoped to HubSpot data, or would you imagine extending it beyond that? Like, for example, market trends, you know, like how things are happening outside of that specific set of data?

100%. So, you know, when I think about like what makes actually us as humans intelligent, we're like always up to date. I, I did an inbound talk with Olivier on OpenAI, and one way he kind of framed sometimes the models of like mini versus not was like, think of an intern genius in a room. This intern is incredibly smart, but if it knew nothing about street smarts, it's not that intelligent. You know, it's like kind of my interpretation of it. So anything we build that has context needs intelligence from the external world. So we built something called Smart Properties, our data agent today.

Mm-hmm.

Data Agent is the way to get external context. And so you can create a smart property. You can tell Data Agent, I want to know about market trends, filings, competitive reviews on products. And that agent continuously runs and brings into the context layer that external knowledge.

So, yeah.

The data nerd in me is just like geeking out right now because of all of that, because I think that Like, when I look at this, a lot of people have said like, oh, it's going to take a job away, right? But like what you're describing is an aid. It is like an empowerment of your role to be able to do that because like there's going to be people out there that don't, that can't like quickly go and grab those market trends or know exactly where to look for those market trends for specific niches that they're in. And as the way you're describing it is like you literally have your own assistant who's Helping train you to be better at your position in the things that you may not have the skill set fully wrapped up for. But at the same time, it's providing like all of this data. And like, to me, like, that is so empowering as a salesperson to know when I step into a meeting, I'm not walking in blind or I'm not, you know, half— I have half the information I need. I'm getting that power to be able to streamline and go further because like the same thing, like the automation nerd in me is like going to like level 30 right now in my head. Of like how this can all interconnect. So this is fantastic.

Yeah, I nerd out on data all the time. Like, this is exactly, it's so fascinating, right? If you can take unstructured data, put it to structured and get proactive context, everything you've ever done with automations and workflows, you can build that.

Mm-hmm. What's, yeah. No, 'cause I remember, 'cause you were mentioning like, AJ, like the, it's not going to replace, 'cause I can't remember the exact phrase, but Dharmesh had said something in inbound too where like AI is going to do these 3 things, but it's not going to replace your thinking. It's going to enhance it. It's going to refine it, but it's not going to replace it. So I think that's going to be huge. And then I also wonder too, as things start to roll on and the context for this data comes both internally and externally, what is it going to look like with a— probably within the same vertical, obviously, to make a fair comparison— what is it going to look like with a company that runs like that versus one that's kind of like, maybe that's not for us? for us yet, or we don't know how to like make that happen yet. Just like the, you know, like, like I always think of it, and obviously Dennis will laugh, but in terms of real estate, like if you had 2 real estate companies, one who was able to have that sort of unified data and one that doesn't, like the performance of the company overall, like what is that going to look like? Because I imagine it'll be just as kind of the scale there will be, one will be light years ahead of the other.

So.

Yeah, yeah, I totally agree with that. Like, you know, I think in early, early innings of, I would say like this, this version of AI was what after ChatGPT release, it was when AI became so accessible. And in the early innings, it was like, oh, is AI gonna take your job? Is AI gonna replace you? And like, I don't think so, you know, cuz it's not replacing the thinking. Um, but it is gonna replace the teams that don't keep up. And it is hard to keep up.

Mm-hmm.

Because if you had, you know, a real estate company that could tell, like, let's, let's say for real estate, right? Like if you're a buyer or seller, you wanna know comparables. And then you don't, you want, you don't want to underprice and you don't want to overprice. Um, but like a company that like kind of knew all historicals in all neighborhoods, even when you cross a street, like sometimes neighborhoods are like that where like there's a big difference entering a different neighborhood market. If I was a real estate agent equipped with that, like we, we'll do better than the ones.

Yeah.

That's amazing. Yeah. That's, that, that's near and dear to my heart. So that was that. That's, I love that. I absolutely love that. As a little bit of a selfish question, you know, as a platform-first leader, which we all consider you to be, whether you like it or not, but, you know, why is developer relations so important to the success of a company that wants to build out a platform and be platform-first?

Yeah, developer relations has also been always very, like, near and dear to my heart. Platforms, dev rel is like the connective tissue between people and like product. And if you truly want to build a platform, it is actually less about the features that you're building, like the APIs or the features that you're building, and more about the ecosystem of builders on top of it. And those ecosystems of builders, it's how do they, you know, I look at dev advocacy of like, how do you help people be builders, trust the platform, see what's coming? And so like, to me, that's the connective tissue. And like, It's not just about the platform product you build, it's about the people that like work around us and build on top of us are able to create businesses and lives. And DevRel for me is that connective tissue, both out, which is like, how do we tell our stories around, okay, how's the platform changing? So like when we did UIE or like customization on CRM, like it's the first time we released that. It's going to be hard for people to build, like where are the samples? Where are the docs? Where are the bootcamps we might do on it? And then it's also a signal in for me. Which is, you know, you're talking every day to people who are building on our platform. Like, where are the gotchas? What hurts? What APIs are missing? Where are people doing workarounds? And like, for me, that's a signal as a platform builder, like on product, like where do I need to build out so that there needs to be more capabilities?

Yeah. Double down on that. Cause I just want to say to everyone listening to this right now, we do actually pay attention to what is being said in the community.

Oh yeah.

Yeah, I could specifically speak for app cards where it was like the rollout of that. Everybody, especially the external developers, were extremely appreciative of the way that we were able to communicate with them, problem solve for them, and almost educate in a way where it was like, here, let me show you there's more than just— I see what you've done, but these are all the other things you can do as well. And just kind of pushing that from not just our team, but every other team that helped kind of make that happen. It was, it was kind of one of the, the nicest kind of stories, like mini story inside of the bigger story that I saw happen like firsthand. And it was just, uh, the sheer happiness that some of these developers had when they were like, hey, check it. Because some of them would even message me on Dev Slack and be like, hey, uh, you know, I'm working on this thing, can I just show you this? This is like super cool. And I'm like, yeah, of course. Like, I'm not gonna say no, I'm into it, I want to see super cool stuff too. But I was like, yeah. But they were just, they were, they were so more excited are so much more excited just to show what they were doing than they were worrying about like, oh yeah, yeah, I gotta submit this still. It's like, yeah, don't forget, you wanna put this on the marketplace, right? So it was like one of those things where it was like, yeah, just the sheer love of it was just, you could just, it was palpable. You could definitely tell. So, but no, I, you know, being in DevRel, obviously I'm kind of biased, but yeah, I think it's fantastic because—

I think one of the really cool shifts too, and you've said it multiple times now, Kieran, builders, right? This idea of developers used to be like, you're just behind a screen coding 24/7 doing all these things. But the definition of a developer has expanded a lot now since the enablement of AI.

Mm-hmm.

So we have like builders and we have developers and we have all of these people, right? And I think that that's also just like this area where Debra also kind of comes into play because I, I mean, I was pretty close to the launches that happened with OpenAI and Claude and doing those. And like seeing the prompt crafting and the, the things that like we're trying to like convey to other people, like, hey, you don't have to be a developer to build the perfect thing to grab your insights that you want. You just have to have the imagination and the drive to basically do it. And I think that that's been like such a cool story that has also been told through a DevRel lens of not just focusing on developers, but builders who are developing as well.

So.

Yeah, absolutely. Yeah. That trend, it's, it's funny, like I feel like when this wave of AI happened, I thought it would unlock intelligence. It's unlocked creativity. It's unlocked soldiers. Like, it's made people who probably would've been, you know, had a hard time, like the whole vibe coding and like creativity. It has just really unlocked all of that, which is really exciting. Um, but to the Connectors work, like just kind of some of the behind the scenes, I remember when we started it, like I, I sat there and I was like, okay, We can only release like 3 types of objects, like contacts, companies, and deals. And we couldn't even do tickets. And I was like, well, we can't do associations either. So I remember like using, I built like, I first led the prop library, which was like, can we do anything useful in the sales or marketing space when you have no associations?

Mm-hmm.

So I remember pointing to all of the API docs we had and then saying like, what are the things I can do? Like, can I do? analysis on largest deals closed of people coming in, in these types of verticals? Can I ask these questions? And so those were the initial set of how the prompt library came about that AJ, you were very close to it when we shipped part of the prompt library on the website. That was the challenge, which was like, can I help match prompting and what is of the real value was like the object layer. And like, it's the connection of the objects we were able to do and what was valuable. Associations can give you so much more. So, it's great that we got there.

Yeah.

The definition of developer has just dramatically changed just with that. I mean, just when we first started the show, like, it was really gonna be just like telling the developer story. And then after, you know, what, a couple months at the very least, it was like, well, A developer isn't exactly a developer anymore. Like there's a complete, so yeah, I, I, it's one of those rides that's remarkable and just so fast that if you blink, you're going to miss it. And in fact, even it affects the name or the tagline of a podcast. Like that's how, that's how it's like, it goes beyond, right?

So, spicy take too, though. Like, I feel like there's things we learn in this whole age, right? At first it was like prompt libraries and then everybody learned hallucinations and like how that's doing. Now I feel like we're in this world of context engineering and like compaction. So, like, how do you— you run out of token window. So, like, how do you tell, you know, even AI to, like, pack it all in and, like, pass it over? And then I think there was this rise of vibe coding where it was like, oh, do we even need engineers anymore? Do we even need developers? And I feel like we're coming out of that cycle, which is you can't build production software still just on vibes. Like, it is incredibly hard to debug.

Mm-hmm.

Incredibly hard to have scalability. So, You know, I, I have a, my, my background is like in programming languages. And so one of the things I forced my 12-year-old to do is learn C# since I worked on C#.

Wow.

So he had to learn C# since he was like 8. And I'm like, is that still a valuable skill? And I'm like, it, I think it actually is because you still, it's, it doesn't replace the thinking. And so like understanding how systems work or how code works, I think is still like really key. Even though you can go so much further now with like live coding.

So, right. Because there's, there's like sets of problem solving skills just within code debugging themselves that isn't just like specific to one particular. So yeah, I would imagine you get those basics down, whether you're using C# or something else, that, that's still, that transfers. Yeah, I have a 12-year-old as well. I'm going to have to get her started on PHP or something else. Change the curriculum. Get ready.

If you start too early, they like, they hate it forever.

So that's my other lesson.

Don't start too early.

That's awesome. Um, let me, let me kind of, uh, transition into this because this was, this was one too where, um, there was a milestone that we hit a couple weeks ago just on the App Marketplace alone, and that was the 2,000 apps and then 2.5 million— and that's million with an M— active installs, which by itself is a huge feat. Now, you had a LinkedIn post actually where you mentioned the milestone representing something a little bit bigger, and when I read this, I really liked I really like the way it was kind of phrased, but you have thousands of entrepreneurs and developers and teams that saw this opportunity to build after, and like you said, you've had previous experience with building ecosystems, like that type of success. Was there a pattern that you saw emerge with creating that success that one would, let's say you're at a new place and you're trying to build out an ecosystem?

Yeah.

Did you see any patterns?

Yeah, I, I, I definitely do. It's funny, before HubSpot, I, um, actually we, I, I worked on a startup called Comroom and Comroom's like initial take was, can you build recurring champions in communities? And so like one, that's the history behind Comroom, or initial take was, this is the pattern I saw in platforms and ecosystem. You need one part, which is the platform itself needs to expose enough things that builders can build off. So when I was at, when I came to HubSpot, it was like, do we have the right APIs? Like we did not have programmatic access to some data. So you need those. You also need like surfaces. So like for us, and as a CRM company, it was the cards or like some ways you customize. So you need enough of the product pieces that you could unlock, and then you need people to actually build value on it. So, and those are the creatives. So like, it's the people who are innovating and creatives and like in every platform I've seen, people surprise you with how incredible what they can build off of it, like ideas that you never even thought of. So that's the second piece that starts happening. And then you start getting champions. And so when champions in ecosystems, like, uh, what did you call this? Inject joy in, you know, our LLM of life. When you see that injection of like—

I'm gonna trademark that.

It's a great, it's a great quote. But when you started seeing those sparks of like joy, like, this is fun, I want to keep building this, or I can build a business off it, which means I can actually make money off of it, then you start seeing a cycle where like they get excited, they build more things, then customers on the platform who are not developers start using their things, they find value. So like in our case, CRM customization, CMS templates, like, you know, I worked on mobile and like it was apps. Apps were really important because if you have a phone with no apps, The phone sucks, like literally can do nothing. So then you have to—

Oh wait, my, my Razr flip phone.

In this way, I gotta take this.

Hold on.

So that's, that's the cycle. And so ecosystems thrive when you get a self-sustaining cycle. So that is always what I've not like looked for, but I've tried to think about how to build it. So like, that's coming back to DevRel. That's why it's exciting. You open the platform, but you need people to build off of it. So like, we didn't have a Slack channel. Like, how do people ask questions? How do they get samples? How do they get started? And then how do you get user groups? And then if we didn't have, like, we have CAB, we have like, um, Partner Day, then we don't get to share, like, here's where the platform's going. And so there's no trust. And so that self-sustaining cycle. that's when an ecosystem for me has like made it or like is successful. It's when it keeps growing like this. And it's not about a platform person pushing platform. It just keeps going because you have people who are finding value in building it, customers find value in using it, and then it keeps going. So, self-sustaining cycle.

And the best kind of marketing is when your developers can market for you. So, it's like, it's a total win all over the place. I know I keep going back to inbound, but that's obviously because inbound was huge and there was a lot introduced and there was a lot of like really, really useful information there. But you talked about, and you did talk about studying machine learning, right? At MIT and part of your inbound talk because of the impact that you saw as far as like what it could do, right? And then you wanted to build, and correct me if I'm wrong on this, but you wanted to build the expertise, but the technology wasn't there. Which reminds me a lot about— and to be honest, and you don't have to credit me for saying this, but you're kind of like a superhero because like Tony Stark's dad did the same thing. He wanted to build the arc reactor, but he couldn't. So he had to wait because the technology wasn't there. So I mean, you are a little bit of a superhero just in that respect.

Wow.

Wow. Wow.

I was like, that would be my dream to build like, not arc, but I want an assistant like Tony Stark has.

Yeah.

But you're—

but we're seeing—

yeah. So that's guns.

What was that, Dennis? All right.

Nick just wants to keep his job here. He's really, really sucking up as much as he possibly can.

That's usually what Dennis accuses me of, but it's not necessarily true. Not necessarily. But no, but we're seeing that impact now. So like the technology not being there before, but we're seeing it now. So like if I was an engineer that was interested in like moving into product, what kind of advice would you give? Because, you know, there's definitely a huge difference between you know, where it came from, you know, not having that technology, but having the knowledge to know that there's a possibility, to now where it's almost like it's breakneck speed as far as like the things are moving. You know, if somebody wants to get into, you know, being an engineer in product, what kind of advice would you have for them?

Yeah. You know, my biggest advice is like, you know, we use it here, solve for the customer, but the hardest part about product sometimes is knowing what product to build. And so like, in a lot of ways, like my job is a pattern match. Like I look at market trends, I look at competition, I look at what I think, you know, I watch a lot of videos, I read a bunch of stuff, I get customer feedback and I'm trying to distill all those patterns into what customers, like what problems customers have. And then some of them might tell me how to solve it, but like, that's, that's, it's actually my job to figure out like, how do I take a problem space and then figure out what solution to build? So like, if you started today and you want to like vibe code an app, or if you're, you're building an app in AI age, what would you build? That is actually a hard problem.

Mm-hmm.

And you know, like, if you think about your own life, like, I often think about solving things, um, that my life has. Like, I try to find ways to try to give myself more time. So like, as a, as a parent, like, I miss like every school email that comes. So I'm like, I have an agent that runs in my personal life that tries to like comb through all the email. That tries to give me a little widget, you know, that, that tells me what school events I'm missing. But that's like what I would advise, like knowing what to build and how to build. Um, it starts from a pain point or it starts from an opportunity. And those are the 2 kind of dimensions. And then figure out like, that's what product craft is. So like sometimes it's too saturated or sometimes like you'll try to solve a problem that like only you care about and nobody else does. So, you know, it doesn't make a great product and market, but that's what eBay buys. And then by the way, my, uh, I'm so like the, the Tony Stark arc. My children are named superheroes. So I'm like a big, what could be person, like epic. So my oldest son is, uh, named after Top Gun and Wolverine, Logan Maverick. My youngest, much more like responsible. So Parker Maximus Gladiator.

That's awesome. I love that. Fantastic. I'm gonna steal that with the school emails thing because I'm constantly— my wife's like, didn't you get the email? And I was like, oh yeah.

Right, right. Yeah, now I cheat and I'm like, oh, okay.

No, I mean, it's productivity. I love that. Is there something that surprised you the most? Because we're talking about this whole kind of like this jump where it's like, okay, I know this can be a thing. We're just We're not there yet with the technology. And it's like, oh, technology's here now, but now there's 50,000 different options. Like, is it, is there anything in that span so far that's surprised you the most about this evolution?

Um, well, it's, it's happening faster and faster. So how quickly it is since like the beginning, like pre-GPT, ChatGPT is like the moment where like it changed because it was, that's how accessible it was. Um, I think we're still like, you know, things are getting better and better constantly, but they're like not there yet. Like you have to chain together things in your current workflows to make them work. Like, I think this, this wave of the browsers that have assistants are like very interesting. Like everyone's released one now. Um, turns out you live our, we live our lives a lot in the browser. Like if you're on Gmail and if you're on websites, you're on enterprise SaaS, HubSpot is one as well. So like that assistant in that workflow.

Yeah.

can now take actions. And like, that was one thing that it's, the technology is there, but you still have to piece it together. The ones that will figure out like how to bring it into your core workflows, I think that will matter a lot. People chain together a lot of tools still.

Do your kids do this? I mean, are they operating differently? Are they like, are they still browser-based as well? Do you see that as the future there?

Uh, I don't, I don't think browsers will actually be the future. Like I look at like this AIOS concept, like. Man, I wish I could, like, I, I can, you know, my whole house is like automated in different ways, which means it's broken like 20% of the time.

Yes.

I can't figure out how no one, no one who lives here, who's like a guest here can figure out how to turn on the lights. But, you know, we'll get there. I believe in the technology one day. Um, but like, I think, um, um, like with my, with my kids. Yeah. Sorry, Dennis. Going back to that, that kind of question. Like, I think, I think it will, it won't just be browser-based. They are browser-based today, right? In schools. Um, where I live, it's Microsoft Teams. And then they have a lot of, you know, they each have laptops and iPads now. And so like a lot of that work is there. Um, they're not allowed to use AI yet. So we'll see how some of that stuff changes. Um, how some of that stuff changes.

Yeah.

There's been some introductions here, at least in Illinois, where at some of the schools where it's, they've teased and had them again, it's, it's more of like refining the thoughts. It's not replacing, but like, You've done the work. Now see, let's see what this, and of course I'm sure they're vetted, but like see what this can do to help you kind of make it a little bit better. So I understand. And then voice too for this. Like I never thought, like especially with my son who he's about like he's 8, but the use of that as his like mechanism to start to like interact with like any type of chatbot or any type of like AI. It's always via the voice versus typing it out. For me, that always felt weird. I was always like, I don't wanna, like, no, I'm gonna type it out cuz I know exactly what I want to, like, I always, I thought that was kind of a weird— Yeah.

Well, I'm, I'm actually seeing that shift too. Like, so we use, um, like voice in the car. So like, and you know, it knows like how old, you know, what would range. So like if we're learning about a topic like architecture, um, it'll give us 3 different ways. So it appeals to all 3 of us if it knows that we're, it's 3 of us are in a conversation. So that's really interesting.

Wow.

And then even for myself, by the way, like Like I've, I've started using both Whisper Flow and Willow, but like, I will do Slack, like, you know, probably in places I'm not supposed to be, but like, we'll use that just to respond. And it's so good at like getting into those pieces. So I expect them to probably use voice a lot. Um, I mentioned my house is like very automated. That means my children, they don't use light switches. They like have little, uh, scenes that they use for each of the rooms, which is, which is pretty funny.

But.

That's, yeah, my, my girls have been doing the same when it comes to using technology. It's interesting seeing kids using it and One, just how quick they are to pick up on it, because that has kind of blown my mind. I was telling somebody, a neighbor, a little bit ago, I was working one time from the kitchen island, and all of a sudden my screen on the TV clicks on and it's why we should get a hamster. I'm looking around and I'm like, what is happening? And my one daughter put together a Google Slide deck because she's 9. They're both 9, they're twins. But like, that's what they were learning is how to use Google Slide decks. Well, I showed them how to layer pictures because of like the graphic stuff that I do. This was one of the best like slide decks I've seen in a while, but it was just crazy to me that like they, they knew to cast it to a TV. Like I never showed them how to do Chromecasting, right?

Mm-hmm.

But then listening to them interact with AI even, or with things, it's, thank you. It's, can you please do this? And I'm like, like part of me is like, it's so interesting to hear how they're treating the AI as if it's a person almost. Like it's become a part of their life and their, this stuff that they do. Like, one uses it every night to meditate before bed. So like, it's really interesting to see how they're interacting, but then like where their brain goes, because they're really big into Minecraft being 9. Like, that's their world right now. And like, they're interested in like creator packs and doing those things.

Yes.

So now they're researching like how to go about building those things and stuff like that. So like, it is really interesting to see the effects of technology and even AI on youth. Right now. And because, I mean, growing up, us, what it was, MySpace was like the closest I had to coding.

And that was like looking at the things that you had. You had MySpace and a pager, and then when the lights went on, you went home.

That was it. Exactly. And then that's why, like, to see where it is now though, is just, it's flooring to see the adoption rate for sure.

Yeah, I know. That's why, like, even like the home automation or robots or like crossing into physical digital worlds, like, I see it, like, I wish it was better than, you know, I wish, I wish it was better on the home devices already.

Yeah.

What I would expect.

Yeah.

Now, AJ, you mentioned, actually you brought up a really good point, uh, you know, the way that your kids talk to AI as like, as a person. Uh, a question I have for you, Karen, is like, there's been a lot of talk just, you know, just recently in the industry about agents as teammates, right? So as, you know, as particular people, because I find I argue with Cursor all the time and then I'm like, I'm sitting there like, why am I like, this is not, you know, just, you know, Take a step back. This is, you know, but like, you know, how will HubSpot kind of make this happen, um, in our world as far as like that kind of balance, that hybrid, if you will?

Yeah, yeah. We have a deep belief that the world will be around hybrid teams, and that is like people like us that are kind of supercharged with context, with, uh, tools that could make you like kind of achieve more. You know, we did that with like Marketing Studio where like we know everybody wants to reach people and like have discovery. Um, one of the things that's really hard about it is like, if you switch channels like LinkedIn or Instagram or websites or blogs, like each format does matter. And so does the metadata on that matters. So like, what if we could generate marketing content in all of those channels for you that follow all best practices? Or there's this big shift right now happening from SEO. So search engine optimization to AEO, the AI ones, like discovered through LLMs. What if we were ahead of the game there? And what if we helped you when you generate that content, have whatever mechanism needed to get discovered better in those channels? We helped you do that. So that's like the supercharge. The other part of what we're trying to do is, um, hybrid teams of agents. So if agents can do work for part of your team, so our customer agent is probably the strongest one right now, which is, you know, 65% resolution rate on support tickets. Great. Instead of having just humans resolve those support tickets, that's an easier one to deploy, which I can say, I'll put a support agent on here the same way that you had, uh, the chatbots on websites. Now agents is just there. So, you know, that's how we're thinking through in a GTM team. Like, what are the kinds of jobs that you, like, as you grow from a really small team to a medium-sized team, like, what do you need? You know, you probably need somebody who's a customer support rep. You probably need somebody who like is pretty good at like prospecting. You probably need somebody who's like, like this data agent that Dennis was asking about. Like, it's hard to keep your data up to date. Mm-hmm. That's based on market trends or things that are coming out every day. Um, I don't wanna do that. I'm gonna have my agent watch, you know, like, right. I don't wanna have to go pull school emails, for instance. Like, I want to know when it comes.

Mm-hmm.

And so that's how we're thinking through our agents, which are, what do you need on teams? What jobs do they do? And then how do we build agents into those most painful areas first and like get those right? And then we'll like have more, but that's how we think about hybrid teams.

I'd love the SEO take on that too, because I think about before AI, like just the learning curve and the amount of time it took just to kind of I don't want to say be up to date because it's— you're never up to date on what's going on in SEO. That's kind of a vague kind of term. But the, the amount of effort it took to keep up with it, now adding AI to it where again speed becomes, you know, 10x, 100x, whatever you want to call it, having the ability to be able to take your content and be able to give that sort of, um, help, I guess you want to call it, or insight For these, like for the content creation kind of trims down that time to learn. 'Cause like SEO itself can be so complicated and now you're gonna throw on top of it a layer of, you know, LLMs and things like that where it's not black box, but somebody who like knew traditional SEO, it sort of is in a certain respect, at least at the jumping off point.

So.

Yeah, it still is, right? Like it's still so early for that field that I think there's, there's no real standard yet and it's still pretty early. I mean, that's why like we, we do a lot of what we do at HubSpot because it's, it shouldn't be this hard. And if, if we can figure out how to take some of that complexity and make it easier, easier to use and faster time to value, like you can get leads through this. Like we added AI referrals, you know, as a discovery source at HubSpot. Like, great. Now you know that, that like if, if that's your discovery source, somebody had it through an LLM, like I can guarantee you. That, that is a more highly qualified, higher intent lead because they already did some research.

Yeah, that's perfect.

Yeah.

Um, now with respect to that, because we're talking about like the speed of the enhancement of AI just in general, like how do you balance investing in what's existing now and then keeping your options open for future? Because I, and I think I've said this kind of like probably multiple times on a couple of other episodes, but I can't watch a YouTube video without getting bombarded with an AI tool, the next AI tool, quote unquote, that's going to build the next thing for me. All I gotta do is sign up for this. How, where's the balance there? Because there's so many things to choose from, but then there's, there's such that need to keep up.

Yeah.

You know, I, I, we, people and teams will probably swing back and forth. Like at the end of the day, AI is a technology that's like, a technology that is here to help us do what we were doing before, and it's unlocking some things that were really hard to do. So for me, like, it kind of goes back to, are we creating the value? It will be really fun and exciting to work on some of the new stuff or have that new tool that promises everything, but how many promises have there been on, you know, personalized email over the years as well? And like, that holy grail has never really truly been there. So. I guess the balance for me is like making sure that we're grounded in customer value and that it's like real people's jobs and lives that like we can make better. So we'll swing some part, you know, into what's new and like we should keep up and like learn about what's out there. But at the end of the day, like it's building practical value. So we'll swing back this way as well. You know, at HubSpot right now, we're like looking a lot at like the current product. And is it living up to the promise and value that, that we want it to be? And so, you know, we're working a lot on existing product quality. Some of those will use AI to get— identify those issues, but we have to make the core product valuable or value to customers.

Now, let's say I'm a— let's say, let's say I'm an experienced developer and I'm in HubSpot now. What are some of the things with regards to like the future of the platform that I should be excited about?

Um, we'll definitely use MCP. I mean, we have our, um, we have our remote server out, the connectors all power that, but there's MCP servers and clients. And so like being able to use multiple tools at once, like learn, experiment, like understand that protocol. You know, I get really interested in other protocols like ADA and others. Um, those are built off of like APIs that are needed as well. And, you know, we've, we've done an incredible job releasing more APIs over the last couple of years. And so what, what's missing? Um, I think one big take for me, like, going forward was, like, the permissions, user permissions matter a lot. Like, we kind of used to think that it was just portal permissions. And for me, like, if you think about agents, if they are teammates, if they are humans, like, they are only allowed to see what they're supposed to see.

Mm-hmm.

same way a human is. And so user permissions became like incredibly important. So I would say all of those are like, spend time in MCP for sure. That one's super fun to see like what you could do. And, you know, we're heading in a lot of different directions, I think with app objects, I think with event streams and like unified events.

Mm-hmm.

I think that's really exciting, but there's like so many capabilities. I think we're trying to build still. And, you know, going back to where I started from, like, that's how platforms grow. Uh, we unlock a little bit, we see what's innovated on it, people become champions, and then you do, you do it again and again and again. So, um, we're in a, like, a very exciting, like, new era. Um, the world ahead are slightly different things. Like, people talk about agents a lot, but, like, agents are going to need permissions. They're going to need access. They're going to need to talk to things the right way. I don't think security and trust and the handshakes are all there yet, but, you know, context is also going to be big. Like, how do you give agents context?

That's where you get the Terminator 2 believers. When you don't have user-based permissions is when they start talking about Cyberdyne Systems. But no, that's definitely something you definitely want to take into consideration. We definitely want to be respectful of time because we know we only have an hour with you, but Um, I, were there any other questions, AJ and Dennis, from like the AI perspective that you wanted to ask? Because I really wanted, uh, for the last leg of this to kind of get to know who Karen is. Like we, we usually like, yeah.

With a buildup like that, you have to ask the questions now. Yeah.

No, I don't, I don't really have any other questions. I'm, I'm excited to learn just Karen's backstory because like I said, bringing her to LinkedIn, I'm, I'm, I was floored day one when I first interacted with you when you were at Common Room. Like, I remember Dennis and I were talking, we're like, oh my God, this is really, this person really knows a lot.

Like, I'm just excited. One of the, uh, one of the questions we love to ask, uh, our guests is, you know, what is some of the best advice a mentor has given you in your career? Because we'd love, like, there's such a variety of answers and they're all like a really good lesson to learn. Do you have one in particular?

You know, I think mine that's always stuck with me is, um, like, where do you find your joy? Because like, we'll all have jobs often that like have different roles. And depending on like, like I've worked at startups, I've worked at big companies, like some of them are more defined. But like, where do you find that joy is where you like, like, I love the craft of building products, but it is because you get to talk to customers and like figure out what a problem is and then figure out like what's the right way to build it. Um, you explore kind of new technologies and like at the end of the day, it's a, it's an altruistic and helpful, like. Am I building a product that matters and helps people? So I would say joy. The other is another one that a mentor has told me is like, there's a book called Discover Your Strengths that I read very early and I think was very helpful because it was like, you don't have to be great at everything, but oftentimes your strengths will align with those joys. And don't be a liability in the other areas and know what they are in your current role, but spend more of your energy and calories on your strengths and don't be a liability.

In the other areas.

No, that's excellent. I love that.

Solid.

Yeah.

What is, um, what are some of the things that outside of, I guess you would say outside that are non-techie, I guess, that you do to like unwind? Like what do you do when you're not being the platform-first leader in HubSpot? Yeah.

Um, you know, around after COVID, I picked up tennis where I tried, I tried to play competitive tennis. Which is like getting ranked and like playing evenings and weekends on matches. Um, my career is heavily declining. However, like, that's one of the fun things because tennis for me is like, it's actually both like a physical challenge and like a mental challenge because you have to like think about where to place.

Mm-hmm.

You have to play out the shots. Um, I'm a big strategy board game player as well, so I like to like think ahead multiple steps and You have to— for tennis, you have to do that like while running and try to hit the ball back.

So that's got to be one of the harder ones to like try to pick up, like one of the more physical ones to try to pick up. Like, you know, you could, you could start with chess, you could start with that, you know what I mean? But like to go straight into tennis, which could be like a marathon depending on the match sets, if you've ever watched that TV, it's like, uh, kudos.

It really is. Like, I know to win— I think somebody, I think a coach told me like to win a tennis match, like you have to be able to do like 43 sprints. And so like, I'm like, okay, I'm not practicing long-term running at all. So I literally like will only practice like 3-minute runs.

That's a lot to ask. That's huge.

That's why I play pickleball. We don't, we don't have to do any of that kind of stuff.

I'll play pickleball too. So definitely up for a pickleball event.

Is there now like tennis and then, you know, with every, with everything else in life, is there a, um, How do you balance the demand with like the tech demand and then with like personal life? Because I, you know, we've talked to a lot of people that, you know, surprisingly there's a lot of different answers because you would think it's either, well, I either do this or I do that, but sometimes it's not so black and white when it comes to trying to create that balance.

Yeah, it's not. It's interesting. I'm so like, I love tech. I grew up in the heart of Silicon Valley, which is probably where I got some of these These, these, these characteristics. Um, I am finding myself needing to focus more on non-tech digital, especially with my kids. So we've been playing like a ton more board games because it's so easy to like be on a screen or try an AI thing or, you know, and I'm, I'm finding that, um, presence is like, that's what's hard to recreate. And like, I don't want my, you know, I don't, I don't want to be the, I don't want the presence to be the thing that like I want that to be a thing that my kids remember, you know? So it's interesting, like, after all this, actually figuring out and finding time. I use ChatGPT to give me suggestions to be more present and like, I'll match our, like, you know, um, I'll match our different, like, things that we like, but then I'll try to, try to be more present, like, in human life.

Just got to be careful with the hallucinations. Pick up the Nintendo Switch. No, no. That's awesome. Um, were there any things that you, uh, you know, before we wrap up everything, uh, were there any things that you wanted to talk about that we didn't cover today that you wanted to let everyone know about from the developer world?

You know, I think, uh, we're, it's exciting to hit that like 2,000 milestone and like, we're kind of entering like a new era for me. Again, it's the, how do we open up and create value? And then what can people do and innovate? And then the thing I feel like I, yeah, I still think a lot about is like, how do we help businesses thrive on this ecosystem? Because at the end of the day, that's what we're trying to do too. So, um, I just think about that platform, but it is an exciting time. And I think, I think everyone should play with, with things, you know, outside of work too, because it's, it's moving so fast. And, um, developers, one of the reasons actually through a lot of my career, I've worked on developer platforms. So with developers is like this set of people are like not afraid that things break. And so it's like almost bleeding edge on learning, and like, that's exciting.

I think that's a lost art sometimes. I mean, I think it's coming back, but the ability to, you know, innovate, fail, and then move on to the next thing, at least I always felt like there was a time period in between there where it was like people were almost afraid to do that. And would it be fair to say that kind of AI with the pace, At least with like using agents and things like that, the pace that you can actually innovate kind of alleviates that, oh, it's okay to make a mistake because I can quickly fix it now. Or I can, you know, somewhat, I mean, there's going to be outliers, but do you see that? Because I've always, I think I've noticed that, but maybe it's me. I don't know.

Yeah, I think so. You know, it's kind of like the growth mindset and fixed mindset thinking. Like, growth mindset assumes that failure is going to happen. And like, it's, it's more about what I can learn at every point in time than it is around, did I get it right? Or did I, you know, I think that has served me well. Like, I come from immigrant parents, like, kind of in the, uh, growing up in the back of Silicon Valley. And it was like this, it is in order to grow and in order, like, life wasn't going to move forward, you are going to fail. So like, I'm looking for the moments, failure moments as learning. And like, it's probably just a different mindset shift.

So. Now, where can, if people want to follow you or they want to learn more about You work in a— if you want to be found. I mean, that's— I always reference that with people because sometimes it's like, I'm good. Just, just, you know, where can people find me?

So, you know, I'll often post like stuff that's on top of my mind. Um, I'm on X too, but I'm on LinkedIn more.

So. Excellent.

Excellent. Um, I always like to, uh, to end with, please, you know, if you're watching this, please make sure to like, and then make sure to subscribe to the HubSpot Developers YouTube channel. And since we've been talking about apps and APIs, and this is AJ's favorite part because this is always where I come up with something off the cuff, but Uh, you know, I've also, you know, no secret, done some vibe coding myself. And, you know, and my app is in fact, you know, dating an API and they're really resting well together. Resting well together. I'm going to repeat that for the community value. Oh my Lord. It just— and ends it on a low note. No, I'm just kidding.

That's the ending recording now.